HADOOP-2506 Remove the algebra package
git-svn-id: https://svn.apache.org/repos/asf/lucene/hadoop/trunk/src/contrib/hbase@608289 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
parent
9ebe18e2b4
commit
b0f524f3d7
|
@ -5,6 +5,8 @@ Trunk (unreleased changes)
|
|||
INCOMPATIBLE CHANGES
|
||||
HADOOP-2056 A table with row keys containing colon fails to split regions
|
||||
HADOOP-2079 Fix generated HLog, HRegion names
|
||||
HADOOP-2495 Minor performance improvements: Slim-down BatchOperation, etc.
|
||||
HADOOP-2506 Remove the algebra package
|
||||
|
||||
NEW FEATURES
|
||||
HADOOP-2061 Add new Base64 dialects
|
||||
|
@ -22,7 +24,6 @@ Trunk (unreleased changes)
|
|||
HADOOP-2479 Save on number of Text object creations
|
||||
HADOOP-2485 Make mapfile index interval configurable (Set default to 32
|
||||
instead of 128)
|
||||
HADOOP-2495 Minor performance improvements: Slim-down BatchOperation, etc.
|
||||
|
||||
BUG FIXES
|
||||
HADOOP-2059 In tests, exceptions in min dfs shutdown should not fail test
|
||||
|
|
|
@ -101,7 +101,6 @@ TOKEN: /** for HQL statements */
|
|||
| <AND: "and">
|
||||
| <OR: "or">
|
||||
| <COMMA: ",">
|
||||
| <DOT: ".">
|
||||
| <LPAREN: "(">
|
||||
| <RPAREN: ")">
|
||||
| <EQUALS: "=">
|
||||
|
@ -126,20 +125,11 @@ TOKEN: /** for HQL statements */
|
|||
| <NUM_ENTRIES: "num_entries">
|
||||
| <ADD: "add">
|
||||
| <CHANGE: "change">
|
||||
| <META_TABLE: ".META.">
|
||||
| <ROOT_TABLE: "-ROOT-">
|
||||
}
|
||||
|
||||
TOKEN :
|
||||
{
|
||||
<SAVE: "save">
|
||||
| <GROUP: "group">
|
||||
| <BY: "by">
|
||||
}
|
||||
|
||||
TOKEN : /** Literals */
|
||||
{
|
||||
<ID: ["A"-"Z","a"-"z","_","-",":","/","~"] (["A"-"Z","a"-"z","0"-"9","_","-",":","/","~"])* >
|
||||
<ID: ["A"-"Z","a"-"z","_","-",":","/","~","."] (["A"-"Z","a"-"z","0"-"9","_","-",":","/","~","."])* >
|
||||
| <INTEGER_LITERAL: (["0"-"9"])+ >
|
||||
| <FLOATING_POINT_LITERAL:
|
||||
(["0"-"9"])+ "." (["0"-"9"])+ (<EXPONENT>)?
|
||||
|
@ -190,8 +180,6 @@ Command cmdStatement() :
|
|||
| cmd = clearCommand()
|
||||
| cmd = fsCommand()
|
||||
| cmd = jarCommand()
|
||||
| cmd = substituteCommand()
|
||||
| cmd = saveCommand()
|
||||
)
|
||||
{
|
||||
return cmd;
|
||||
|
@ -235,7 +223,7 @@ JarCommand jarCommand() :
|
|||
{
|
||||
<JAR>
|
||||
(
|
||||
( t=<ID> | t=<INTEGER_LITERAL> | t=<DOT> | t=<FLOATING_POINT_LITERAL> )
|
||||
( t=<ID> | t=<INTEGER_LITERAL> | t=<FLOATING_POINT_LITERAL> )
|
||||
{ path += t.image.toString(); }
|
||||
)*
|
||||
t=<JAR>
|
||||
|
@ -274,8 +262,6 @@ HelpCommand helpCommand() :
|
|||
| t=<CLEAR>
|
||||
| t=<FS>
|
||||
| t=<JAR>
|
||||
| t=<GROUP>
|
||||
| t=<SAVE>
|
||||
| t=<ID>
|
||||
) { argument = t.image.toString(); }
|
||||
]
|
||||
|
@ -650,96 +636,6 @@ ClearCommand clearCommand() :
|
|||
}
|
||||
}
|
||||
|
||||
SubstituteCommand substituteCommand() :
|
||||
{
|
||||
Token key = null;
|
||||
Token chainKey = null;
|
||||
Token operation = null;
|
||||
String tableName = null;
|
||||
String condition = "";
|
||||
List<String> notInList = new ArrayList<String>();
|
||||
SubstituteCommand substitute = new SubstituteCommand(this.out);
|
||||
Token extendedKey = null;
|
||||
}
|
||||
{
|
||||
key=<ID><EQUALS>
|
||||
{ substitute.setKey(key.image.toString()); }
|
||||
(
|
||||
chainKey=<ID><DOT>
|
||||
{ substitute.setChainKey(chainKey.image.toString()); }
|
||||
operation = <ID>
|
||||
{ substitute.setOperation(operation.image.toString()); }
|
||||
<LPAREN>
|
||||
{
|
||||
String operationType = operation.image.toLowerCase();
|
||||
if(operationType.equals("projection")) {
|
||||
List<String> columnList = columnList();
|
||||
for (int i = 0; i < columnList.size(); i++) {
|
||||
condition += appendIndicator(columnList.get(i)) + " ";
|
||||
}
|
||||
|
||||
} else {
|
||||
condition = booleanTerm();
|
||||
}
|
||||
|
||||
substitute.setCondition(condition);
|
||||
}
|
||||
<RPAREN>
|
||||
[
|
||||
<AND> extendedKey=<ID>
|
||||
{
|
||||
String eKey = extendedKey.image.toString();
|
||||
String cKey = chainKey.image.toString();
|
||||
substitute.setCondition(condition);
|
||||
if (secondR.equals(eKey)) {
|
||||
substitute.resetVariableRelation(eKey, cKey);
|
||||
} else {
|
||||
substitute.resetVariableRelation(cKey, eKey);
|
||||
}
|
||||
}
|
||||
]
|
||||
| <TABLE><LPAREN>
|
||||
tableName = identifier()
|
||||
{ substitute.setInput(tableName); }
|
||||
<RPAREN>
|
||||
| operation=<GROUP>chainKey=<ID><BY>
|
||||
<LPAREN>
|
||||
{
|
||||
List<String> columnList = columnList();
|
||||
for (int i = 0; i < columnList.size(); i++) {
|
||||
condition += appendIndicator(columnList.get(i));
|
||||
}
|
||||
}
|
||||
<RPAREN>
|
||||
{
|
||||
substitute.setChainKey(chainKey.image.toString());
|
||||
substitute.setOperation(operation.image.toString());
|
||||
substitute.setCondition(condition);
|
||||
}
|
||||
)
|
||||
|
||||
{
|
||||
return substitute;
|
||||
}
|
||||
}
|
||||
|
||||
SaveCommand saveCommand() :
|
||||
{
|
||||
Token t = null;
|
||||
String tableName;
|
||||
SaveCommand save = new SaveCommand(this.out);
|
||||
}
|
||||
{
|
||||
<SAVE> t=<ID>
|
||||
{ save.setStatement(t.image.toString()); }
|
||||
<INTO><TABLE><LPAREN>
|
||||
tableName = identifier() { save.setOutput(tableName); }
|
||||
<RPAREN>
|
||||
{
|
||||
return save;
|
||||
}
|
||||
}
|
||||
|
||||
List<String> getLiteralValues() :
|
||||
{
|
||||
List<String> values = new ArrayList<String>();
|
||||
|
@ -872,153 +768,13 @@ String identifier() :
|
|||
}
|
||||
{
|
||||
(
|
||||
( t=<ID> | t=<META_TABLE> | t=<ROOT_TABLE> )
|
||||
t=<ID>
|
||||
{ return t.image.toString(); }
|
||||
| ( t=<QUOTED_IDENTIFIER> | t=<STRING_LITERAL> )
|
||||
{ return t.image.substring(1,t.image.toString().length() - 1); }
|
||||
)
|
||||
}
|
||||
|
||||
String booleanTerm() :
|
||||
{
|
||||
String query = "";
|
||||
String tmp = null;
|
||||
String joinTerm = null;
|
||||
String connector = null;
|
||||
}
|
||||
{
|
||||
query = booleanTerms()
|
||||
{
|
||||
if(query.endsWith(".ROW")) {
|
||||
joinTerm = query;
|
||||
query = "";
|
||||
}
|
||||
}
|
||||
(
|
||||
(
|
||||
<AND> { connector = " AND "; }
|
||||
| <OR> { connector = " OR "; }
|
||||
) tmp = booleanTerms() {
|
||||
if(tmp.endsWith(".ROW")) {
|
||||
joinTerm = tmp;
|
||||
} else {
|
||||
if(!query.equals(""))
|
||||
query += connector;
|
||||
query += tmp;
|
||||
}
|
||||
}
|
||||
)*
|
||||
{
|
||||
if(joinTerm != null) {
|
||||
return joinTerm + " BOOL " + query;
|
||||
} else {
|
||||
return query;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
String booleanTerms() :
|
||||
{
|
||||
Token tSearchName, tComparand;
|
||||
List<String> inList = new ArrayList<String>();
|
||||
String searchName=null,comparator=null,comparand=null;
|
||||
Token joinColumn = null;
|
||||
Token joinKey = null;
|
||||
String result = null;
|
||||
}
|
||||
{
|
||||
(
|
||||
tSearchName=<ID> { searchName = tSearchName.image.toString(); }
|
||||
[
|
||||
<DOT>
|
||||
(
|
||||
joinColumn=<ID>
|
||||
{ searchName += "." + joinColumn.image.toString(); }
|
||||
| <ROW>
|
||||
{
|
||||
secondR = searchName;
|
||||
searchName += ".ROW";
|
||||
}
|
||||
)
|
||||
]
|
||||
)
|
||||
comparator = getComparator()
|
||||
(
|
||||
tComparand=<INTEGER_LITERAL>
|
||||
{ comparand = tComparand.image.toString(); }
|
||||
| tComparand=<STRING_LITERAL>
|
||||
{ comparand = tComparand.image.substring(1,tComparand.image.length() - 1); }
|
||||
| tComparand=<ID>
|
||||
{ comparand = tComparand.image.toString(); }
|
||||
[
|
||||
<DOT>
|
||||
(
|
||||
<ROW>
|
||||
{
|
||||
secondR = comparand;
|
||||
comparand += ".ROW";
|
||||
}
|
||||
| joinColumn=<ID>
|
||||
{ comparand += "." + joinColumn.image.toString(); }
|
||||
)
|
||||
]
|
||||
| inList = getColumns()
|
||||
{
|
||||
if(comparator == null) {
|
||||
comparator = "==";
|
||||
}
|
||||
comparand = "";
|
||||
try{
|
||||
for(int i=0; i<inList.size(); i++) {
|
||||
comparand += URLEncoder.encode(inList.get(i), "UTF-8");
|
||||
if(inList.size() != (i+1)) { comparand += "|"; }
|
||||
}
|
||||
} catch (UnsupportedEncodingException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
}
|
||||
)
|
||||
{
|
||||
if(searchName.endsWith(".ROW")) {
|
||||
result = appendIndicator(comparand)
|
||||
+ " " + comparator + " " + searchName;
|
||||
} else {
|
||||
result = appendIndicator(searchName)
|
||||
+ " " + comparator + " " + comparand;
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
String getComparator() :
|
||||
{
|
||||
Token t = null;
|
||||
String comparator = null;
|
||||
}
|
||||
{
|
||||
(
|
||||
t=<LCOMP>
|
||||
{ comparator = t.image.toString(); }
|
||||
[<EQUALS> { comparator += "="; }]
|
||||
| t=<RCOMP>
|
||||
{ comparator = t.image.toString(); }
|
||||
[<EQUALS> { comparator += "="; }]
|
||||
| t=<EQUALS>
|
||||
{ comparator = t.image.toString(); }
|
||||
[<LCOMP> { comparator = ">" + comparator; } ]
|
||||
[<RCOMP> { comparator = "<" + comparator; } ]
|
||||
| t=<NOTEQUAL>
|
||||
{ comparator = t.image.toString(); }
|
||||
| <NOT><IN>
|
||||
{ comparator = "!!"; }
|
||||
| <IN>
|
||||
{ comparator = "=="; }
|
||||
)
|
||||
|
||||
{ return comparator; }
|
||||
}
|
||||
|
||||
String appendIndicator(String columnName) :
|
||||
{
|
||||
String column = columnName;
|
||||
|
|
|
@ -121,39 +121,6 @@ public class HelpCommand extends BasicCommand {
|
|||
|
||||
load.put("EXIT", new String[] { "Exit shell", "EXIT;" });
|
||||
|
||||
// A Algebraic Query Commands
|
||||
// this is a tentative query language based on a hbase which uses relational
|
||||
// model of
|
||||
// data.
|
||||
|
||||
load.put("TABLE",
|
||||
new String[] { "Load a table", "A = table('table_name');" });
|
||||
load.put("SUBSTITUTE", new String[] { "Substitute expression to [A~Z]",
|
||||
"D = A.projection('cf_name1'[, 'cf_name2']);" });
|
||||
load.put("SAVE", new String[] {
|
||||
"Save results into specified table (It runs a mapreduce job)",
|
||||
"SAVE A INTO table('table_name');" });
|
||||
|
||||
// Relational Operations
|
||||
load.put("PROJECTION", new String[] {
|
||||
"Selects a subset of the columnfamilies of a relation",
|
||||
"A = TABLE('table_name');"
|
||||
+ " B = A.Projection('cf_name1'[, 'cf_name2']);" });
|
||||
load
|
||||
.put(
|
||||
"SELECTION",
|
||||
new String[] {
|
||||
"Selects a subset of the rows in a relation that satisfy a selection condition (>, <, AND, OR, etc.)",
|
||||
"A = Table('table_name');"
|
||||
+ " B = A.Selection(cf_name1 > 100 [AND cf_name2 = 'string_value']);" });
|
||||
|
||||
// Aggregation Functions
|
||||
// TODO : and apply aggregate function independently to each group of rows
|
||||
load.put("GROUP", new String[] {
|
||||
"Group rows by value of an attribute",
|
||||
"A = Table('table_name');"
|
||||
+ " B = Group A by ('cf_name1'[, 'cf_name2']);" });
|
||||
|
||||
return load;
|
||||
}
|
||||
|
||||
|
|
|
@ -1,114 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.Writer;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseAdmin;
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HConnection;
|
||||
import org.apache.hadoop.hbase.HConnectionManager;
|
||||
import org.apache.hadoop.hbase.shell.algebra.OperationEvaluator;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobClient;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.RunningJob;
|
||||
|
||||
/**
|
||||
* Save results to specified table.
|
||||
*/
|
||||
public class SaveCommand extends BasicCommand {
|
||||
private String chainKey;
|
||||
private String output;
|
||||
|
||||
public SaveCommand(Writer o) {
|
||||
super(o);
|
||||
}
|
||||
|
||||
public ReturnMsg execute(HBaseConfiguration conf) {
|
||||
try {
|
||||
HConnection conn = HConnectionManager.getConnection(conf);
|
||||
|
||||
if (!conn.tableExists(new Text(output))) {
|
||||
OperationEvaluator eval = new OperationEvaluator(conf, chainKey, output);
|
||||
JobConf jobConf = eval.getJobConf();
|
||||
if (submitJob(jobConf)) {
|
||||
return new ReturnMsg(0, "Successfully complete.");
|
||||
} else {
|
||||
HBaseAdmin admin = new HBaseAdmin(conf);
|
||||
admin.deleteTable(new Text(output));
|
||||
|
||||
return new ReturnMsg(0, "Job failed.");
|
||||
}
|
||||
} else {
|
||||
return new ReturnMsg(0, "'" + output + "' table already exist.");
|
||||
}
|
||||
|
||||
} catch (IOException e) {
|
||||
return new ReturnMsg(0, e.toString());
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Submit a job to job tracker.
|
||||
*
|
||||
* @param job
|
||||
* @return result
|
||||
* @throws IOException
|
||||
*/
|
||||
public boolean submitJob(JobConf job) throws IOException {
|
||||
JobClient jc = new JobClient(job);
|
||||
boolean success = true;
|
||||
RunningJob running = null;
|
||||
try {
|
||||
running = jc.submitJob(job);
|
||||
String jobId = running.getJobID();
|
||||
|
||||
while (!running.isComplete()) {
|
||||
try {
|
||||
Thread.sleep(5000);
|
||||
} catch (InterruptedException e) {
|
||||
}
|
||||
running = jc.getJob(jobId);
|
||||
}
|
||||
success = running.isSuccessful();
|
||||
} finally {
|
||||
if (!success && (running != null)) {
|
||||
running.killJob();
|
||||
}
|
||||
jc.close();
|
||||
}
|
||||
return success;
|
||||
}
|
||||
|
||||
public void setOutput(String output) {
|
||||
this.output = output;
|
||||
}
|
||||
|
||||
public void setStatement(String chainKey) {
|
||||
this.chainKey = chainKey;
|
||||
}
|
||||
|
||||
@Override
|
||||
public CommandType getCommandType() {
|
||||
return CommandType.SHELL;
|
||||
}
|
||||
}
|
|
@ -45,7 +45,8 @@ public class ShowCommand extends BasicCommand {
|
|||
this(o, f, null);
|
||||
}
|
||||
|
||||
public ShowCommand(final Writer o, final TableFormatter f, final String argument) {
|
||||
public ShowCommand(final Writer o, final TableFormatter f,
|
||||
final String argument) {
|
||||
super(o);
|
||||
this.formatter = f;
|
||||
this.command = argument;
|
||||
|
@ -58,7 +59,6 @@ public class ShowCommand extends BasicCommand {
|
|||
try {
|
||||
HBaseAdmin admin = new HBaseAdmin(conf);
|
||||
int tableLength = 0;
|
||||
if ("tables".equals(this.command)) {
|
||||
HTableDescriptor[] tables = admin.listTables();
|
||||
tableLength = tables.length;
|
||||
if (tableLength == 0) {
|
||||
|
@ -71,33 +71,6 @@ public class ShowCommand extends BasicCommand {
|
|||
}
|
||||
formatter.footer();
|
||||
return new ReturnMsg(1, tableLength + " table(s) in set.");
|
||||
} else {
|
||||
Map<String, VariableRef> refer = VariablesPool.get(command);
|
||||
if (refer == null) {
|
||||
return new ReturnMsg(0, "Unknown arguments.");
|
||||
}
|
||||
|
||||
String msg = null;
|
||||
for (Map.Entry<String, VariableRef> e : refer.entrySet()) {
|
||||
msg = command + " = ";
|
||||
if (e.getKey() != null) {
|
||||
msg += e.getKey() + ".";
|
||||
}
|
||||
msg += e.getValue().getOperation() + "(";
|
||||
if (e.getValue().getOperation().equals("projection")) {
|
||||
String[] proj = e.getValue().getArgument().split(" ");
|
||||
for (int i = 0; i < proj.length; i++) {
|
||||
msg += "'" + proj[i] + "'";
|
||||
if (i + 1 != proj.length)
|
||||
msg += ", ";
|
||||
}
|
||||
} else {
|
||||
msg += e.getValue().getArgument().replace(" BOOL ", " and ");
|
||||
}
|
||||
msg += ");\n";
|
||||
}
|
||||
return new ReturnMsg(0, msg);
|
||||
}
|
||||
} catch (IOException e) {
|
||||
return new ReturnMsg(0, "error msg : " + e.toString());
|
||||
}
|
||||
|
|
|
@ -1,79 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell;
|
||||
|
||||
import java.io.Writer;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.shell.algebra.Constants;
|
||||
|
||||
/**
|
||||
* This class represents a substitute command.
|
||||
*/
|
||||
public class SubstituteCommand extends BasicCommand {
|
||||
private String key;
|
||||
private String chainKey;
|
||||
private String operation;
|
||||
private String condition;
|
||||
|
||||
public SubstituteCommand(Writer o) {
|
||||
super(o);
|
||||
}
|
||||
|
||||
public ReturnMsg execute(HBaseConfiguration conf) {
|
||||
VariableRef formula = new VariableRef(operation, condition);
|
||||
VariablesPool.put(key, chainKey, formula);
|
||||
return null;
|
||||
}
|
||||
|
||||
public void setInput(String input) {
|
||||
this.operation = "table";
|
||||
this.condition = input;
|
||||
}
|
||||
|
||||
public void setKey(String key) {
|
||||
this.key = key;
|
||||
}
|
||||
|
||||
public void setChainKey(String chainKey) {
|
||||
this.chainKey = chainKey;
|
||||
}
|
||||
|
||||
public void setOperation(String operation) {
|
||||
this.operation = operation;
|
||||
}
|
||||
|
||||
public void setCondition(String condition) {
|
||||
this.condition = condition;
|
||||
}
|
||||
|
||||
public void resetVariableRelation(String r1, String r2) {
|
||||
setChainKey(r1);
|
||||
String tableName = VariablesPool.get(r1).get(null).getArgument();
|
||||
VariableRef formula = new VariableRef(Constants.JOIN_SECOND_RELATION,
|
||||
tableName);
|
||||
VariablesPool.put(r1, r2, formula);
|
||||
}
|
||||
|
||||
@Override
|
||||
public CommandType getCommandType() {
|
||||
return CommandType.SHELL;
|
||||
}
|
||||
}
|
|
@ -1,52 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell;
|
||||
|
||||
/**
|
||||
* VariableRef is used to reference declared Variables.
|
||||
*/
|
||||
public class VariableRef {
|
||||
String operation;
|
||||
String argument;
|
||||
|
||||
/** Constructor */
|
||||
public VariableRef(String operation, String condition) {
|
||||
this.operation = operation;
|
||||
this.argument = condition;
|
||||
}
|
||||
|
||||
/**
|
||||
* Return argument of an operation
|
||||
*
|
||||
* @return argument
|
||||
*/
|
||||
public String getArgument() {
|
||||
return argument;
|
||||
}
|
||||
|
||||
/**
|
||||
* Return operation
|
||||
*
|
||||
* @return operation
|
||||
*/
|
||||
public String getOperation() {
|
||||
return operation;
|
||||
}
|
||||
}
|
|
@ -1,52 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell;
|
||||
|
||||
import java.util.HashMap;
|
||||
|
||||
/**
|
||||
* Variable pool is a collection of substitution variables.
|
||||
*/
|
||||
public class VariablesPool {
|
||||
static HashMap<String, HashMap<String, VariableRef>> variables = new HashMap<String, HashMap<String, VariableRef>>();
|
||||
|
||||
/**
|
||||
* puts the date in the substitution variable.
|
||||
*
|
||||
* @param key
|
||||
* @param parentKey
|
||||
* @param statement
|
||||
*/
|
||||
public static void put(String key, String parentKey, VariableRef statement) {
|
||||
HashMap<String, VariableRef> value = new HashMap<String, VariableRef>();
|
||||
value.put(parentKey, statement);
|
||||
variables.put(key, value);
|
||||
}
|
||||
|
||||
/**
|
||||
* returns the substitution variable's value.
|
||||
*
|
||||
* @param key
|
||||
* @return HashMap<String, VariableRef>
|
||||
*/
|
||||
public static HashMap<String, VariableRef> get(String key) {
|
||||
return variables.get(key);
|
||||
}
|
||||
}
|
|
@ -1,36 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
/**
|
||||
* List of access control algebraic operations constants.
|
||||
*/
|
||||
public class Constants {
|
||||
public static final String OUTPUT_TABLE_EXIST = "job.config.output.table";
|
||||
public static final String CONFIG_INPUT = "input";
|
||||
public static final String CONFIG_OUTPUT = "output";
|
||||
public static final String EXPRESSION_FILTER_LIST = "expression.filter.list";
|
||||
|
||||
public static final String RELATIONAL_PROJECTION = "projection";
|
||||
public static final String RELATIONAL_SELECTION = "selection";
|
||||
public static final String RELATIONAL_GROUP = "group";
|
||||
public static final String RELATIONAL_JOIN = "join";
|
||||
public static final String JOIN_SECOND_RELATION = "secondR";
|
||||
}
|
|
@ -1,50 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Map;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableMap;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableReduce;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Duplicates Table. R1 to R3 in O(N)
|
||||
*/
|
||||
public class DuplicateTable extends RelationalOperation {
|
||||
public DuplicateTable(HBaseConfiguration conf, Map<String, String> condition) {
|
||||
super(conf, condition);
|
||||
}
|
||||
|
||||
@Override
|
||||
public JobConf getConf() throws IOException, RuntimeException {
|
||||
HColumnDescriptor[] columns = getInputColumnDescriptor();
|
||||
outputTableCreate(columns, null);
|
||||
|
||||
IdentityTableMap.initJob(input, getColumnStringArray(columns),
|
||||
IdentityTableMap.class, jobConf);
|
||||
IdentityTableReduce.initJob(output, IdentityTableReduce.class, jobConf);
|
||||
|
||||
return jobConf;
|
||||
}
|
||||
}
|
|
@ -1,559 +0,0 @@
|
|||
options {
|
||||
STATIC = false;
|
||||
IGNORE_CASE = true;
|
||||
}
|
||||
|
||||
PARSER_BEGIN(ExpressionParser)
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.Reader;
|
||||
import java.io.StringReader;
|
||||
import java.io.UnsupportedEncodingException;
|
||||
import java.net.URLDecoder;
|
||||
import java.util.ArrayList;
|
||||
import java.util.HashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.SortedMap;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HTable;
|
||||
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.io.Writable;
|
||||
|
||||
/**
|
||||
* Parsing command line.
|
||||
*/
|
||||
public class ExpressionParser {
|
||||
HBaseConfiguration conf = new HBaseConfiguration();
|
||||
private String expression;
|
||||
private String joinKeyColumn;
|
||||
private String secondRelation;
|
||||
private Map<String, List<String>> unionSet = new HashMap<String, List<String>>();
|
||||
private Map<String, List<String>> intersectionSet = new HashMap<String, List<String>>();
|
||||
private SortedMap<Text, byte[]> secondValue = null;
|
||||
|
||||
public ExpressionParser(final String expression) {
|
||||
this((Reader)(new StringReader(expression)));
|
||||
this.expression = expression;
|
||||
}
|
||||
|
||||
public String getExpression() {
|
||||
return this.expression;
|
||||
}
|
||||
}
|
||||
|
||||
PARSER_END(ExpressionParser)
|
||||
|
||||
SKIP :
|
||||
{
|
||||
" "
|
||||
| "\t"
|
||||
| "\r"
|
||||
| "\n"
|
||||
}
|
||||
|
||||
TOKEN : /** Literals */
|
||||
{
|
||||
<ROW: "row">
|
||||
| <AND: "and">
|
||||
| <OR: "or">
|
||||
| <DOT: ".">
|
||||
| <EQUALS: "=">
|
||||
| <LCOMP: ">">
|
||||
| <RCOMP: "<">
|
||||
| <LCOMPE: ">=">
|
||||
| <RCOMPE: "<=">
|
||||
| <IN: "==">
|
||||
| <NOTIN: "!!">
|
||||
| <BAR: "|">
|
||||
| <BOOL: "bool">
|
||||
| <ID: ["A"-"Z","a"-"z","_","-",":","/"] (["A"-"Z","a"-"z","0"-"9","_","-",":","/"])* >
|
||||
| <INTEGER_LITERAL: (["0"-"9"])+ >
|
||||
| <FLOATING_POINT_LITERAL:
|
||||
(["0"-"9"])+ "." (["0"-"9"])+ (<EXPONENT>)?
|
||||
| "." (["0"-"9"])+ (<EXPONENT>)?
|
||||
| (["0"-"9"])+ <EXPONENT>
|
||||
| (["0"-"9"])+ (<EXPONENT>)?
|
||||
>
|
||||
| <#EXPONENT: ["e","E"] (["+","-"])? (["0"-"9"])+ >
|
||||
| <QUOTED_IDENTIFIER: "\"" (~["\""])+ "\"" >
|
||||
| <STRING_LITERAL: "'" (~["'"])* ( "''" (~["'"])* )* "'" >
|
||||
}
|
||||
|
||||
void booleanExpressionParse() :
|
||||
{
|
||||
Map<String, List<String>> temp = new HashMap<String, List<String>>();
|
||||
}
|
||||
{
|
||||
temp = booleanTerm()
|
||||
(
|
||||
(
|
||||
<AND>
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
}
|
||||
| <OR>
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
}
|
||||
)
|
||||
)*
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void joinExpressionParse() :
|
||||
{
|
||||
Map<String, List<String>> temp = new HashMap<String, List<String>>();
|
||||
}
|
||||
{
|
||||
indexJoinCondition()
|
||||
<BOOL>
|
||||
[
|
||||
temp = booleanTerm()
|
||||
(
|
||||
(
|
||||
<AND>
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
}
|
||||
| <OR>
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
}
|
||||
)
|
||||
)*
|
||||
{
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
|
||||
void indexJoinCondition() :
|
||||
{
|
||||
Token firstR = null;
|
||||
Token joinKey = null;
|
||||
Token secondR = null;
|
||||
}
|
||||
{
|
||||
firstR=<ID><DOT>joinKey=<ID>
|
||||
<EQUALS>
|
||||
secondR=<ID><DOT><ROW>
|
||||
{
|
||||
joinKeyColumn = joinKey.image.toString();
|
||||
secondRelation = secondR.image.toString();
|
||||
}
|
||||
}
|
||||
|
||||
Map<String, List<String>> booleanTerm() :
|
||||
{
|
||||
Token tSearchName = null;
|
||||
Token tComparator = null;
|
||||
Token tFirst = null;
|
||||
Token tSecond = null;
|
||||
Map<String, List<String>> result = new HashMap<String, List<String>>();
|
||||
List<String> valueList = new ArrayList<String>();
|
||||
String comparand = null;
|
||||
}
|
||||
{
|
||||
(
|
||||
tSearchName=<ID>
|
||||
| tSearchName=<INTEGER_LITERAL>
|
||||
) [<DOT> ( tFirst=<ID> | tFirst=<INTEGER_LITERAL> ) ]
|
||||
( tComparator=<EQUALS> | tComparator=<LCOMP> | tComparator=<LCOMPE> | tComparator=<RCOMPE>
|
||||
| tComparator=<RCOMP> | tComparator=<IN> | tComparator=<NOTIN> )
|
||||
comparand = getValueList()
|
||||
[<DOT> ( tSecond=<ID> | tSecond=<INTEGER_LITERAL> )]
|
||||
|
||||
{
|
||||
if(tFirst == null && tSecond == null) {
|
||||
valueList.add(tComparator.image.toString() + " " +comparand);
|
||||
result.put(tSearchName.image.toString(), valueList);
|
||||
} else if (tFirst != null && tSecond != null ){
|
||||
if(tSearchName.image.toString().equals(secondRelation)) {
|
||||
valueList.add(tComparator.image.toString() + " "
|
||||
+ secondValue.get(new Text(tSearchName.image.toString())));
|
||||
result.put(tFirst.image.toString(), valueList);
|
||||
} else {
|
||||
valueList.add(tComparator.image.toString() + " " + tSecond.image.toString());
|
||||
result.put(tFirst.image.toString(), valueList);
|
||||
}
|
||||
}
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
String getValueList() :
|
||||
{
|
||||
Token tComparand = null;
|
||||
Token tList = null;
|
||||
String result = "";
|
||||
}
|
||||
{
|
||||
( tComparand=<ID> | tComparand=<INTEGER_LITERAL> )
|
||||
{
|
||||
result = tComparand.image.toString();
|
||||
}
|
||||
[
|
||||
<BAR>
|
||||
(
|
||||
(tList=<ID> | tList=<INTEGER_LITERAL> )
|
||||
{
|
||||
result += "|" + tList.image.toString();
|
||||
}
|
||||
[ <BAR> ]
|
||||
)*
|
||||
]
|
||||
{ return result; }
|
||||
}
|
||||
|
||||
boolean checkConstraints(MapWritable data) :
|
||||
{
|
||||
Map<String, String> record = getComparisonObject(data);
|
||||
boolean result = false;
|
||||
if(intersectionSet.size() == 0 && unionSet.size() == 0) {
|
||||
return true;
|
||||
}
|
||||
if (data == null || record.size() == 0) {
|
||||
return result; // return false if data is null.
|
||||
}
|
||||
}
|
||||
{
|
||||
{
|
||||
if (intersectionSet.size() == record.size()) {
|
||||
result = booleanCompare(intersectionSet, record, true);
|
||||
} else if (unionSet.size() == record.size()) {
|
||||
result = booleanCompare(unionSet, record, false);
|
||||
} else {
|
||||
result = getCompareResult(record);
|
||||
}
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
Map<String, String> getComparisonObject(MapWritable data) :
|
||||
{
|
||||
Map<String, String> result = new HashMap<String, String>();
|
||||
}
|
||||
{
|
||||
{
|
||||
for (Map.Entry<Writable, Writable> e : data.entrySet()) {
|
||||
String cKey = e.getKey().toString();
|
||||
String val = new String(((ImmutableBytesWritable) e.getValue()).get());
|
||||
|
||||
if (intersectionSet.containsKey(cKey) || unionSet.containsKey(cKey)) {
|
||||
result.put(cKey, val);
|
||||
}
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
MapWritable getJoinColumns(MapWritable value, int numCols, String secondRelation) :
|
||||
{
|
||||
MapWritable appendValue = new MapWritable();
|
||||
String joinKey = null;
|
||||
this.secondRelation = secondRelation;
|
||||
}
|
||||
{
|
||||
{
|
||||
if (numCols > 0) {
|
||||
for (Map.Entry<Writable, Writable> e : value.entrySet()) {
|
||||
Text column = (Text) e.getKey();
|
||||
for (int i = 0; i < numCols; i++) {
|
||||
if (column.equals(new Text(joinKeyColumn))) {
|
||||
joinKey = new String(((ImmutableBytesWritable) e.getValue()).get());
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
//If joinKey is null, just return.
|
||||
if(joinKey == null) {
|
||||
return appendValue;
|
||||
}
|
||||
|
||||
try {
|
||||
HTable table = new HTable(conf, new Text(secondRelation));
|
||||
secondValue = table.getRow(new Text(joinKey));
|
||||
for (Map.Entry<Text, byte[]> e : secondValue.entrySet()) {
|
||||
appendValue.put(e.getKey(), new ImmutableBytesWritable(e.getValue()));
|
||||
}
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
return appendValue;
|
||||
}
|
||||
}
|
||||
|
||||
boolean getCompareResult(Map<String, String> record) :
|
||||
{
|
||||
}
|
||||
{
|
||||
{
|
||||
return (booleanCompare(intersectionSet, record, true) &&
|
||||
booleanCompare(unionSet, record, false)) ? true : false;
|
||||
}
|
||||
}
|
||||
|
||||
boolean booleanCompare(Map<String, List<String>> expression,
|
||||
Map<String, String> record, boolean isIntersection) :
|
||||
{
|
||||
boolean negative = true;
|
||||
boolean positive = false;
|
||||
}
|
||||
{
|
||||
{
|
||||
try{
|
||||
for (Map.Entry<String, List<String>> e : expression.entrySet()) {
|
||||
String key = e.getKey();
|
||||
List<String> valueList = e.getValue();
|
||||
String recordValueList = record.get(key);
|
||||
|
||||
for (int i = 0; i < valueList.size(); i++) {
|
||||
String[] term = valueList.get(i).split(" ");
|
||||
String comparator = term[0];
|
||||
String comparand = term[1];
|
||||
|
||||
switch (comparator.charAt(0)) {
|
||||
case '>':
|
||||
if (isSecond(comparator, "=")) {
|
||||
if (Integer.parseInt(comparand) > Integer
|
||||
.parseInt(recordValueList)) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
} else {
|
||||
if (Integer.parseInt(comparand) > Integer
|
||||
.parseInt(recordValueList)
|
||||
|| comparand.equals(recordValueList)) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
}
|
||||
break;
|
||||
case '<':
|
||||
if (isSecond(comparator, "=")) {
|
||||
if (Integer.parseInt(comparand) < Integer
|
||||
.parseInt(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
} else {
|
||||
if (Integer.parseInt(comparand) < Integer
|
||||
.parseInt(recordValueList)
|
||||
|| comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
|
||||
case '!':
|
||||
if (isSecond(comparator, "!")) {
|
||||
boolean checkBool = true;
|
||||
String[] coms = comparand.split("[|]");
|
||||
for (int j = 0; j < coms.length; j++) {
|
||||
if (URLDecoder.decode(coms[j], "UTF-8").equals(recordValueList)) {
|
||||
checkBool = false;
|
||||
}
|
||||
}
|
||||
|
||||
if (!checkBool) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
|
||||
} else {
|
||||
if (comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
case '=':
|
||||
if (isSecond(comparator, "=")) {
|
||||
|
||||
boolean checkBool = true;
|
||||
String[] coms = comparand.split("[|]");
|
||||
for (int j = 0; j < coms.length; j++) {
|
||||
if (URLDecoder.decode(coms[j], "UTF-8").equals(recordValueList)) {
|
||||
checkBool = false;
|
||||
}
|
||||
}
|
||||
|
||||
if (checkBool) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
|
||||
} else {
|
||||
if (!comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
} catch (UnsupportedEncodingException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
boolean result = false;
|
||||
if (isIntersection) {
|
||||
result = negative;
|
||||
} else {
|
||||
result = positive;
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
boolean isSecond(String comparator, String string) :
|
||||
{
|
||||
}
|
||||
{
|
||||
{
|
||||
return (comparator.length() == 2 && string.charAt(0) == comparator.charAt(1))
|
||||
? true : false;
|
||||
}
|
||||
}
|
|
@ -1,83 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
|
||||
import org.apache.hadoop.hbase.HStoreKey;
|
||||
import org.apache.hadoop.hbase.mapred.GroupingTableMap;
|
||||
import org.apache.hadoop.hbase.mapred.TableMap;
|
||||
import org.apache.hadoop.hbase.mapred.TableOutputCollector;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ExpressionParser;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ParseException;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.Reporter;
|
||||
|
||||
/**
|
||||
* Extract grouping columns from filtered records.
|
||||
*/
|
||||
public class GroupingFilterMap extends GroupingTableMap {
|
||||
ExpressionParser expressionParser;
|
||||
public static final String EXPRESSION = "shell.mapred.filtertablemap.exps";
|
||||
|
||||
public static void initJob(String table, String columns, String groupColumns,
|
||||
String expression, Class<? extends TableMap> mapper, JobConf job) {
|
||||
initJob(table, columns, mapper, job);
|
||||
job.set(GROUP_COLUMNS, groupColumns);
|
||||
job.set(EXPRESSION, expression);
|
||||
}
|
||||
|
||||
/** {@inheritDoc} */
|
||||
@Override
|
||||
public void configure(JobConf job) {
|
||||
super.configure(job);
|
||||
String[] cols = job.get(GROUP_COLUMNS, "").split(" ");
|
||||
m_columns = new Text[cols.length];
|
||||
for (int i = 0; i < cols.length; i++) {
|
||||
m_columns[i] = new Text(cols[i]);
|
||||
}
|
||||
expressionParser = new ExpressionParser(job.get(EXPRESSION, ""));
|
||||
try {
|
||||
expressionParser.booleanExpressionParse();
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
|
||||
public void map(@SuppressWarnings("unused")
|
||||
HStoreKey key, MapWritable value, TableOutputCollector output,
|
||||
@SuppressWarnings("unused")
|
||||
Reporter reporter) throws IOException {
|
||||
byte[][] keyVals = extractKeyValues(value);
|
||||
if (keyVals != null) {
|
||||
Text tKey = createGroupKey(keyVals);
|
||||
|
||||
try {
|
||||
if (expressionParser.checkConstraints(value)) {
|
||||
output.collect(tKey, value);
|
||||
}
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,78 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
|
||||
import org.apache.hadoop.hbase.HStoreKey;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableMap;
|
||||
import org.apache.hadoop.hbase.mapred.TableMap;
|
||||
import org.apache.hadoop.hbase.mapred.TableOutputCollector;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ExpressionParser;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ParseException;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.Reporter;
|
||||
|
||||
/**
|
||||
* Extract filtered records.
|
||||
*/
|
||||
public class IdentityFilterMap extends IdentityTableMap {
|
||||
ExpressionParser expressionParser;
|
||||
public static final String EXPRESSION = "shell.mapred.filtertablemap.exps";
|
||||
|
||||
@SuppressWarnings("deprecation")
|
||||
public static void initJob(String table, String columns, String expression,
|
||||
Class<? extends TableMap> mapper, JobConf job) {
|
||||
initJob(table, columns, mapper, job);
|
||||
job.set(EXPRESSION, expression);
|
||||
}
|
||||
|
||||
/*
|
||||
* (non-Javadoc)
|
||||
*
|
||||
* @see org.apache.hadoop.hbase.mapred.TableMap#configure(org.apache.hadoop.mapred.JobConf)
|
||||
*/
|
||||
public void configure(JobConf job) {
|
||||
super.configure(job);
|
||||
expressionParser = new ExpressionParser(job.get(EXPRESSION, ""));
|
||||
try {
|
||||
expressionParser.booleanExpressionParse();
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Filter the value for each specified column family.
|
||||
*/
|
||||
public void map(HStoreKey key, MapWritable value,
|
||||
TableOutputCollector output, Reporter reporter) throws IOException {
|
||||
Text tKey = key.getRow();
|
||||
try {
|
||||
if (expressionParser.checkConstraints(value)) {
|
||||
output.collect(tKey, value);
|
||||
}
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,76 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Map;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Perform a index join using MapReduce.
|
||||
*/
|
||||
public class IndexJoin extends RelationalOperation {
|
||||
public IndexJoin(HBaseConfiguration conf, Map<String, String> condition) {
|
||||
super(conf, condition);
|
||||
}
|
||||
|
||||
@Override
|
||||
public JobConf getConf() throws IOException, RuntimeException {
|
||||
String secondRelation = condition.get(Constants.JOIN_SECOND_RELATION);
|
||||
|
||||
HColumnDescriptor[] firstColumns = null;
|
||||
HColumnDescriptor[] secondColumns = null;
|
||||
for (int i = 0; i < tables.length; i++) {
|
||||
if (tables[i].getName().equals(new Text(input))) {
|
||||
firstColumns = tables[i].getFamilies().values().toArray(
|
||||
new HColumnDescriptor[] {});
|
||||
} else if (tables[i].getName().equals(new Text(secondRelation))) {
|
||||
secondColumns = tables[i].getFamilies().values().toArray(
|
||||
new HColumnDescriptor[] {});
|
||||
}
|
||||
}
|
||||
|
||||
String firstColumnsStr = "";
|
||||
String secondColumnsStr = "";
|
||||
|
||||
for (int i = 0; i < firstColumns.length; i++) {
|
||||
desc.addFamily(firstColumns[i]);
|
||||
firstColumnsStr += firstColumns[i].getName() + " ";
|
||||
}
|
||||
|
||||
for (int i = 0; i < secondColumns.length; i++) {
|
||||
desc.addFamily(secondColumns[i]);
|
||||
secondColumnsStr += secondColumns[i].getName() + " ";
|
||||
}
|
||||
|
||||
admin.createTable(desc); // create output table.
|
||||
|
||||
IndexJoinMap.initJob(input, secondRelation, firstColumnsStr,
|
||||
secondColumnsStr, condition.get(Constants.RELATIONAL_JOIN),
|
||||
IndexJoinMap.class, jobConf);
|
||||
IndexJoinReduce.initJob(output, IndexJoinReduce.class, jobConf);
|
||||
|
||||
return jobConf;
|
||||
}
|
||||
}
|
|
@ -1,108 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
|
||||
import org.apache.hadoop.hbase.HStoreKey;
|
||||
import org.apache.hadoop.hbase.mapred.TableMap;
|
||||
import org.apache.hadoop.hbase.mapred.TableOutputCollector;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ExpressionParser;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ParseException;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.Reporter;
|
||||
|
||||
/**
|
||||
* An index join exploits the existence of an row index for one of the relations
|
||||
* used in the join to find matching rows more quickly.
|
||||
*
|
||||
* Index join (using R2 row index) takes time O(i+m)/map function number.
|
||||
*/
|
||||
public class IndexJoinMap extends TableMap {
|
||||
ExpressionParser expressionParser;
|
||||
private String secondRelation;
|
||||
public static final String JOIN_EXPRESSION = "shell.mapred.join.expression";
|
||||
public static final String SECOND_RELATION = "shell.mapred.join.second.relation";
|
||||
public static final String FIRST_COLUMNS = "shell.mapred.first.columns";
|
||||
private Text[] first_columns;
|
||||
|
||||
/** constructor */
|
||||
public IndexJoinMap() {
|
||||
super();
|
||||
}
|
||||
|
||||
/**
|
||||
* @param firstRelation R1
|
||||
* @param secondRelation R2
|
||||
* @param firstColumns (A 1,A 2,...,A n)
|
||||
* @param secondColumns (B~1~,B~2~,...,B~m~)
|
||||
* @param joinExpression join condition expression
|
||||
* @param mapper mapper class
|
||||
* @param job jobConf
|
||||
*/
|
||||
public static void initJob(String firstRelation, String secondRelation,
|
||||
String firstColumns, String secondColumns, String joinExpression,
|
||||
Class<? extends TableMap> mapper, JobConf job) {
|
||||
initJob(firstRelation, firstColumns, mapper, job);
|
||||
job.set(JOIN_EXPRESSION, joinExpression);
|
||||
job.set(SECOND_RELATION, secondRelation);
|
||||
job.set(FIRST_COLUMNS, firstColumns);
|
||||
}
|
||||
|
||||
/** {@inheritDoc} */
|
||||
@Override
|
||||
public void configure(JobConf job) {
|
||||
super.configure(job);
|
||||
secondRelation = job.get(SECOND_RELATION, "");
|
||||
String[] cols = job.get(FIRST_COLUMNS, "").split(" ");
|
||||
first_columns = new Text[cols.length];
|
||||
for (int i = 0; i < cols.length; i++) {
|
||||
first_columns[i] = new Text(cols[i]);
|
||||
}
|
||||
|
||||
expressionParser = new ExpressionParser(job.get(JOIN_EXPRESSION, ""));
|
||||
try {
|
||||
expressionParser.joinExpressionParse();
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
|
||||
@Override
|
||||
public void map(HStoreKey key, MapWritable value,
|
||||
TableOutputCollector output, Reporter reporter) throws IOException {
|
||||
Text tKey = key.getRow();
|
||||
try {
|
||||
MapWritable appendValue = expressionParser.getJoinColumns(value,
|
||||
first_columns.length, secondRelation);
|
||||
|
||||
if (appendValue.size() != 0) {
|
||||
value.putAll(appendValue);
|
||||
if (expressionParser.checkConstraints(value)) {
|
||||
output.collect(tKey, value);
|
||||
}
|
||||
}
|
||||
} catch (ParseException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,49 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Iterator;
|
||||
|
||||
import org.apache.hadoop.hbase.mapred.TableOutputCollector;
|
||||
import org.apache.hadoop.hbase.mapred.TableReduce;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.Reporter;
|
||||
|
||||
/**
|
||||
* Table join Reduce class
|
||||
*/
|
||||
public class IndexJoinReduce extends TableReduce {
|
||||
/** constructor */
|
||||
public IndexJoinReduce() {
|
||||
super();
|
||||
}
|
||||
|
||||
@Override
|
||||
public void reduce(Text key, @SuppressWarnings("unchecked")
|
||||
Iterator values, TableOutputCollector output, Reporter reporter)
|
||||
throws IOException {
|
||||
while (values.hasNext()) {
|
||||
MapWritable r = (MapWritable) values.next();
|
||||
output.collect(key, r);
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,41 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Operation interface for one of algebra operations like relational algebra
|
||||
* operations, matrix algebra operations, linear algebra operations, topological
|
||||
* algebra operations, etc.
|
||||
*/
|
||||
public interface Operation {
|
||||
|
||||
/**
|
||||
* return the Map/Reduce job configuration for performing operations.
|
||||
*
|
||||
* @return JobConf
|
||||
* @throws IOException
|
||||
* @throws RuntimeException
|
||||
*/
|
||||
JobConf getConf() throws IOException, RuntimeException;
|
||||
}
|
|
@ -1,88 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.HashMap;
|
||||
import java.util.Map;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.shell.VariableRef;
|
||||
import org.apache.hadoop.hbase.shell.VariablesPool;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Each algebra operation can be evaluated one of several different algorithms.
|
||||
*
|
||||
* So, It should be query executor/optimizer later. And It will become the core
|
||||
* module that regulates the query-performance of Hbase Shell.
|
||||
*
|
||||
* @see <a
|
||||
* href="http://wiki.apache.org/lucene-hadoop/Hbase/HbaseShell/Executor">Intergrated
|
||||
* query executor architecture</a>
|
||||
*/
|
||||
public class OperationEvaluator {
|
||||
private HBaseConfiguration conf;
|
||||
Map<String, String> condition = new HashMap<String, String>();
|
||||
|
||||
/** Constructor */
|
||||
public OperationEvaluator(HBaseConfiguration conf, String chainKey, String output) {
|
||||
this.conf = conf;
|
||||
String chain = chainKey;
|
||||
String input = null;
|
||||
|
||||
while (chain != null) {
|
||||
for (Map.Entry<String, VariableRef> e : VariablesPool.get(chain).entrySet()) {
|
||||
if (e.getKey() == null) {
|
||||
input = e.getValue().getArgument();
|
||||
} else {
|
||||
condition.put(e.getValue().getOperation(), e.getValue()
|
||||
.getArgument());
|
||||
}
|
||||
chain = e.getKey();
|
||||
}
|
||||
}
|
||||
condition.put(Constants.CONFIG_INPUT, input);
|
||||
condition.put(Constants.CONFIG_OUTPUT, output);
|
||||
}
|
||||
|
||||
|
||||
/**
|
||||
* Returns the job configuration object for statements type
|
||||
*
|
||||
* @return JobConf
|
||||
* @throws IOException
|
||||
* @throws RuntimeException
|
||||
*/
|
||||
public JobConf getJobConf() throws IOException, RuntimeException {
|
||||
RelationalOperation operation;
|
||||
if (condition.containsKey(Constants.RELATIONAL_SELECTION)) {
|
||||
operation = new Selection(conf, condition);
|
||||
} else if (condition.containsKey(Constants.RELATIONAL_PROJECTION)) {
|
||||
operation = new Projection(conf, condition);
|
||||
} else if (condition.containsKey(Constants.RELATIONAL_JOIN)) {
|
||||
operation = new IndexJoin(conf, condition);
|
||||
} else {
|
||||
operation = new DuplicateTable(conf, condition);
|
||||
}
|
||||
|
||||
return operation.getOperation().getConf();
|
||||
}
|
||||
}
|
|
@ -1,54 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Map;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.mapred.GroupingTableMap;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableMap;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableReduce;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Perform a relational projection using MapReduce.
|
||||
*/
|
||||
public class Projection extends RelationalOperation {
|
||||
public Projection(HBaseConfiguration conf, Map<String, String> condition) {
|
||||
super(conf, condition);
|
||||
}
|
||||
|
||||
@Override
|
||||
public JobConf getConf() throws IOException, RuntimeException {
|
||||
outputTableCreate(null, condition.get(Constants.RELATIONAL_PROJECTION));
|
||||
|
||||
if (condition.containsKey(Constants.RELATIONAL_GROUP)) {
|
||||
GroupingTableMap.initJob(input, getProjColumns(), getGroupColumns(),
|
||||
GroupingTableMap.class, jobConf);
|
||||
} else {
|
||||
IdentityTableMap.initJob(input, getProjColumns(), IdentityTableMap.class,
|
||||
jobConf);
|
||||
}
|
||||
|
||||
IdentityTableReduce.initJob(output, IdentityTableReduce.class, jobConf);
|
||||
return jobConf;
|
||||
}
|
||||
}
|
|
@ -1,175 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.HashSet;
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseAdmin;
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.hbase.HConnection;
|
||||
import org.apache.hadoop.hbase.HConnectionManager;
|
||||
import org.apache.hadoop.hbase.HTableDescriptor;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.ClusterStatus;
|
||||
import org.apache.hadoop.mapred.JobClient;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Represents the interface to an relational algebra operation like projection,
|
||||
* selection, join, group.
|
||||
*/
|
||||
public abstract class RelationalOperation implements Operation {
|
||||
protected JobConf jobConf;
|
||||
protected HConnection conn;
|
||||
protected HBaseAdmin admin;
|
||||
protected JobClient jobClient;
|
||||
protected HTableDescriptor desc;
|
||||
protected String input;
|
||||
protected String output;
|
||||
protected Map<String, String> condition;
|
||||
protected HTableDescriptor[] tables;
|
||||
protected Set<String> projSet = new HashSet<String>();
|
||||
|
||||
/**
|
||||
* Constructor
|
||||
*
|
||||
* @param conf
|
||||
* @param statements
|
||||
*/
|
||||
public RelationalOperation(HBaseConfiguration conf,
|
||||
Map<String, String> statements) {
|
||||
this.jobConf = new JobConf(conf);
|
||||
this.conn = HConnectionManager.getConnection(conf);
|
||||
this.condition = statements;
|
||||
this.input = statements.get(Constants.CONFIG_INPUT);
|
||||
this.output = statements.get(Constants.CONFIG_OUTPUT);
|
||||
jobConf.setJobName("shell.mapred-" + +System.currentTimeMillis());
|
||||
desc = new HTableDescriptor(output);
|
||||
|
||||
try {
|
||||
this.admin = new HBaseAdmin(conf);
|
||||
this.jobClient = new JobClient(jobConf);
|
||||
tables = conn.listTables();
|
||||
|
||||
ClusterStatus cluster = jobClient.getClusterStatus();
|
||||
jobConf.setNumMapTasks(cluster.getMapTasks());
|
||||
jobConf.setNumReduceTasks(1);
|
||||
} catch (Exception e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Gets the input table column descriptor[]
|
||||
*
|
||||
* @return columns
|
||||
*/
|
||||
public HColumnDescriptor[] getInputColumnDescriptor() {
|
||||
HColumnDescriptor[] columns = null;
|
||||
for (int i = 0; i < tables.length; i++) {
|
||||
if (tables[i].getName().equals(new Text(input))) {
|
||||
columns = tables[i].getFamilies().values().toArray(
|
||||
new HColumnDescriptor[] {});
|
||||
break;
|
||||
}
|
||||
}
|
||||
return columns;
|
||||
}
|
||||
|
||||
/**
|
||||
* Convert HColumnDescriptor[] to String
|
||||
*
|
||||
* @param columns
|
||||
* @return columns string
|
||||
*/
|
||||
public String getColumnStringArray(HColumnDescriptor[] columns) {
|
||||
String result = "";
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
desc.addFamily(columns[i]);
|
||||
result += columns[i].getName() + " ";
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* Creates the output table
|
||||
*
|
||||
* @param columns
|
||||
* @param columnString
|
||||
* @throws IOException
|
||||
*/
|
||||
public void outputTableCreate(HColumnDescriptor[] columns, String columnString)
|
||||
throws IOException {
|
||||
if (columnString == null) {
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
if (projSet.size() > 0) {
|
||||
desc.addFamily(columns[i]);
|
||||
} else {
|
||||
if (projSet.contains(columns[i].getName().toString())) {
|
||||
desc.addFamily(columns[i]);
|
||||
}
|
||||
}
|
||||
}
|
||||
} else {
|
||||
String[] cols = columnString.split(" ");
|
||||
for (int i = 0; i < cols.length; i++) {
|
||||
desc.addFamily(new HColumnDescriptor(cols[i]));
|
||||
}
|
||||
}
|
||||
|
||||
admin.createTable(desc);
|
||||
}
|
||||
|
||||
/**
|
||||
* Return the jobConf
|
||||
*/
|
||||
public JobConf getConf() throws IOException, RuntimeException {
|
||||
return jobConf;
|
||||
}
|
||||
|
||||
/**
|
||||
* @return projection conditions
|
||||
*/
|
||||
public String getProjColumns() {
|
||||
return condition.get(Constants.RELATIONAL_PROJECTION);
|
||||
}
|
||||
|
||||
/**
|
||||
* @return selection conditions
|
||||
*/
|
||||
public String getExpression() {
|
||||
return condition.get(Constants.RELATIONAL_SELECTION);
|
||||
}
|
||||
|
||||
/**
|
||||
* @return group conditions
|
||||
*/
|
||||
public String getGroupColumns() {
|
||||
return condition.get(Constants.RELATIONAL_GROUP);
|
||||
}
|
||||
|
||||
public Operation getOperation() {
|
||||
return this;
|
||||
}
|
||||
}
|
|
@ -1,74 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Map;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableReduce;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
|
||||
/**
|
||||
* Perform a relational selection by linear search algorithm on each Map
|
||||
* Functions.
|
||||
*
|
||||
* Scan each file block an test all records to see whether they satisfy the
|
||||
* selection condition.
|
||||
*/
|
||||
public class Selection extends RelationalOperation {
|
||||
public Selection(HBaseConfiguration conf, Map<String, String> condition) {
|
||||
super(conf, condition);
|
||||
}
|
||||
|
||||
@Override
|
||||
public JobConf getConf() throws IOException, RuntimeException {
|
||||
HColumnDescriptor[] columns = getInputColumnDescriptor();
|
||||
String groupColumns = getGroupColumns(columns);
|
||||
outputTableCreate(columns, null);
|
||||
|
||||
if (condition.containsKey(Constants.RELATIONAL_GROUP)) {
|
||||
GroupingFilterMap.initJob(input, groupColumns, getGroupColumns(),
|
||||
getExpression(), GroupingFilterMap.class, jobConf);
|
||||
} else {
|
||||
IdentityFilterMap.initJob(input, groupColumns, getExpression(),
|
||||
IdentityFilterMap.class, jobConf);
|
||||
}
|
||||
|
||||
IdentityTableReduce.initJob(output, IdentityTableReduce.class, jobConf);
|
||||
return jobConf;
|
||||
}
|
||||
|
||||
private String getGroupColumns(HColumnDescriptor[] columns) {
|
||||
String result = null;
|
||||
if (condition.containsKey(Constants.RELATIONAL_PROJECTION)) {
|
||||
result = getProjColumns();
|
||||
} else {
|
||||
result = getColumnStringArray(columns);
|
||||
}
|
||||
|
||||
String[] groups = result.split(" ");
|
||||
for (int i = 0; i < groups.length; i++) {
|
||||
projSet.add(groups[i]);
|
||||
}
|
||||
return result;
|
||||
}
|
||||
}
|
|
@ -1,774 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. ExpressionParser.java */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.Reader;
|
||||
import java.io.StringReader;
|
||||
import java.io.UnsupportedEncodingException;
|
||||
import java.net.URLDecoder;
|
||||
import java.util.ArrayList;
|
||||
import java.util.HashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.SortedMap;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HTable;
|
||||
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.io.Writable;
|
||||
|
||||
/**
|
||||
* Parsing command line.
|
||||
*/
|
||||
public class ExpressionParser implements ExpressionParserConstants {
|
||||
HBaseConfiguration conf = new HBaseConfiguration();
|
||||
private String expression;
|
||||
private String joinKeyColumn;
|
||||
private String secondRelation;
|
||||
private Map<String, List<String>> unionSet = new HashMap<String, List<String>>();
|
||||
private Map<String, List<String>> intersectionSet = new HashMap<String, List<String>>();
|
||||
private SortedMap<Text, byte[]> secondValue = null;
|
||||
|
||||
public ExpressionParser(final String expression) {
|
||||
this((Reader)(new StringReader(expression)));
|
||||
this.expression = expression;
|
||||
}
|
||||
|
||||
public String getExpression() {
|
||||
return this.expression;
|
||||
}
|
||||
|
||||
final public void booleanExpressionParse() throws ParseException {
|
||||
Map<String, List<String>> temp = new HashMap<String, List<String>>();
|
||||
temp = booleanTerm();
|
||||
label_1:
|
||||
while (true) {
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
case OR:
|
||||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[0] = jj_gen;
|
||||
break label_1;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
jj_consume_token(AND);
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
break;
|
||||
case OR:
|
||||
jj_consume_token(OR);
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
break;
|
||||
default:
|
||||
jj_la1[1] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
}
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
final public void joinExpressionParse() throws ParseException {
|
||||
Map<String, List<String>> temp = new HashMap<String, List<String>>();
|
||||
indexJoinCondition();
|
||||
jj_consume_token(BOOL);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
case INTEGER_LITERAL:
|
||||
temp = booleanTerm();
|
||||
label_2:
|
||||
while (true) {
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
case OR:
|
||||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[2] = jj_gen;
|
||||
break label_2;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
jj_consume_token(AND);
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = intersectionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
intersectionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
break;
|
||||
case OR:
|
||||
jj_consume_token(OR);
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
temp = booleanTerm();
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
temp = null;
|
||||
break;
|
||||
default:
|
||||
jj_la1[3] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
}
|
||||
if(temp != null) {
|
||||
for(Map.Entry<String, List<String>> e : temp.entrySet()) {
|
||||
List<String> newList = unionSet.get(e.getKey());
|
||||
if(newList != null) {
|
||||
newList.addAll(e.getValue());
|
||||
} else {
|
||||
newList = e.getValue();
|
||||
}
|
||||
unionSet.put(e.getKey(), newList);
|
||||
}
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[4] = jj_gen;
|
||||
;
|
||||
}
|
||||
}
|
||||
|
||||
final public void indexJoinCondition() throws ParseException {
|
||||
Token firstR = null;
|
||||
Token joinKey = null;
|
||||
Token secondR = null;
|
||||
firstR = jj_consume_token(ID);
|
||||
jj_consume_token(DOT);
|
||||
joinKey = jj_consume_token(ID);
|
||||
jj_consume_token(EQUALS);
|
||||
secondR = jj_consume_token(ID);
|
||||
jj_consume_token(DOT);
|
||||
jj_consume_token(ROW);
|
||||
joinKeyColumn = joinKey.image.toString();
|
||||
secondRelation = secondR.image.toString();
|
||||
}
|
||||
|
||||
final public Map<String, List<String>> booleanTerm() throws ParseException {
|
||||
Token tSearchName = null;
|
||||
Token tComparator = null;
|
||||
Token tFirst = null;
|
||||
Token tSecond = null;
|
||||
Map<String, List<String>> result = new HashMap<String, List<String>>();
|
||||
List<String> valueList = new ArrayList<String>();
|
||||
String comparand = null;
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
tSearchName = jj_consume_token(ID);
|
||||
break;
|
||||
case INTEGER_LITERAL:
|
||||
tSearchName = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[5] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case DOT:
|
||||
jj_consume_token(DOT);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
tFirst = jj_consume_token(ID);
|
||||
break;
|
||||
case INTEGER_LITERAL:
|
||||
tFirst = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[6] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[7] = jj_gen;
|
||||
;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case EQUALS:
|
||||
tComparator = jj_consume_token(EQUALS);
|
||||
break;
|
||||
case LCOMP:
|
||||
tComparator = jj_consume_token(LCOMP);
|
||||
break;
|
||||
case LCOMPE:
|
||||
tComparator = jj_consume_token(LCOMPE);
|
||||
break;
|
||||
case RCOMPE:
|
||||
tComparator = jj_consume_token(RCOMPE);
|
||||
break;
|
||||
case RCOMP:
|
||||
tComparator = jj_consume_token(RCOMP);
|
||||
break;
|
||||
case IN:
|
||||
tComparator = jj_consume_token(IN);
|
||||
break;
|
||||
case NOTIN:
|
||||
tComparator = jj_consume_token(NOTIN);
|
||||
break;
|
||||
default:
|
||||
jj_la1[8] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
comparand = getValueList();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case DOT:
|
||||
jj_consume_token(DOT);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
tSecond = jj_consume_token(ID);
|
||||
break;
|
||||
case INTEGER_LITERAL:
|
||||
tSecond = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[9] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[10] = jj_gen;
|
||||
;
|
||||
}
|
||||
if(tFirst == null && tSecond == null) {
|
||||
valueList.add(tComparator.image.toString() + " " +comparand);
|
||||
result.put(tSearchName.image.toString(), valueList);
|
||||
} else if (tFirst != null && tSecond != null ){
|
||||
if(tSearchName.image.toString().equals(secondRelation)) {
|
||||
valueList.add(tComparator.image.toString() + " "
|
||||
+ secondValue.get(new Text(tSearchName.image.toString())));
|
||||
result.put(tFirst.image.toString(), valueList);
|
||||
} else {
|
||||
valueList.add(tComparator.image.toString() + " " + tSecond.image.toString());
|
||||
result.put(tFirst.image.toString(), valueList);
|
||||
}
|
||||
}
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public String getValueList() throws ParseException {
|
||||
Token tComparand = null;
|
||||
Token tList = null;
|
||||
String result = "";
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
tComparand = jj_consume_token(ID);
|
||||
break;
|
||||
case INTEGER_LITERAL:
|
||||
tComparand = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[11] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
result = tComparand.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case BAR:
|
||||
jj_consume_token(BAR);
|
||||
label_3:
|
||||
while (true) {
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
case INTEGER_LITERAL:
|
||||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[12] = jj_gen;
|
||||
break label_3;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
tList = jj_consume_token(ID);
|
||||
break;
|
||||
case INTEGER_LITERAL:
|
||||
tList = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[13] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
result += "|" + tList.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case BAR:
|
||||
jj_consume_token(BAR);
|
||||
break;
|
||||
default:
|
||||
jj_la1[14] = jj_gen;
|
||||
;
|
||||
}
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[15] = jj_gen;
|
||||
;
|
||||
}
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public boolean checkConstraints(MapWritable data) throws ParseException {
|
||||
Map<String, String> record = getComparisonObject(data);
|
||||
boolean result = false;
|
||||
if(intersectionSet.size() == 0 && unionSet.size() == 0) {
|
||||
return true;
|
||||
}
|
||||
if (data == null || record.size() == 0) {
|
||||
return result; // return false if data is null.
|
||||
}
|
||||
if (intersectionSet.size() == record.size()) {
|
||||
result = booleanCompare(intersectionSet, record, true);
|
||||
} else if (unionSet.size() == record.size()) {
|
||||
result = booleanCompare(unionSet, record, false);
|
||||
} else {
|
||||
result = getCompareResult(record);
|
||||
}
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public Map<String, String> getComparisonObject(MapWritable data) throws ParseException {
|
||||
Map<String, String> result = new HashMap<String, String>();
|
||||
for (Map.Entry<Writable, Writable> e : data.entrySet()) {
|
||||
String cKey = e.getKey().toString();
|
||||
String val = new String(((ImmutableBytesWritable) e.getValue()).get());
|
||||
|
||||
if (intersectionSet.containsKey(cKey) || unionSet.containsKey(cKey)) {
|
||||
result.put(cKey, val);
|
||||
}
|
||||
}
|
||||
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public MapWritable getJoinColumns(MapWritable value, int numCols, String secondRelation) throws ParseException {
|
||||
MapWritable appendValue = new MapWritable();
|
||||
String joinKey = null;
|
||||
this.secondRelation = secondRelation;
|
||||
if (numCols > 0) {
|
||||
for (Map.Entry<Writable, Writable> e : value.entrySet()) {
|
||||
Text column = (Text) e.getKey();
|
||||
for (int i = 0; i < numCols; i++) {
|
||||
if (column.equals(new Text(joinKeyColumn))) {
|
||||
joinKey = new String(((ImmutableBytesWritable) e.getValue()).get());
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
//If joinKey is null, just return.
|
||||
if(joinKey == null) {
|
||||
{if (true) return appendValue;}
|
||||
}
|
||||
|
||||
try {
|
||||
HTable table = new HTable(conf, new Text(secondRelation));
|
||||
secondValue = table.getRow(new Text(joinKey));
|
||||
for (Map.Entry<Text, byte[]> e : secondValue.entrySet()) {
|
||||
appendValue.put(e.getKey(), new ImmutableBytesWritable(e.getValue()));
|
||||
}
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
{if (true) return appendValue;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public boolean getCompareResult(Map<String, String> record) throws ParseException {
|
||||
{if (true) return (booleanCompare(intersectionSet, record, true) &&
|
||||
booleanCompare(unionSet, record, false)) ? true : false;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public boolean booleanCompare(Map<String, List<String>> expression,
|
||||
Map<String, String> record, boolean isIntersection) throws ParseException {
|
||||
boolean negative = true;
|
||||
boolean positive = false;
|
||||
try{
|
||||
for (Map.Entry<String, List<String>> e : expression.entrySet()) {
|
||||
String key = e.getKey();
|
||||
List<String> valueList = e.getValue();
|
||||
String recordValueList = record.get(key);
|
||||
|
||||
for (int i = 0; i < valueList.size(); i++) {
|
||||
String[] term = valueList.get(i).split(" ");
|
||||
String comparator = term[0];
|
||||
String comparand = term[1];
|
||||
|
||||
switch (comparator.charAt(0)) {
|
||||
case '>':
|
||||
if (isSecond(comparator, "=")) {
|
||||
if (Integer.parseInt(comparand) > Integer
|
||||
.parseInt(recordValueList)) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
} else {
|
||||
if (Integer.parseInt(comparand) > Integer
|
||||
.parseInt(recordValueList)
|
||||
|| comparand.equals(recordValueList)) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
}
|
||||
break;
|
||||
case '<':
|
||||
if (isSecond(comparator, "=")) {
|
||||
if (Integer.parseInt(comparand) < Integer
|
||||
.parseInt(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
} else {
|
||||
if (Integer.parseInt(comparand) < Integer
|
||||
.parseInt(recordValueList)
|
||||
|| comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
|
||||
case '!':
|
||||
if (isSecond(comparator, "!")) {
|
||||
boolean checkBool = true;
|
||||
String[] coms = comparand.split("[|]");
|
||||
for (int j = 0; j < coms.length; j++) {
|
||||
if (URLDecoder.decode(coms[j], "UTF-8").equals(recordValueList)) {
|
||||
checkBool = false;
|
||||
}
|
||||
}
|
||||
|
||||
if (!checkBool) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
|
||||
} else {
|
||||
if (comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
case '=':
|
||||
if (isSecond(comparator, "=")) {
|
||||
|
||||
boolean checkBool = true;
|
||||
String[] coms = comparand.split("[|]");
|
||||
for (int j = 0; j < coms.length; j++) {
|
||||
if (URLDecoder.decode(coms[j], "UTF-8").equals(recordValueList)) {
|
||||
checkBool = false;
|
||||
}
|
||||
}
|
||||
|
||||
if (checkBool) {
|
||||
negative = false;
|
||||
} else {
|
||||
positive = true;
|
||||
}
|
||||
|
||||
} else {
|
||||
if (!comparand.equals(recordValueList))
|
||||
negative = false;
|
||||
else
|
||||
positive = true;
|
||||
}
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
} catch (UnsupportedEncodingException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
boolean result = false;
|
||||
if (isIntersection) {
|
||||
result = negative;
|
||||
} else {
|
||||
result = positive;
|
||||
}
|
||||
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public boolean isSecond(String comparator, String string) throws ParseException {
|
||||
{if (true) return (comparator.length() == 2 && string.charAt(0) == comparator.charAt(1))
|
||||
? true : false;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
public ExpressionParserTokenManager token_source;
|
||||
SimpleCharStream jj_input_stream;
|
||||
public Token token, jj_nt;
|
||||
private int jj_ntk;
|
||||
private int jj_gen;
|
||||
final private int[] jj_la1 = new int[16];
|
||||
static private int[] jj_la1_0;
|
||||
static {
|
||||
jj_la1_0();
|
||||
}
|
||||
private static void jj_la1_0() {
|
||||
jj_la1_0 = new int[] {0xc0,0xc0,0xc0,0xc0,0xc0000,0xc0000,0xc0000,0x100,0xfe00,0xc0000,0x100,0xc0000,0xc0000,0xc0000,0x10000,0x10000,};
|
||||
}
|
||||
|
||||
public ExpressionParser(java.io.InputStream stream) {
|
||||
this(stream, null);
|
||||
}
|
||||
public ExpressionParser(java.io.InputStream stream, String encoding) {
|
||||
try { jj_input_stream = new SimpleCharStream(stream, encoding, 1, 1); } catch(java.io.UnsupportedEncodingException e) { throw new RuntimeException(e); }
|
||||
token_source = new ExpressionParserTokenManager(jj_input_stream);
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
public void ReInit(java.io.InputStream stream) {
|
||||
ReInit(stream, null);
|
||||
}
|
||||
public void ReInit(java.io.InputStream stream, String encoding) {
|
||||
try { jj_input_stream.ReInit(stream, encoding, 1, 1); } catch(java.io.UnsupportedEncodingException e) { throw new RuntimeException(e); }
|
||||
token_source.ReInit(jj_input_stream);
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
public ExpressionParser(java.io.Reader stream) {
|
||||
jj_input_stream = new SimpleCharStream(stream, 1, 1);
|
||||
token_source = new ExpressionParserTokenManager(jj_input_stream);
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
public void ReInit(java.io.Reader stream) {
|
||||
jj_input_stream.ReInit(stream, 1, 1);
|
||||
token_source.ReInit(jj_input_stream);
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
public ExpressionParser(ExpressionParserTokenManager tm) {
|
||||
token_source = tm;
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
public void ReInit(ExpressionParserTokenManager tm) {
|
||||
token_source = tm;
|
||||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 16; i++) jj_la1[i] = -1;
|
||||
}
|
||||
|
||||
final private Token jj_consume_token(int kind) throws ParseException {
|
||||
Token oldToken;
|
||||
if ((oldToken = token).next != null) token = token.next;
|
||||
else token = token.next = token_source.getNextToken();
|
||||
jj_ntk = -1;
|
||||
if (token.kind == kind) {
|
||||
jj_gen++;
|
||||
return token;
|
||||
}
|
||||
token = oldToken;
|
||||
jj_kind = kind;
|
||||
throw generateParseException();
|
||||
}
|
||||
|
||||
final public Token getNextToken() {
|
||||
if (token.next != null) token = token.next;
|
||||
else token = token.next = token_source.getNextToken();
|
||||
jj_ntk = -1;
|
||||
jj_gen++;
|
||||
return token;
|
||||
}
|
||||
|
||||
final public Token getToken(int index) {
|
||||
Token t = token;
|
||||
for (int i = 0; i < index; i++) {
|
||||
if (t.next != null) t = t.next;
|
||||
else t = t.next = token_source.getNextToken();
|
||||
}
|
||||
return t;
|
||||
}
|
||||
|
||||
final private int jj_ntk() {
|
||||
if ((jj_nt=token.next) == null)
|
||||
return (jj_ntk = (token.next=token_source.getNextToken()).kind);
|
||||
else
|
||||
return (jj_ntk = jj_nt.kind);
|
||||
}
|
||||
|
||||
private java.util.Vector jj_expentries = new java.util.Vector();
|
||||
private int[] jj_expentry;
|
||||
private int jj_kind = -1;
|
||||
|
||||
public ParseException generateParseException() {
|
||||
jj_expentries.removeAllElements();
|
||||
boolean[] la1tokens = new boolean[24];
|
||||
for (int i = 0; i < 24; i++) {
|
||||
la1tokens[i] = false;
|
||||
}
|
||||
if (jj_kind >= 0) {
|
||||
la1tokens[jj_kind] = true;
|
||||
jj_kind = -1;
|
||||
}
|
||||
for (int i = 0; i < 16; i++) {
|
||||
if (jj_la1[i] == jj_gen) {
|
||||
for (int j = 0; j < 32; j++) {
|
||||
if ((jj_la1_0[i] & (1<<j)) != 0) {
|
||||
la1tokens[j] = true;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
for (int i = 0; i < 24; i++) {
|
||||
if (la1tokens[i]) {
|
||||
jj_expentry = new int[1];
|
||||
jj_expentry[0] = i;
|
||||
jj_expentries.addElement(jj_expentry);
|
||||
}
|
||||
}
|
||||
int[][] exptokseq = new int[jj_expentries.size()][];
|
||||
for (int i = 0; i < jj_expentries.size(); i++) {
|
||||
exptokseq[i] = (int[])jj_expentries.elementAt(i);
|
||||
}
|
||||
return new ParseException(token, exptokseq, tokenImage);
|
||||
}
|
||||
|
||||
final public void enable_tracing() {
|
||||
}
|
||||
|
||||
final public void disable_tracing() {
|
||||
}
|
||||
|
||||
}
|
|
@ -1,56 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. ExpressionParserConstants.java */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
public interface ExpressionParserConstants {
|
||||
|
||||
int EOF = 0;
|
||||
int ROW = 5;
|
||||
int AND = 6;
|
||||
int OR = 7;
|
||||
int DOT = 8;
|
||||
int EQUALS = 9;
|
||||
int LCOMP = 10;
|
||||
int RCOMP = 11;
|
||||
int LCOMPE = 12;
|
||||
int RCOMPE = 13;
|
||||
int IN = 14;
|
||||
int NOTIN = 15;
|
||||
int BAR = 16;
|
||||
int BOOL = 17;
|
||||
int ID = 18;
|
||||
int INTEGER_LITERAL = 19;
|
||||
int FLOATING_POINT_LITERAL = 20;
|
||||
int EXPONENT = 21;
|
||||
int QUOTED_IDENTIFIER = 22;
|
||||
int STRING_LITERAL = 23;
|
||||
|
||||
int DEFAULT = 0;
|
||||
|
||||
String[] tokenImage = {
|
||||
"<EOF>",
|
||||
"\" \"",
|
||||
"\"\\t\"",
|
||||
"\"\\r\"",
|
||||
"\"\\n\"",
|
||||
"\"row\"",
|
||||
"\"and\"",
|
||||
"\"or\"",
|
||||
"\".\"",
|
||||
"\"=\"",
|
||||
"\">\"",
|
||||
"\"<\"",
|
||||
"\">=\"",
|
||||
"\"<=\"",
|
||||
"\"==\"",
|
||||
"\"!!\"",
|
||||
"\"|\"",
|
||||
"\"bool\"",
|
||||
"<ID>",
|
||||
"<INTEGER_LITERAL>",
|
||||
"<FLOATING_POINT_LITERAL>",
|
||||
"<EXPONENT>",
|
||||
"<QUOTED_IDENTIFIER>",
|
||||
"<STRING_LITERAL>",
|
||||
};
|
||||
|
||||
}
|
|
@ -1,658 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. ExpressionParserTokenManager.java */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.Reader;
|
||||
import java.io.StringReader;
|
||||
import java.io.UnsupportedEncodingException;
|
||||
import java.net.URLDecoder;
|
||||
import java.util.ArrayList;
|
||||
import java.util.HashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.SortedMap;
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HTable;
|
||||
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.io.Writable;
|
||||
|
||||
public class ExpressionParserTokenManager implements ExpressionParserConstants
|
||||
{
|
||||
public java.io.PrintStream debugStream = System.out;
|
||||
public void setDebugStream(java.io.PrintStream ds) { debugStream = ds; }
|
||||
private final int jjStopStringLiteralDfa_0(int pos, long active0)
|
||||
{
|
||||
switch (pos)
|
||||
{
|
||||
case 0:
|
||||
if ((active0 & 0x200e0L) != 0L)
|
||||
{
|
||||
jjmatchedKind = 18;
|
||||
return 1;
|
||||
}
|
||||
if ((active0 & 0x100L) != 0L)
|
||||
return 3;
|
||||
return -1;
|
||||
case 1:
|
||||
if ((active0 & 0x20060L) != 0L)
|
||||
{
|
||||
jjmatchedKind = 18;
|
||||
jjmatchedPos = 1;
|
||||
return 1;
|
||||
}
|
||||
if ((active0 & 0x80L) != 0L)
|
||||
return 1;
|
||||
return -1;
|
||||
case 2:
|
||||
if ((active0 & 0x20000L) != 0L)
|
||||
{
|
||||
jjmatchedKind = 18;
|
||||
jjmatchedPos = 2;
|
||||
return 1;
|
||||
}
|
||||
if ((active0 & 0x60L) != 0L)
|
||||
return 1;
|
||||
return -1;
|
||||
default :
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
private final int jjStartNfa_0(int pos, long active0)
|
||||
{
|
||||
return jjMoveNfa_0(jjStopStringLiteralDfa_0(pos, active0), pos + 1);
|
||||
}
|
||||
private final int jjStopAtPos(int pos, int kind)
|
||||
{
|
||||
jjmatchedKind = kind;
|
||||
jjmatchedPos = pos;
|
||||
return pos + 1;
|
||||
}
|
||||
private final int jjStartNfaWithStates_0(int pos, int kind, int state)
|
||||
{
|
||||
jjmatchedKind = kind;
|
||||
jjmatchedPos = pos;
|
||||
try { curChar = input_stream.readChar(); }
|
||||
catch(java.io.IOException e) { return pos + 1; }
|
||||
return jjMoveNfa_0(state, pos + 1);
|
||||
}
|
||||
private final int jjMoveStringLiteralDfa0_0()
|
||||
{
|
||||
switch(curChar)
|
||||
{
|
||||
case 33:
|
||||
return jjMoveStringLiteralDfa1_0(0x8000L);
|
||||
case 46:
|
||||
return jjStartNfaWithStates_0(0, 8, 3);
|
||||
case 60:
|
||||
jjmatchedKind = 11;
|
||||
return jjMoveStringLiteralDfa1_0(0x2000L);
|
||||
case 61:
|
||||
jjmatchedKind = 9;
|
||||
return jjMoveStringLiteralDfa1_0(0x4000L);
|
||||
case 62:
|
||||
jjmatchedKind = 10;
|
||||
return jjMoveStringLiteralDfa1_0(0x1000L);
|
||||
case 65:
|
||||
case 97:
|
||||
return jjMoveStringLiteralDfa1_0(0x40L);
|
||||
case 66:
|
||||
case 98:
|
||||
return jjMoveStringLiteralDfa1_0(0x20000L);
|
||||
case 79:
|
||||
case 111:
|
||||
return jjMoveStringLiteralDfa1_0(0x80L);
|
||||
case 82:
|
||||
case 114:
|
||||
return jjMoveStringLiteralDfa1_0(0x20L);
|
||||
case 124:
|
||||
return jjStopAtPos(0, 16);
|
||||
default :
|
||||
return jjMoveNfa_0(0, 0);
|
||||
}
|
||||
}
|
||||
private final int jjMoveStringLiteralDfa1_0(long active0)
|
||||
{
|
||||
try { curChar = input_stream.readChar(); }
|
||||
catch(java.io.IOException e) {
|
||||
jjStopStringLiteralDfa_0(0, active0);
|
||||
return 1;
|
||||
}
|
||||
switch(curChar)
|
||||
{
|
||||
case 33:
|
||||
if ((active0 & 0x8000L) != 0L)
|
||||
return jjStopAtPos(1, 15);
|
||||
break;
|
||||
case 61:
|
||||
if ((active0 & 0x1000L) != 0L)
|
||||
return jjStopAtPos(1, 12);
|
||||
else if ((active0 & 0x2000L) != 0L)
|
||||
return jjStopAtPos(1, 13);
|
||||
else if ((active0 & 0x4000L) != 0L)
|
||||
return jjStopAtPos(1, 14);
|
||||
break;
|
||||
case 78:
|
||||
case 110:
|
||||
return jjMoveStringLiteralDfa2_0(active0, 0x40L);
|
||||
case 79:
|
||||
case 111:
|
||||
return jjMoveStringLiteralDfa2_0(active0, 0x20020L);
|
||||
case 82:
|
||||
case 114:
|
||||
if ((active0 & 0x80L) != 0L)
|
||||
return jjStartNfaWithStates_0(1, 7, 1);
|
||||
break;
|
||||
default :
|
||||
break;
|
||||
}
|
||||
return jjStartNfa_0(0, active0);
|
||||
}
|
||||
private final int jjMoveStringLiteralDfa2_0(long old0, long active0)
|
||||
{
|
||||
if (((active0 &= old0)) == 0L)
|
||||
return jjStartNfa_0(0, old0);
|
||||
try { curChar = input_stream.readChar(); }
|
||||
catch(java.io.IOException e) {
|
||||
jjStopStringLiteralDfa_0(1, active0);
|
||||
return 2;
|
||||
}
|
||||
switch(curChar)
|
||||
{
|
||||
case 68:
|
||||
case 100:
|
||||
if ((active0 & 0x40L) != 0L)
|
||||
return jjStartNfaWithStates_0(2, 6, 1);
|
||||
break;
|
||||
case 79:
|
||||
case 111:
|
||||
return jjMoveStringLiteralDfa3_0(active0, 0x20000L);
|
||||
case 87:
|
||||
case 119:
|
||||
if ((active0 & 0x20L) != 0L)
|
||||
return jjStartNfaWithStates_0(2, 5, 1);
|
||||
break;
|
||||
default :
|
||||
break;
|
||||
}
|
||||
return jjStartNfa_0(1, active0);
|
||||
}
|
||||
private final int jjMoveStringLiteralDfa3_0(long old0, long active0)
|
||||
{
|
||||
if (((active0 &= old0)) == 0L)
|
||||
return jjStartNfa_0(1, old0);
|
||||
try { curChar = input_stream.readChar(); }
|
||||
catch(java.io.IOException e) {
|
||||
jjStopStringLiteralDfa_0(2, active0);
|
||||
return 3;
|
||||
}
|
||||
switch(curChar)
|
||||
{
|
||||
case 76:
|
||||
case 108:
|
||||
if ((active0 & 0x20000L) != 0L)
|
||||
return jjStartNfaWithStates_0(3, 17, 1);
|
||||
break;
|
||||
default :
|
||||
break;
|
||||
}
|
||||
return jjStartNfa_0(2, active0);
|
||||
}
|
||||
private final void jjCheckNAdd(int state)
|
||||
{
|
||||
if (jjrounds[state] != jjround)
|
||||
{
|
||||
jjstateSet[jjnewStateCnt++] = state;
|
||||
jjrounds[state] = jjround;
|
||||
}
|
||||
}
|
||||
private final void jjAddStates(int start, int end)
|
||||
{
|
||||
do {
|
||||
jjstateSet[jjnewStateCnt++] = jjnextStates[start];
|
||||
} while (start++ != end);
|
||||
}
|
||||
private final void jjCheckNAddTwoStates(int state1, int state2)
|
||||
{
|
||||
jjCheckNAdd(state1);
|
||||
jjCheckNAdd(state2);
|
||||
}
|
||||
private final void jjCheckNAddStates(int start, int end)
|
||||
{
|
||||
do {
|
||||
jjCheckNAdd(jjnextStates[start]);
|
||||
} while (start++ != end);
|
||||
}
|
||||
private final void jjCheckNAddStates(int start)
|
||||
{
|
||||
jjCheckNAdd(jjnextStates[start]);
|
||||
jjCheckNAdd(jjnextStates[start + 1]);
|
||||
}
|
||||
static final long[] jjbitVec0 = {
|
||||
0x0L, 0x0L, 0xffffffffffffffffL, 0xffffffffffffffffL
|
||||
};
|
||||
private final int jjMoveNfa_0(int startState, int curPos)
|
||||
{
|
||||
int[] nextStates;
|
||||
int startsAt = 0;
|
||||
jjnewStateCnt = 32;
|
||||
int i = 1;
|
||||
jjstateSet[0] = startState;
|
||||
int j, kind = 0x7fffffff;
|
||||
for (;;)
|
||||
{
|
||||
if (++jjround == 0x7fffffff)
|
||||
ReInitRounds();
|
||||
if (curChar < 64)
|
||||
{
|
||||
long l = 1L << curChar;
|
||||
MatchLoop: do
|
||||
{
|
||||
switch(jjstateSet[--i])
|
||||
{
|
||||
case 0:
|
||||
if ((0x3ff000000000000L & l) != 0L)
|
||||
{
|
||||
if (kind > 19)
|
||||
kind = 19;
|
||||
jjCheckNAddStates(0, 6);
|
||||
}
|
||||
else if ((0x400a00000000000L & l) != 0L)
|
||||
{
|
||||
if (kind > 18)
|
||||
kind = 18;
|
||||
jjCheckNAdd(1);
|
||||
}
|
||||
else if (curChar == 39)
|
||||
jjCheckNAddStates(7, 9);
|
||||
else if (curChar == 34)
|
||||
jjCheckNAdd(8);
|
||||
else if (curChar == 46)
|
||||
jjCheckNAdd(3);
|
||||
break;
|
||||
case 1:
|
||||
if ((0x7ffa00000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 18)
|
||||
kind = 18;
|
||||
jjCheckNAdd(1);
|
||||
break;
|
||||
case 2:
|
||||
if (curChar == 46)
|
||||
jjCheckNAdd(3);
|
||||
break;
|
||||
case 3:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAddTwoStates(3, 4);
|
||||
break;
|
||||
case 5:
|
||||
if ((0x280000000000L & l) != 0L)
|
||||
jjCheckNAdd(6);
|
||||
break;
|
||||
case 6:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAdd(6);
|
||||
break;
|
||||
case 7:
|
||||
if (curChar == 34)
|
||||
jjCheckNAdd(8);
|
||||
break;
|
||||
case 8:
|
||||
if ((0xfffffffbffffffffL & l) != 0L)
|
||||
jjCheckNAddTwoStates(8, 9);
|
||||
break;
|
||||
case 9:
|
||||
if (curChar == 34 && kind > 22)
|
||||
kind = 22;
|
||||
break;
|
||||
case 10:
|
||||
if (curChar == 39)
|
||||
jjCheckNAddStates(7, 9);
|
||||
break;
|
||||
case 11:
|
||||
if ((0xffffff7fffffffffL & l) != 0L)
|
||||
jjCheckNAddStates(7, 9);
|
||||
break;
|
||||
case 12:
|
||||
if (curChar == 39)
|
||||
jjCheckNAddStates(10, 12);
|
||||
break;
|
||||
case 13:
|
||||
if (curChar == 39)
|
||||
jjstateSet[jjnewStateCnt++] = 12;
|
||||
break;
|
||||
case 14:
|
||||
if ((0xffffff7fffffffffL & l) != 0L)
|
||||
jjCheckNAddStates(10, 12);
|
||||
break;
|
||||
case 15:
|
||||
if (curChar == 39 && kind > 23)
|
||||
kind = 23;
|
||||
break;
|
||||
case 16:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 19)
|
||||
kind = 19;
|
||||
jjCheckNAddStates(0, 6);
|
||||
break;
|
||||
case 17:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 19)
|
||||
kind = 19;
|
||||
jjCheckNAdd(17);
|
||||
break;
|
||||
case 18:
|
||||
if ((0x3ff000000000000L & l) != 0L)
|
||||
jjCheckNAddTwoStates(18, 19);
|
||||
break;
|
||||
case 19:
|
||||
if (curChar == 46)
|
||||
jjCheckNAdd(20);
|
||||
break;
|
||||
case 20:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAddTwoStates(20, 21);
|
||||
break;
|
||||
case 22:
|
||||
if ((0x280000000000L & l) != 0L)
|
||||
jjCheckNAdd(23);
|
||||
break;
|
||||
case 23:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAdd(23);
|
||||
break;
|
||||
case 24:
|
||||
if ((0x3ff000000000000L & l) != 0L)
|
||||
jjCheckNAddTwoStates(24, 25);
|
||||
break;
|
||||
case 26:
|
||||
if ((0x280000000000L & l) != 0L)
|
||||
jjCheckNAdd(27);
|
||||
break;
|
||||
case 27:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAdd(27);
|
||||
break;
|
||||
case 28:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAddTwoStates(28, 29);
|
||||
break;
|
||||
case 30:
|
||||
if ((0x280000000000L & l) != 0L)
|
||||
jjCheckNAdd(31);
|
||||
break;
|
||||
case 31:
|
||||
if ((0x3ff000000000000L & l) == 0L)
|
||||
break;
|
||||
if (kind > 20)
|
||||
kind = 20;
|
||||
jjCheckNAdd(31);
|
||||
break;
|
||||
default : break;
|
||||
}
|
||||
} while(i != startsAt);
|
||||
}
|
||||
else if (curChar < 128)
|
||||
{
|
||||
long l = 1L << (curChar & 077);
|
||||
MatchLoop: do
|
||||
{
|
||||
switch(jjstateSet[--i])
|
||||
{
|
||||
case 0:
|
||||
case 1:
|
||||
if ((0x7fffffe87fffffeL & l) == 0L)
|
||||
break;
|
||||
if (kind > 18)
|
||||
kind = 18;
|
||||
jjCheckNAdd(1);
|
||||
break;
|
||||
case 4:
|
||||
if ((0x2000000020L & l) != 0L)
|
||||
jjAddStates(13, 14);
|
||||
break;
|
||||
case 8:
|
||||
jjAddStates(15, 16);
|
||||
break;
|
||||
case 11:
|
||||
jjCheckNAddStates(7, 9);
|
||||
break;
|
||||
case 14:
|
||||
jjCheckNAddStates(10, 12);
|
||||
break;
|
||||
case 21:
|
||||
if ((0x2000000020L & l) != 0L)
|
||||
jjAddStates(17, 18);
|
||||
break;
|
||||
case 25:
|
||||
if ((0x2000000020L & l) != 0L)
|
||||
jjAddStates(19, 20);
|
||||
break;
|
||||
case 29:
|
||||
if ((0x2000000020L & l) != 0L)
|
||||
jjAddStates(21, 22);
|
||||
break;
|
||||
default : break;
|
||||
}
|
||||
} while(i != startsAt);
|
||||
}
|
||||
else
|
||||
{
|
||||
int i2 = (curChar & 0xff) >> 6;
|
||||
long l2 = 1L << (curChar & 077);
|
||||
MatchLoop: do
|
||||
{
|
||||
switch(jjstateSet[--i])
|
||||
{
|
||||
case 8:
|
||||
if ((jjbitVec0[i2] & l2) != 0L)
|
||||
jjAddStates(15, 16);
|
||||
break;
|
||||
case 11:
|
||||
if ((jjbitVec0[i2] & l2) != 0L)
|
||||
jjCheckNAddStates(7, 9);
|
||||
break;
|
||||
case 14:
|
||||
if ((jjbitVec0[i2] & l2) != 0L)
|
||||
jjCheckNAddStates(10, 12);
|
||||
break;
|
||||
default : break;
|
||||
}
|
||||
} while(i != startsAt);
|
||||
}
|
||||
if (kind != 0x7fffffff)
|
||||
{
|
||||
jjmatchedKind = kind;
|
||||
jjmatchedPos = curPos;
|
||||
kind = 0x7fffffff;
|
||||
}
|
||||
++curPos;
|
||||
if ((i = jjnewStateCnt) == (startsAt = 32 - (jjnewStateCnt = startsAt)))
|
||||
return curPos;
|
||||
try { curChar = input_stream.readChar(); }
|
||||
catch(java.io.IOException e) { return curPos; }
|
||||
}
|
||||
}
|
||||
static final int[] jjnextStates = {
|
||||
17, 18, 19, 24, 25, 28, 29, 11, 13, 15, 13, 14, 15, 5, 6, 8,
|
||||
9, 22, 23, 26, 27, 30, 31,
|
||||
};
|
||||
public static final String[] jjstrLiteralImages = {
|
||||
"", null, null, null, null, null, null, null, "\56", "\75", "\76", "\74",
|
||||
"\76\75", "\74\75", "\75\75", "\41\41", "\174", null, null, null, null, null, null, null, };
|
||||
public static final String[] lexStateNames = {
|
||||
"DEFAULT",
|
||||
};
|
||||
static final long[] jjtoToken = {
|
||||
0xdfffe1L,
|
||||
};
|
||||
static final long[] jjtoSkip = {
|
||||
0x1eL,
|
||||
};
|
||||
protected SimpleCharStream input_stream;
|
||||
private final int[] jjrounds = new int[32];
|
||||
private final int[] jjstateSet = new int[64];
|
||||
protected char curChar;
|
||||
public ExpressionParserTokenManager(SimpleCharStream stream){
|
||||
if (SimpleCharStream.staticFlag)
|
||||
throw new Error("ERROR: Cannot use a static CharStream class with a non-static lexical analyzer.");
|
||||
input_stream = stream;
|
||||
}
|
||||
public ExpressionParserTokenManager(SimpleCharStream stream, int lexState){
|
||||
this(stream);
|
||||
SwitchTo(lexState);
|
||||
}
|
||||
public void ReInit(SimpleCharStream stream)
|
||||
{
|
||||
jjmatchedPos = jjnewStateCnt = 0;
|
||||
curLexState = defaultLexState;
|
||||
input_stream = stream;
|
||||
ReInitRounds();
|
||||
}
|
||||
private final void ReInitRounds()
|
||||
{
|
||||
int i;
|
||||
jjround = 0x80000001;
|
||||
for (i = 32; i-- > 0;)
|
||||
jjrounds[i] = 0x80000000;
|
||||
}
|
||||
public void ReInit(SimpleCharStream stream, int lexState)
|
||||
{
|
||||
ReInit(stream);
|
||||
SwitchTo(lexState);
|
||||
}
|
||||
public void SwitchTo(int lexState)
|
||||
{
|
||||
if (lexState >= 1 || lexState < 0)
|
||||
throw new TokenMgrError("Error: Ignoring invalid lexical state : " + lexState + ". State unchanged.", TokenMgrError.INVALID_LEXICAL_STATE);
|
||||
else
|
||||
curLexState = lexState;
|
||||
}
|
||||
|
||||
protected Token jjFillToken()
|
||||
{
|
||||
Token t = Token.newToken(jjmatchedKind);
|
||||
t.kind = jjmatchedKind;
|
||||
String im = jjstrLiteralImages[jjmatchedKind];
|
||||
t.image = (im == null) ? input_stream.GetImage() : im;
|
||||
t.beginLine = input_stream.getBeginLine();
|
||||
t.beginColumn = input_stream.getBeginColumn();
|
||||
t.endLine = input_stream.getEndLine();
|
||||
t.endColumn = input_stream.getEndColumn();
|
||||
return t;
|
||||
}
|
||||
|
||||
int curLexState = 0;
|
||||
int defaultLexState = 0;
|
||||
int jjnewStateCnt;
|
||||
int jjround;
|
||||
int jjmatchedPos;
|
||||
int jjmatchedKind;
|
||||
|
||||
public Token getNextToken()
|
||||
{
|
||||
int kind;
|
||||
Token specialToken = null;
|
||||
Token matchedToken;
|
||||
int curPos = 0;
|
||||
|
||||
EOFLoop :
|
||||
for (;;)
|
||||
{
|
||||
try
|
||||
{
|
||||
curChar = input_stream.BeginToken();
|
||||
}
|
||||
catch(java.io.IOException e)
|
||||
{
|
||||
jjmatchedKind = 0;
|
||||
matchedToken = jjFillToken();
|
||||
return matchedToken;
|
||||
}
|
||||
|
||||
try { input_stream.backup(0);
|
||||
while (curChar <= 32 && (0x100002600L & (1L << curChar)) != 0L)
|
||||
curChar = input_stream.BeginToken();
|
||||
}
|
||||
catch (java.io.IOException e1) { continue EOFLoop; }
|
||||
jjmatchedKind = 0x7fffffff;
|
||||
jjmatchedPos = 0;
|
||||
curPos = jjMoveStringLiteralDfa0_0();
|
||||
if (jjmatchedKind != 0x7fffffff)
|
||||
{
|
||||
if (jjmatchedPos + 1 < curPos)
|
||||
input_stream.backup(curPos - jjmatchedPos - 1);
|
||||
if ((jjtoToken[jjmatchedKind >> 6] & (1L << (jjmatchedKind & 077))) != 0L)
|
||||
{
|
||||
matchedToken = jjFillToken();
|
||||
return matchedToken;
|
||||
}
|
||||
else
|
||||
{
|
||||
continue EOFLoop;
|
||||
}
|
||||
}
|
||||
int error_line = input_stream.getEndLine();
|
||||
int error_column = input_stream.getEndColumn();
|
||||
String error_after = null;
|
||||
boolean EOFSeen = false;
|
||||
try { input_stream.readChar(); input_stream.backup(1); }
|
||||
catch (java.io.IOException e1) {
|
||||
EOFSeen = true;
|
||||
error_after = curPos <= 1 ? "" : input_stream.GetImage();
|
||||
if (curChar == '\n' || curChar == '\r') {
|
||||
error_line++;
|
||||
error_column = 0;
|
||||
}
|
||||
else
|
||||
error_column++;
|
||||
}
|
||||
if (!EOFSeen) {
|
||||
input_stream.backup(1);
|
||||
error_after = curPos <= 1 ? "" : input_stream.GetImage();
|
||||
}
|
||||
throw new TokenMgrError(EOFSeen, curLexState, error_line, error_column, error_after, curChar, TokenMgrError.LEXICAL_ERROR);
|
||||
}
|
||||
}
|
||||
|
||||
}
|
|
@ -1,192 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. ParseException.java Version 3.0 */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
/**
|
||||
* This exception is thrown when parse errors are encountered.
|
||||
* You can explicitly create objects of this exception type by
|
||||
* calling the method generateParseException in the generated
|
||||
* parser.
|
||||
*
|
||||
* You can modify this class to customize your error reporting
|
||||
* mechanisms so long as you retain the public fields.
|
||||
*/
|
||||
public class ParseException extends Exception {
|
||||
|
||||
/**
|
||||
* This constructor is used by the method "generateParseException"
|
||||
* in the generated parser. Calling this constructor generates
|
||||
* a new object of this type with the fields "currentToken",
|
||||
* "expectedTokenSequences", and "tokenImage" set. The boolean
|
||||
* flag "specialConstructor" is also set to true to indicate that
|
||||
* this constructor was used to create this object.
|
||||
* This constructor calls its super class with the empty string
|
||||
* to force the "toString" method of parent class "Throwable" to
|
||||
* print the error message in the form:
|
||||
* ParseException: <result of getMessage>
|
||||
*/
|
||||
public ParseException(Token currentTokenVal,
|
||||
int[][] expectedTokenSequencesVal,
|
||||
String[] tokenImageVal
|
||||
)
|
||||
{
|
||||
super("");
|
||||
specialConstructor = true;
|
||||
currentToken = currentTokenVal;
|
||||
expectedTokenSequences = expectedTokenSequencesVal;
|
||||
tokenImage = tokenImageVal;
|
||||
}
|
||||
|
||||
/**
|
||||
* The following constructors are for use by you for whatever
|
||||
* purpose you can think of. Constructing the exception in this
|
||||
* manner makes the exception behave in the normal way - i.e., as
|
||||
* documented in the class "Throwable". The fields "errorToken",
|
||||
* "expectedTokenSequences", and "tokenImage" do not contain
|
||||
* relevant information. The JavaCC generated code does not use
|
||||
* these constructors.
|
||||
*/
|
||||
|
||||
public ParseException() {
|
||||
super();
|
||||
specialConstructor = false;
|
||||
}
|
||||
|
||||
public ParseException(String message) {
|
||||
super(message);
|
||||
specialConstructor = false;
|
||||
}
|
||||
|
||||
/**
|
||||
* This variable determines which constructor was used to create
|
||||
* this object and thereby affects the semantics of the
|
||||
* "getMessage" method (see below).
|
||||
*/
|
||||
protected boolean specialConstructor;
|
||||
|
||||
/**
|
||||
* This is the last token that has been consumed successfully. If
|
||||
* this object has been created due to a parse error, the token
|
||||
* followng this token will (therefore) be the first error token.
|
||||
*/
|
||||
public Token currentToken;
|
||||
|
||||
/**
|
||||
* Each entry in this array is an array of integers. Each array
|
||||
* of integers represents a sequence of tokens (by their ordinal
|
||||
* values) that is expected at this point of the parse.
|
||||
*/
|
||||
public int[][] expectedTokenSequences;
|
||||
|
||||
/**
|
||||
* This is a reference to the "tokenImage" array of the generated
|
||||
* parser within which the parse error occurred. This array is
|
||||
* defined in the generated ...Constants interface.
|
||||
*/
|
||||
public String[] tokenImage;
|
||||
|
||||
/**
|
||||
* This method has the standard behavior when this object has been
|
||||
* created using the standard constructors. Otherwise, it uses
|
||||
* "currentToken" and "expectedTokenSequences" to generate a parse
|
||||
* error message and returns it. If this object has been created
|
||||
* due to a parse error, and you do not catch it (it gets thrown
|
||||
* from the parser), then this method is called during the printing
|
||||
* of the final stack trace, and hence the correct error message
|
||||
* gets displayed.
|
||||
*/
|
||||
public String getMessage() {
|
||||
if (!specialConstructor) {
|
||||
return super.getMessage();
|
||||
}
|
||||
StringBuffer expected = new StringBuffer();
|
||||
int maxSize = 0;
|
||||
for (int i = 0; i < expectedTokenSequences.length; i++) {
|
||||
if (maxSize < expectedTokenSequences[i].length) {
|
||||
maxSize = expectedTokenSequences[i].length;
|
||||
}
|
||||
for (int j = 0; j < expectedTokenSequences[i].length; j++) {
|
||||
expected.append(tokenImage[expectedTokenSequences[i][j]]).append(" ");
|
||||
}
|
||||
if (expectedTokenSequences[i][expectedTokenSequences[i].length - 1] != 0) {
|
||||
expected.append("...");
|
||||
}
|
||||
expected.append(eol).append(" ");
|
||||
}
|
||||
String retval = "Encountered \"";
|
||||
Token tok = currentToken.next;
|
||||
for (int i = 0; i < maxSize; i++) {
|
||||
if (i != 0) retval += " ";
|
||||
if (tok.kind == 0) {
|
||||
retval += tokenImage[0];
|
||||
break;
|
||||
}
|
||||
retval += add_escapes(tok.image);
|
||||
tok = tok.next;
|
||||
}
|
||||
retval += "\" at line " + currentToken.next.beginLine + ", column " + currentToken.next.beginColumn;
|
||||
retval += "." + eol;
|
||||
if (expectedTokenSequences.length == 1) {
|
||||
retval += "Was expecting:" + eol + " ";
|
||||
} else {
|
||||
retval += "Was expecting one of:" + eol + " ";
|
||||
}
|
||||
retval += expected.toString();
|
||||
return retval;
|
||||
}
|
||||
|
||||
/**
|
||||
* The end of line string for this machine.
|
||||
*/
|
||||
protected String eol = System.getProperty("line.separator", "\n");
|
||||
|
||||
/**
|
||||
* Used to convert raw characters to their escaped version
|
||||
* when these raw version cannot be used as part of an ASCII
|
||||
* string literal.
|
||||
*/
|
||||
protected String add_escapes(String str) {
|
||||
StringBuffer retval = new StringBuffer();
|
||||
char ch;
|
||||
for (int i = 0; i < str.length(); i++) {
|
||||
switch (str.charAt(i))
|
||||
{
|
||||
case 0 :
|
||||
continue;
|
||||
case '\b':
|
||||
retval.append("\\b");
|
||||
continue;
|
||||
case '\t':
|
||||
retval.append("\\t");
|
||||
continue;
|
||||
case '\n':
|
||||
retval.append("\\n");
|
||||
continue;
|
||||
case '\f':
|
||||
retval.append("\\f");
|
||||
continue;
|
||||
case '\r':
|
||||
retval.append("\\r");
|
||||
continue;
|
||||
case '\"':
|
||||
retval.append("\\\"");
|
||||
continue;
|
||||
case '\'':
|
||||
retval.append("\\\'");
|
||||
continue;
|
||||
case '\\':
|
||||
retval.append("\\\\");
|
||||
continue;
|
||||
default:
|
||||
if ((ch = str.charAt(i)) < 0x20 || ch > 0x7e) {
|
||||
String s = "0000" + Integer.toString(ch, 16);
|
||||
retval.append("\\u" + s.substring(s.length() - 4, s.length()));
|
||||
} else {
|
||||
retval.append(ch);
|
||||
}
|
||||
continue;
|
||||
}
|
||||
}
|
||||
return retval.toString();
|
||||
}
|
||||
|
||||
}
|
|
@ -1,439 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. SimpleCharStream.java Version 4.0 */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
/**
|
||||
* An implementation of interface CharStream, where the stream is assumed to
|
||||
* contain only ASCII characters (without unicode processing).
|
||||
*/
|
||||
|
||||
public class SimpleCharStream
|
||||
{
|
||||
public static final boolean staticFlag = false;
|
||||
int bufsize;
|
||||
int available;
|
||||
int tokenBegin;
|
||||
public int bufpos = -1;
|
||||
protected int bufline[];
|
||||
protected int bufcolumn[];
|
||||
|
||||
protected int column = 0;
|
||||
protected int line = 1;
|
||||
|
||||
protected boolean prevCharIsCR = false;
|
||||
protected boolean prevCharIsLF = false;
|
||||
|
||||
protected java.io.Reader inputStream;
|
||||
|
||||
protected char[] buffer;
|
||||
protected int maxNextCharInd = 0;
|
||||
protected int inBuf = 0;
|
||||
protected int tabSize = 8;
|
||||
|
||||
protected void setTabSize(int i) { tabSize = i; }
|
||||
protected int getTabSize(int i) { return tabSize; }
|
||||
|
||||
|
||||
protected void ExpandBuff(boolean wrapAround)
|
||||
{
|
||||
char[] newbuffer = new char[bufsize + 2048];
|
||||
int newbufline[] = new int[bufsize + 2048];
|
||||
int newbufcolumn[] = new int[bufsize + 2048];
|
||||
|
||||
try
|
||||
{
|
||||
if (wrapAround)
|
||||
{
|
||||
System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
|
||||
System.arraycopy(buffer, 0, newbuffer,
|
||||
bufsize - tokenBegin, bufpos);
|
||||
buffer = newbuffer;
|
||||
|
||||
System.arraycopy(bufline, tokenBegin, newbufline, 0, bufsize - tokenBegin);
|
||||
System.arraycopy(bufline, 0, newbufline, bufsize - tokenBegin, bufpos);
|
||||
bufline = newbufline;
|
||||
|
||||
System.arraycopy(bufcolumn, tokenBegin, newbufcolumn, 0, bufsize - tokenBegin);
|
||||
System.arraycopy(bufcolumn, 0, newbufcolumn, bufsize - tokenBegin, bufpos);
|
||||
bufcolumn = newbufcolumn;
|
||||
|
||||
maxNextCharInd = (bufpos += (bufsize - tokenBegin));
|
||||
}
|
||||
else
|
||||
{
|
||||
System.arraycopy(buffer, tokenBegin, newbuffer, 0, bufsize - tokenBegin);
|
||||
buffer = newbuffer;
|
||||
|
||||
System.arraycopy(bufline, tokenBegin, newbufline, 0, bufsize - tokenBegin);
|
||||
bufline = newbufline;
|
||||
|
||||
System.arraycopy(bufcolumn, tokenBegin, newbufcolumn, 0, bufsize - tokenBegin);
|
||||
bufcolumn = newbufcolumn;
|
||||
|
||||
maxNextCharInd = (bufpos -= tokenBegin);
|
||||
}
|
||||
}
|
||||
catch (Throwable t)
|
||||
{
|
||||
throw new Error(t.getMessage());
|
||||
}
|
||||
|
||||
|
||||
bufsize += 2048;
|
||||
available = bufsize;
|
||||
tokenBegin = 0;
|
||||
}
|
||||
|
||||
protected void FillBuff() throws java.io.IOException
|
||||
{
|
||||
if (maxNextCharInd == available)
|
||||
{
|
||||
if (available == bufsize)
|
||||
{
|
||||
if (tokenBegin > 2048)
|
||||
{
|
||||
bufpos = maxNextCharInd = 0;
|
||||
available = tokenBegin;
|
||||
}
|
||||
else if (tokenBegin < 0)
|
||||
bufpos = maxNextCharInd = 0;
|
||||
else
|
||||
ExpandBuff(false);
|
||||
}
|
||||
else if (available > tokenBegin)
|
||||
available = bufsize;
|
||||
else if ((tokenBegin - available) < 2048)
|
||||
ExpandBuff(true);
|
||||
else
|
||||
available = tokenBegin;
|
||||
}
|
||||
|
||||
int i;
|
||||
try {
|
||||
if ((i = inputStream.read(buffer, maxNextCharInd,
|
||||
available - maxNextCharInd)) == -1)
|
||||
{
|
||||
inputStream.close();
|
||||
throw new java.io.IOException();
|
||||
}
|
||||
else
|
||||
maxNextCharInd += i;
|
||||
return;
|
||||
}
|
||||
catch(java.io.IOException e) {
|
||||
--bufpos;
|
||||
backup(0);
|
||||
if (tokenBegin == -1)
|
||||
tokenBegin = bufpos;
|
||||
throw e;
|
||||
}
|
||||
}
|
||||
|
||||
public char BeginToken() throws java.io.IOException
|
||||
{
|
||||
tokenBegin = -1;
|
||||
char c = readChar();
|
||||
tokenBegin = bufpos;
|
||||
|
||||
return c;
|
||||
}
|
||||
|
||||
protected void UpdateLineColumn(char c)
|
||||
{
|
||||
column++;
|
||||
|
||||
if (prevCharIsLF)
|
||||
{
|
||||
prevCharIsLF = false;
|
||||
line += (column = 1);
|
||||
}
|
||||
else if (prevCharIsCR)
|
||||
{
|
||||
prevCharIsCR = false;
|
||||
if (c == '\n')
|
||||
{
|
||||
prevCharIsLF = true;
|
||||
}
|
||||
else
|
||||
line += (column = 1);
|
||||
}
|
||||
|
||||
switch (c)
|
||||
{
|
||||
case '\r' :
|
||||
prevCharIsCR = true;
|
||||
break;
|
||||
case '\n' :
|
||||
prevCharIsLF = true;
|
||||
break;
|
||||
case '\t' :
|
||||
column--;
|
||||
column += (tabSize - (column % tabSize));
|
||||
break;
|
||||
default :
|
||||
break;
|
||||
}
|
||||
|
||||
bufline[bufpos] = line;
|
||||
bufcolumn[bufpos] = column;
|
||||
}
|
||||
|
||||
public char readChar() throws java.io.IOException
|
||||
{
|
||||
if (inBuf > 0)
|
||||
{
|
||||
--inBuf;
|
||||
|
||||
if (++bufpos == bufsize)
|
||||
bufpos = 0;
|
||||
|
||||
return buffer[bufpos];
|
||||
}
|
||||
|
||||
if (++bufpos >= maxNextCharInd)
|
||||
FillBuff();
|
||||
|
||||
char c = buffer[bufpos];
|
||||
|
||||
UpdateLineColumn(c);
|
||||
return (c);
|
||||
}
|
||||
|
||||
/**
|
||||
* @deprecated
|
||||
* @see #getEndColumn
|
||||
*/
|
||||
|
||||
public int getColumn() {
|
||||
return bufcolumn[bufpos];
|
||||
}
|
||||
|
||||
/**
|
||||
* @deprecated
|
||||
* @see #getEndLine
|
||||
*/
|
||||
|
||||
public int getLine() {
|
||||
return bufline[bufpos];
|
||||
}
|
||||
|
||||
public int getEndColumn() {
|
||||
return bufcolumn[bufpos];
|
||||
}
|
||||
|
||||
public int getEndLine() {
|
||||
return bufline[bufpos];
|
||||
}
|
||||
|
||||
public int getBeginColumn() {
|
||||
return bufcolumn[tokenBegin];
|
||||
}
|
||||
|
||||
public int getBeginLine() {
|
||||
return bufline[tokenBegin];
|
||||
}
|
||||
|
||||
public void backup(int amount) {
|
||||
|
||||
inBuf += amount;
|
||||
if ((bufpos -= amount) < 0)
|
||||
bufpos += bufsize;
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.Reader dstream, int startline,
|
||||
int startcolumn, int buffersize)
|
||||
{
|
||||
inputStream = dstream;
|
||||
line = startline;
|
||||
column = startcolumn - 1;
|
||||
|
||||
available = bufsize = buffersize;
|
||||
buffer = new char[buffersize];
|
||||
bufline = new int[buffersize];
|
||||
bufcolumn = new int[buffersize];
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.Reader dstream, int startline,
|
||||
int startcolumn)
|
||||
{
|
||||
this(dstream, startline, startcolumn, 4096);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.Reader dstream)
|
||||
{
|
||||
this(dstream, 1, 1, 4096);
|
||||
}
|
||||
public void ReInit(java.io.Reader dstream, int startline,
|
||||
int startcolumn, int buffersize)
|
||||
{
|
||||
inputStream = dstream;
|
||||
line = startline;
|
||||
column = startcolumn - 1;
|
||||
|
||||
if (buffer == null || buffersize != buffer.length)
|
||||
{
|
||||
available = bufsize = buffersize;
|
||||
buffer = new char[buffersize];
|
||||
bufline = new int[buffersize];
|
||||
bufcolumn = new int[buffersize];
|
||||
}
|
||||
prevCharIsLF = prevCharIsCR = false;
|
||||
tokenBegin = inBuf = maxNextCharInd = 0;
|
||||
bufpos = -1;
|
||||
}
|
||||
|
||||
public void ReInit(java.io.Reader dstream, int startline,
|
||||
int startcolumn)
|
||||
{
|
||||
ReInit(dstream, startline, startcolumn, 4096);
|
||||
}
|
||||
|
||||
public void ReInit(java.io.Reader dstream)
|
||||
{
|
||||
ReInit(dstream, 1, 1, 4096);
|
||||
}
|
||||
public SimpleCharStream(java.io.InputStream dstream, String encoding, int startline,
|
||||
int startcolumn, int buffersize) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
this(encoding == null ? new java.io.InputStreamReader(dstream) : new java.io.InputStreamReader(dstream, encoding), startline, startcolumn, buffersize);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.InputStream dstream, int startline,
|
||||
int startcolumn, int buffersize)
|
||||
{
|
||||
this(new java.io.InputStreamReader(dstream), startline, startcolumn, buffersize);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.InputStream dstream, String encoding, int startline,
|
||||
int startcolumn) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
this(dstream, encoding, startline, startcolumn, 4096);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.InputStream dstream, int startline,
|
||||
int startcolumn)
|
||||
{
|
||||
this(dstream, startline, startcolumn, 4096);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.InputStream dstream, String encoding) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
this(dstream, encoding, 1, 1, 4096);
|
||||
}
|
||||
|
||||
public SimpleCharStream(java.io.InputStream dstream)
|
||||
{
|
||||
this(dstream, 1, 1, 4096);
|
||||
}
|
||||
|
||||
public void ReInit(java.io.InputStream dstream, String encoding, int startline,
|
||||
int startcolumn, int buffersize) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
ReInit(encoding == null ? new java.io.InputStreamReader(dstream) : new java.io.InputStreamReader(dstream, encoding), startline, startcolumn, buffersize);
|
||||
}
|
||||
|
||||
public void ReInit(java.io.InputStream dstream, int startline,
|
||||
int startcolumn, int buffersize)
|
||||
{
|
||||
ReInit(new java.io.InputStreamReader(dstream), startline, startcolumn, buffersize);
|
||||
}
|
||||
|
||||
public void ReInit(java.io.InputStream dstream, String encoding) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
ReInit(dstream, encoding, 1, 1, 4096);
|
||||
}
|
||||
|
||||
public void ReInit(java.io.InputStream dstream)
|
||||
{
|
||||
ReInit(dstream, 1, 1, 4096);
|
||||
}
|
||||
public void ReInit(java.io.InputStream dstream, String encoding, int startline,
|
||||
int startcolumn) throws java.io.UnsupportedEncodingException
|
||||
{
|
||||
ReInit(dstream, encoding, startline, startcolumn, 4096);
|
||||
}
|
||||
public void ReInit(java.io.InputStream dstream, int startline,
|
||||
int startcolumn)
|
||||
{
|
||||
ReInit(dstream, startline, startcolumn, 4096);
|
||||
}
|
||||
public String GetImage()
|
||||
{
|
||||
if (bufpos >= tokenBegin)
|
||||
return new String(buffer, tokenBegin, bufpos - tokenBegin + 1);
|
||||
else
|
||||
return new String(buffer, tokenBegin, bufsize - tokenBegin) +
|
||||
new String(buffer, 0, bufpos + 1);
|
||||
}
|
||||
|
||||
public char[] GetSuffix(int len)
|
||||
{
|
||||
char[] ret = new char[len];
|
||||
|
||||
if ((bufpos + 1) >= len)
|
||||
System.arraycopy(buffer, bufpos - len + 1, ret, 0, len);
|
||||
else
|
||||
{
|
||||
System.arraycopy(buffer, bufsize - (len - bufpos - 1), ret, 0,
|
||||
len - bufpos - 1);
|
||||
System.arraycopy(buffer, 0, ret, len - bufpos - 1, bufpos + 1);
|
||||
}
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
public void Done()
|
||||
{
|
||||
buffer = null;
|
||||
bufline = null;
|
||||
bufcolumn = null;
|
||||
}
|
||||
|
||||
/**
|
||||
* Method to adjust line and column numbers for the start of a token.
|
||||
*/
|
||||
public void adjustBeginLineColumn(int newLine, int newCol)
|
||||
{
|
||||
int start = tokenBegin;
|
||||
int len;
|
||||
|
||||
if (bufpos >= tokenBegin)
|
||||
{
|
||||
len = bufpos - tokenBegin + inBuf + 1;
|
||||
}
|
||||
else
|
||||
{
|
||||
len = bufsize - tokenBegin + bufpos + 1 + inBuf;
|
||||
}
|
||||
|
||||
int i = 0, j = 0, k = 0;
|
||||
int nextColDiff = 0, columnDiff = 0;
|
||||
|
||||
while (i < len &&
|
||||
bufline[j = start % bufsize] == bufline[k = ++start % bufsize])
|
||||
{
|
||||
bufline[j] = newLine;
|
||||
nextColDiff = columnDiff + bufcolumn[k] - bufcolumn[j];
|
||||
bufcolumn[j] = newCol + columnDiff;
|
||||
columnDiff = nextColDiff;
|
||||
i++;
|
||||
}
|
||||
|
||||
if (i < len)
|
||||
{
|
||||
bufline[j] = newLine++;
|
||||
bufcolumn[j] = newCol + columnDiff;
|
||||
|
||||
while (i++ < len)
|
||||
{
|
||||
if (bufline[j = start % bufsize] != bufline[++start % bufsize])
|
||||
bufline[j] = newLine++;
|
||||
else
|
||||
bufline[j] = newLine;
|
||||
}
|
||||
}
|
||||
|
||||
line = bufline[j];
|
||||
column = bufcolumn[j];
|
||||
}
|
||||
|
||||
}
|
|
@ -1,81 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. Token.java Version 3.0 */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
/**
|
||||
* Describes the input token stream.
|
||||
*/
|
||||
|
||||
public class Token {
|
||||
|
||||
/**
|
||||
* An integer that describes the kind of this token. This numbering
|
||||
* system is determined by JavaCCParser, and a table of these numbers is
|
||||
* stored in the file ...Constants.java.
|
||||
*/
|
||||
public int kind;
|
||||
|
||||
/**
|
||||
* beginLine and beginColumn describe the position of the first character
|
||||
* of this token; endLine and endColumn describe the position of the
|
||||
* last character of this token.
|
||||
*/
|
||||
public int beginLine, beginColumn, endLine, endColumn;
|
||||
|
||||
/**
|
||||
* The string image of the token.
|
||||
*/
|
||||
public String image;
|
||||
|
||||
/**
|
||||
* A reference to the next regular (non-special) token from the input
|
||||
* stream. If this is the last token from the input stream, or if the
|
||||
* token manager has not read tokens beyond this one, this field is
|
||||
* set to null. This is true only if this token is also a regular
|
||||
* token. Otherwise, see below for a description of the contents of
|
||||
* this field.
|
||||
*/
|
||||
public Token next;
|
||||
|
||||
/**
|
||||
* This field is used to access special tokens that occur prior to this
|
||||
* token, but after the immediately preceding regular (non-special) token.
|
||||
* If there are no such special tokens, this field is set to null.
|
||||
* When there are more than one such special token, this field refers
|
||||
* to the last of these special tokens, which in turn refers to the next
|
||||
* previous special token through its specialToken field, and so on
|
||||
* until the first special token (whose specialToken field is null).
|
||||
* The next fields of special tokens refer to other special tokens that
|
||||
* immediately follow it (without an intervening regular token). If there
|
||||
* is no such token, this field is null.
|
||||
*/
|
||||
public Token specialToken;
|
||||
|
||||
/**
|
||||
* Returns the image.
|
||||
*/
|
||||
public String toString()
|
||||
{
|
||||
return image;
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns a new Token object, by default. However, if you want, you
|
||||
* can create and return subclass objects based on the value of ofKind.
|
||||
* Simply add the cases to the switch for all those special cases.
|
||||
* For example, if you have a subclass of Token called IDToken that
|
||||
* you want to create if ofKind is ID, simlpy add something like :
|
||||
*
|
||||
* case MyParserConstants.ID : return new IDToken();
|
||||
*
|
||||
* to the following switch statement. Then you can cast matchedToken
|
||||
* variable to the appropriate type and use it in your lexical actions.
|
||||
*/
|
||||
public static final Token newToken(int ofKind)
|
||||
{
|
||||
switch(ofKind)
|
||||
{
|
||||
default : return new Token();
|
||||
}
|
||||
}
|
||||
|
||||
}
|
|
@ -1,133 +0,0 @@
|
|||
/* Generated By:JavaCC: Do not edit this line. TokenMgrError.java Version 3.0 */
|
||||
package org.apache.hadoop.hbase.shell.algebra.generated;
|
||||
|
||||
public class TokenMgrError extends Error
|
||||
{
|
||||
/*
|
||||
* Ordinals for various reasons why an Error of this type can be thrown.
|
||||
*/
|
||||
|
||||
/**
|
||||
* Lexical error occured.
|
||||
*/
|
||||
static final int LEXICAL_ERROR = 0;
|
||||
|
||||
/**
|
||||
* An attempt wass made to create a second instance of a static token manager.
|
||||
*/
|
||||
static final int STATIC_LEXER_ERROR = 1;
|
||||
|
||||
/**
|
||||
* Tried to change to an invalid lexical state.
|
||||
*/
|
||||
static final int INVALID_LEXICAL_STATE = 2;
|
||||
|
||||
/**
|
||||
* Detected (and bailed out of) an infinite loop in the token manager.
|
||||
*/
|
||||
static final int LOOP_DETECTED = 3;
|
||||
|
||||
/**
|
||||
* Indicates the reason why the exception is thrown. It will have
|
||||
* one of the above 4 values.
|
||||
*/
|
||||
int errorCode;
|
||||
|
||||
/**
|
||||
* Replaces unprintable characters by their espaced (or unicode escaped)
|
||||
* equivalents in the given string
|
||||
*/
|
||||
protected static final String addEscapes(String str) {
|
||||
StringBuffer retval = new StringBuffer();
|
||||
char ch;
|
||||
for (int i = 0; i < str.length(); i++) {
|
||||
switch (str.charAt(i))
|
||||
{
|
||||
case 0 :
|
||||
continue;
|
||||
case '\b':
|
||||
retval.append("\\b");
|
||||
continue;
|
||||
case '\t':
|
||||
retval.append("\\t");
|
||||
continue;
|
||||
case '\n':
|
||||
retval.append("\\n");
|
||||
continue;
|
||||
case '\f':
|
||||
retval.append("\\f");
|
||||
continue;
|
||||
case '\r':
|
||||
retval.append("\\r");
|
||||
continue;
|
||||
case '\"':
|
||||
retval.append("\\\"");
|
||||
continue;
|
||||
case '\'':
|
||||
retval.append("\\\'");
|
||||
continue;
|
||||
case '\\':
|
||||
retval.append("\\\\");
|
||||
continue;
|
||||
default:
|
||||
if ((ch = str.charAt(i)) < 0x20 || ch > 0x7e) {
|
||||
String s = "0000" + Integer.toString(ch, 16);
|
||||
retval.append("\\u" + s.substring(s.length() - 4, s.length()));
|
||||
} else {
|
||||
retval.append(ch);
|
||||
}
|
||||
continue;
|
||||
}
|
||||
}
|
||||
return retval.toString();
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns a detailed message for the Error when it is thrown by the
|
||||
* token manager to indicate a lexical error.
|
||||
* Parameters :
|
||||
* EOFSeen : indicates if EOF caused the lexicl error
|
||||
* curLexState : lexical state in which this error occured
|
||||
* errorLine : line number when the error occured
|
||||
* errorColumn : column number when the error occured
|
||||
* errorAfter : prefix that was seen before this error occured
|
||||
* curchar : the offending character
|
||||
* Note: You can customize the lexical error message by modifying this method.
|
||||
*/
|
||||
protected static String LexicalError(boolean EOFSeen, int lexState, int errorLine, int errorColumn, String errorAfter, char curChar) {
|
||||
return("Lexical error at line " +
|
||||
errorLine + ", column " +
|
||||
errorColumn + ". Encountered: " +
|
||||
(EOFSeen ? "<EOF> " : ("\"" + addEscapes(String.valueOf(curChar)) + "\"") + " (" + (int)curChar + "), ") +
|
||||
"after : \"" + addEscapes(errorAfter) + "\"");
|
||||
}
|
||||
|
||||
/**
|
||||
* You can also modify the body of this method to customize your error messages.
|
||||
* For example, cases like LOOP_DETECTED and INVALID_LEXICAL_STATE are not
|
||||
* of end-users concern, so you can return something like :
|
||||
*
|
||||
* "Internal Error : Please file a bug report .... "
|
||||
*
|
||||
* from this method for such cases in the release version of your parser.
|
||||
*/
|
||||
public String getMessage() {
|
||||
return super.getMessage();
|
||||
}
|
||||
|
||||
/*
|
||||
* Constructors of various flavors follow.
|
||||
*/
|
||||
|
||||
public TokenMgrError() {
|
||||
}
|
||||
|
||||
public TokenMgrError(String message, int reason) {
|
||||
super(message);
|
||||
errorCode = reason;
|
||||
}
|
||||
|
||||
public TokenMgrError(boolean EOFSeen, int lexState, int errorLine, int errorColumn, String errorAfter, char curChar, int reason) {
|
||||
this(LexicalError(EOFSeen, lexState, errorLine, errorColumn, errorAfter, curChar), reason);
|
||||
}
|
||||
}
|
|
@ -75,9 +75,7 @@ public class Parser implements ParserConstants {
|
|||
case SELECT:
|
||||
case ENABLE:
|
||||
case DISABLE:
|
||||
case SAVE:
|
||||
case ID:
|
||||
case 71:
|
||||
case 65:
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case HELP:
|
||||
case ALTER:
|
||||
|
@ -95,15 +93,13 @@ public class Parser implements ParserConstants {
|
|||
case SELECT:
|
||||
case ENABLE:
|
||||
case DISABLE:
|
||||
case SAVE:
|
||||
case ID:
|
||||
statement = cmdStatement();
|
||||
break;
|
||||
default:
|
||||
jj_la1[0] = jj_gen;
|
||||
;
|
||||
}
|
||||
jj_consume_token(71);
|
||||
jj_consume_token(65);
|
||||
break;
|
||||
case 0:
|
||||
jj_consume_token(0);
|
||||
|
@ -166,12 +162,6 @@ public class Parser implements ParserConstants {
|
|||
case JAR:
|
||||
cmd = jarCommand();
|
||||
break;
|
||||
case ID:
|
||||
cmd = substituteCommand();
|
||||
break;
|
||||
case SAVE:
|
||||
cmd = saveCommand();
|
||||
break;
|
||||
default:
|
||||
jj_la1[2] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
|
@ -220,7 +210,6 @@ public class Parser implements ParserConstants {
|
|||
label_2:
|
||||
while (true) {
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case DOT:
|
||||
case ID:
|
||||
case INTEGER_LITERAL:
|
||||
case FLOATING_POINT_LITERAL:
|
||||
|
@ -237,9 +226,6 @@ public class Parser implements ParserConstants {
|
|||
case INTEGER_LITERAL:
|
||||
t = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
case DOT:
|
||||
t = jj_consume_token(DOT);
|
||||
break;
|
||||
case FLOATING_POINT_LITERAL:
|
||||
t = jj_consume_token(FLOATING_POINT_LITERAL);
|
||||
break;
|
||||
|
@ -304,8 +290,6 @@ public class Parser implements ParserConstants {
|
|||
case INSERT:
|
||||
case DELETE:
|
||||
case SELECT:
|
||||
case SAVE:
|
||||
case GROUP:
|
||||
case ID:
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case SHOW:
|
||||
|
@ -344,12 +328,6 @@ public class Parser implements ParserConstants {
|
|||
case JAR:
|
||||
t = jj_consume_token(JAR);
|
||||
break;
|
||||
case GROUP:
|
||||
t = jj_consume_token(GROUP);
|
||||
break;
|
||||
case SAVE:
|
||||
t = jj_consume_token(SAVE);
|
||||
break;
|
||||
case ID:
|
||||
t = jj_consume_token(ID);
|
||||
break;
|
||||
|
@ -374,8 +352,6 @@ public class Parser implements ParserConstants {
|
|||
String argument = null;
|
||||
jj_consume_token(SHOW);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case META_TABLE:
|
||||
case ROOT_TABLE:
|
||||
case ID:
|
||||
case QUOTED_IDENTIFIER:
|
||||
case STRING_LITERAL:
|
||||
|
@ -808,104 +784,6 @@ public class Parser implements ParserConstants {
|
|||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public SubstituteCommand substituteCommand() throws ParseException {
|
||||
Token key = null;
|
||||
Token chainKey = null;
|
||||
Token operation = null;
|
||||
String tableName = null;
|
||||
String condition = "";
|
||||
List<String> notInList = new ArrayList<String>();
|
||||
SubstituteCommand substitute = new SubstituteCommand(this.out);
|
||||
Token extendedKey = null;
|
||||
key = jj_consume_token(ID);
|
||||
jj_consume_token(EQUALS);
|
||||
substitute.setKey(key.image.toString());
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
chainKey = jj_consume_token(ID);
|
||||
jj_consume_token(DOT);
|
||||
substitute.setChainKey(chainKey.image.toString());
|
||||
operation = jj_consume_token(ID);
|
||||
substitute.setOperation(operation.image.toString());
|
||||
jj_consume_token(LPAREN);
|
||||
String operationType = operation.image.toLowerCase();
|
||||
if(operationType.equals("projection")) {
|
||||
List<String> columnList = columnList();
|
||||
for (int i = 0; i < columnList.size(); i++) {
|
||||
condition += appendIndicator(columnList.get(i)) + " ";
|
||||
}
|
||||
|
||||
} else {
|
||||
condition = booleanTerm();
|
||||
}
|
||||
|
||||
substitute.setCondition(condition);
|
||||
jj_consume_token(RPAREN);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
jj_consume_token(AND);
|
||||
extendedKey = jj_consume_token(ID);
|
||||
String eKey = extendedKey.image.toString();
|
||||
String cKey = chainKey.image.toString();
|
||||
substitute.setCondition(condition);
|
||||
if (secondR.equals(eKey)) {
|
||||
substitute.resetVariableRelation(eKey, cKey);
|
||||
} else {
|
||||
substitute.resetVariableRelation(cKey, eKey);
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[27] = jj_gen;
|
||||
;
|
||||
}
|
||||
break;
|
||||
case TABLE:
|
||||
jj_consume_token(TABLE);
|
||||
jj_consume_token(LPAREN);
|
||||
tableName = identifier();
|
||||
substitute.setInput(tableName);
|
||||
jj_consume_token(RPAREN);
|
||||
break;
|
||||
case GROUP:
|
||||
operation = jj_consume_token(GROUP);
|
||||
chainKey = jj_consume_token(ID);
|
||||
jj_consume_token(BY);
|
||||
jj_consume_token(LPAREN);
|
||||
List<String> columnList = columnList();
|
||||
for (int i = 0; i < columnList.size(); i++) {
|
||||
condition += appendIndicator(columnList.get(i));
|
||||
}
|
||||
jj_consume_token(RPAREN);
|
||||
substitute.setChainKey(chainKey.image.toString());
|
||||
substitute.setOperation(operation.image.toString());
|
||||
substitute.setCondition(condition);
|
||||
break;
|
||||
default:
|
||||
jj_la1[28] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
{if (true) return substitute;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public SaveCommand saveCommand() throws ParseException {
|
||||
Token t = null;
|
||||
String tableName;
|
||||
SaveCommand save = new SaveCommand(this.out);
|
||||
jj_consume_token(SAVE);
|
||||
t = jj_consume_token(ID);
|
||||
save.setStatement(t.image.toString());
|
||||
jj_consume_token(INTO);
|
||||
jj_consume_token(TABLE);
|
||||
jj_consume_token(LPAREN);
|
||||
tableName = identifier();
|
||||
save.setOutput(tableName);
|
||||
jj_consume_token(RPAREN);
|
||||
{if (true) return save;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public List<String> getLiteralValues() throws ParseException {
|
||||
List<String> values = new ArrayList<String>();
|
||||
String literal = null;
|
||||
|
@ -922,7 +800,7 @@ public class Parser implements ParserConstants {
|
|||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[29] = jj_gen;
|
||||
jj_la1[27] = jj_gen;
|
||||
break label_7;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
|
@ -945,14 +823,14 @@ public class Parser implements ParserConstants {
|
|||
jj_consume_token(QUOTED_IDENTIFIER);
|
||||
break;
|
||||
default:
|
||||
jj_la1[30] = jj_gen;
|
||||
jj_la1[28] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
values.removeAll(values);
|
||||
break;
|
||||
default:
|
||||
jj_la1[31] = jj_gen;
|
||||
jj_la1[29] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
|
@ -972,7 +850,7 @@ public class Parser implements ParserConstants {
|
|||
s = jj_consume_token(QUOTED_IDENTIFIER);
|
||||
break;
|
||||
default:
|
||||
jj_la1[32] = jj_gen;
|
||||
jj_la1[30] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
|
@ -998,7 +876,7 @@ public class Parser implements ParserConstants {
|
|||
col = jj_consume_token(ASTERISK);
|
||||
break;
|
||||
default:
|
||||
jj_la1[33] = jj_gen;
|
||||
jj_la1[31] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
|
@ -1014,14 +892,14 @@ public class Parser implements ParserConstants {
|
|||
col = jj_consume_token(STRING_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[34] = jj_gen;
|
||||
jj_la1[32] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
{if (true) return col.image.substring(1,col.image.toString().length() - 1);}
|
||||
break;
|
||||
default:
|
||||
jj_la1[35] = jj_gen;
|
||||
jj_la1[33] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
|
@ -1041,7 +919,7 @@ public class Parser implements ParserConstants {
|
|||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[36] = jj_gen;
|
||||
jj_la1[34] = jj_gen;
|
||||
break label_8;
|
||||
}
|
||||
jj_consume_token(COMMA);
|
||||
|
@ -1065,7 +943,7 @@ public class Parser implements ParserConstants {
|
|||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[37] = jj_gen;
|
||||
jj_la1[35] = jj_gen;
|
||||
break label_9;
|
||||
}
|
||||
jj_consume_token(COMMA);
|
||||
|
@ -1092,7 +970,7 @@ public class Parser implements ParserConstants {
|
|||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[38] = jj_gen;
|
||||
jj_la1[36] = jj_gen;
|
||||
break label_10;
|
||||
}
|
||||
jj_consume_token(COMMA);
|
||||
|
@ -1114,7 +992,7 @@ public class Parser implements ParserConstants {
|
|||
t = jj_consume_token(INTEGER_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[39] = jj_gen;
|
||||
jj_la1[37] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
|
@ -1128,25 +1006,9 @@ public class Parser implements ParserConstants {
|
|||
|
||||
final public String identifier() throws ParseException {
|
||||
Token t = null;
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case META_TABLE:
|
||||
case ROOT_TABLE:
|
||||
case ID:
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
t = jj_consume_token(ID);
|
||||
break;
|
||||
case META_TABLE:
|
||||
t = jj_consume_token(META_TABLE);
|
||||
break;
|
||||
case ROOT_TABLE:
|
||||
t = jj_consume_token(ROOT_TABLE);
|
||||
break;
|
||||
default:
|
||||
jj_la1[40] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
{if (true) return t.image.toString();}
|
||||
break;
|
||||
case QUOTED_IDENTIFIER:
|
||||
|
@ -1159,247 +1021,20 @@ public class Parser implements ParserConstants {
|
|||
t = jj_consume_token(STRING_LITERAL);
|
||||
break;
|
||||
default:
|
||||
jj_la1[41] = jj_gen;
|
||||
jj_la1[38] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
{if (true) return t.image.substring(1,t.image.toString().length() - 1);}
|
||||
break;
|
||||
default:
|
||||
jj_la1[42] = jj_gen;
|
||||
jj_la1[39] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public String booleanTerm() throws ParseException {
|
||||
String query = "";
|
||||
String tmp = null;
|
||||
String joinTerm = null;
|
||||
String connector = null;
|
||||
query = booleanTerms();
|
||||
if(query.endsWith(".ROW")) {
|
||||
joinTerm = query;
|
||||
query = "";
|
||||
}
|
||||
label_11:
|
||||
while (true) {
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
case OR:
|
||||
;
|
||||
break;
|
||||
default:
|
||||
jj_la1[43] = jj_gen;
|
||||
break label_11;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case AND:
|
||||
jj_consume_token(AND);
|
||||
connector = " AND ";
|
||||
break;
|
||||
case OR:
|
||||
jj_consume_token(OR);
|
||||
connector = " OR ";
|
||||
break;
|
||||
default:
|
||||
jj_la1[44] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
tmp = booleanTerms();
|
||||
if(tmp.endsWith(".ROW")) {
|
||||
joinTerm = tmp;
|
||||
} else {
|
||||
if(!query.equals(""))
|
||||
query += connector;
|
||||
query += tmp;
|
||||
}
|
||||
}
|
||||
if(joinTerm != null) {
|
||||
{if (true) return joinTerm + " BOOL " + query;}
|
||||
} else {
|
||||
{if (true) return query;}
|
||||
}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public String booleanTerms() throws ParseException {
|
||||
Token tSearchName, tComparand;
|
||||
List<String> inList = new ArrayList<String>();
|
||||
String searchName=null,comparator=null,comparand=null;
|
||||
Token joinColumn = null;
|
||||
Token joinKey = null;
|
||||
String result = null;
|
||||
tSearchName = jj_consume_token(ID);
|
||||
searchName = tSearchName.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case DOT:
|
||||
jj_consume_token(DOT);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ID:
|
||||
joinColumn = jj_consume_token(ID);
|
||||
searchName += "." + joinColumn.image.toString();
|
||||
break;
|
||||
case ROW:
|
||||
jj_consume_token(ROW);
|
||||
secondR = searchName;
|
||||
searchName += ".ROW";
|
||||
break;
|
||||
default:
|
||||
jj_la1[45] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[46] = jj_gen;
|
||||
;
|
||||
}
|
||||
comparator = getComparator();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case INTEGER_LITERAL:
|
||||
tComparand = jj_consume_token(INTEGER_LITERAL);
|
||||
comparand = tComparand.image.toString();
|
||||
break;
|
||||
case STRING_LITERAL:
|
||||
tComparand = jj_consume_token(STRING_LITERAL);
|
||||
comparand = tComparand.image.substring(1,tComparand.image.length() - 1);
|
||||
break;
|
||||
case ID:
|
||||
tComparand = jj_consume_token(ID);
|
||||
comparand = tComparand.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case DOT:
|
||||
jj_consume_token(DOT);
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case ROW:
|
||||
jj_consume_token(ROW);
|
||||
secondR = comparand;
|
||||
comparand += ".ROW";
|
||||
break;
|
||||
case ID:
|
||||
joinColumn = jj_consume_token(ID);
|
||||
comparand += "." + joinColumn.image.toString();
|
||||
break;
|
||||
default:
|
||||
jj_la1[47] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[48] = jj_gen;
|
||||
;
|
||||
}
|
||||
break;
|
||||
case LPAREN:
|
||||
inList = getColumns();
|
||||
if(comparator == null) {
|
||||
comparator = "==";
|
||||
}
|
||||
comparand = "";
|
||||
try{
|
||||
for(int i=0; i<inList.size(); i++) {
|
||||
comparand += URLEncoder.encode(inList.get(i), "UTF-8");
|
||||
if(inList.size() != (i+1)) { comparand += "|"; }
|
||||
}
|
||||
} catch (UnsupportedEncodingException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
break;
|
||||
default:
|
||||
jj_la1[49] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
if(searchName.endsWith(".ROW")) {
|
||||
result = appendIndicator(comparand)
|
||||
+ " " + comparator + " " + searchName;
|
||||
} else {
|
||||
result = appendIndicator(searchName)
|
||||
+ " " + comparator + " " + comparand;
|
||||
}
|
||||
|
||||
{if (true) return result;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public String getComparator() throws ParseException {
|
||||
Token t = null;
|
||||
String comparator = null;
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case LCOMP:
|
||||
t = jj_consume_token(LCOMP);
|
||||
comparator = t.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case EQUALS:
|
||||
jj_consume_token(EQUALS);
|
||||
comparator += "=";
|
||||
break;
|
||||
default:
|
||||
jj_la1[50] = jj_gen;
|
||||
;
|
||||
}
|
||||
break;
|
||||
case RCOMP:
|
||||
t = jj_consume_token(RCOMP);
|
||||
comparator = t.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case EQUALS:
|
||||
jj_consume_token(EQUALS);
|
||||
comparator += "=";
|
||||
break;
|
||||
default:
|
||||
jj_la1[51] = jj_gen;
|
||||
;
|
||||
}
|
||||
break;
|
||||
case EQUALS:
|
||||
t = jj_consume_token(EQUALS);
|
||||
comparator = t.image.toString();
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case LCOMP:
|
||||
jj_consume_token(LCOMP);
|
||||
comparator = ">" + comparator;
|
||||
break;
|
||||
default:
|
||||
jj_la1[52] = jj_gen;
|
||||
;
|
||||
}
|
||||
switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
|
||||
case RCOMP:
|
||||
jj_consume_token(RCOMP);
|
||||
comparator = "<" + comparator;
|
||||
break;
|
||||
default:
|
||||
jj_la1[53] = jj_gen;
|
||||
;
|
||||
}
|
||||
break;
|
||||
case NOTEQUAL:
|
||||
t = jj_consume_token(NOTEQUAL);
|
||||
comparator = t.image.toString();
|
||||
break;
|
||||
case NOT:
|
||||
jj_consume_token(NOT);
|
||||
jj_consume_token(IN);
|
||||
comparator = "!!";
|
||||
break;
|
||||
case IN:
|
||||
jj_consume_token(IN);
|
||||
comparator = "==";
|
||||
break;
|
||||
default:
|
||||
jj_la1[54] = jj_gen;
|
||||
jj_consume_token(-1);
|
||||
throw new ParseException();
|
||||
}
|
||||
{if (true) return comparator;}
|
||||
throw new Error("Missing return statement in function");
|
||||
}
|
||||
|
||||
final public String appendIndicator(String columnName) throws ParseException {
|
||||
String column = columnName;
|
||||
{if (true) return (!column.endsWith(":") && column.indexOf(":") == -1)
|
||||
|
@ -1414,42 +1049,34 @@ public class Parser implements ParserConstants {
|
|||
finally { jj_save(0, xla); }
|
||||
}
|
||||
|
||||
final private boolean jj_3R_12() {
|
||||
Token xsp;
|
||||
xsp = jj_scanpos;
|
||||
if (jj_3R_13()) {
|
||||
jj_scanpos = xsp;
|
||||
if (jj_3R_14()) return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
final private boolean jj_3R_14() {
|
||||
Token xsp;
|
||||
xsp = jj_scanpos;
|
||||
if (jj_scan_token(69)) {
|
||||
jj_scanpos = xsp;
|
||||
if (jj_scan_token(70)) return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
final private boolean jj_3R_13() {
|
||||
Token xsp;
|
||||
xsp = jj_scanpos;
|
||||
if (jj_scan_token(65)) {
|
||||
if (jj_scan_token(63)) {
|
||||
jj_scanpos = xsp;
|
||||
if (jj_scan_token(60)) {
|
||||
jj_scanpos = xsp;
|
||||
if (jj_scan_token(61)) return true;
|
||||
}
|
||||
if (jj_scan_token(64)) return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
final private boolean jj_3_1() {
|
||||
if (jj_scan_token(ADD)) return true;
|
||||
if (jj_3R_12()) return true;
|
||||
if (jj_3R_11()) return true;
|
||||
return false;
|
||||
}
|
||||
|
||||
final private boolean jj_3R_12() {
|
||||
if (jj_scan_token(ID)) return true;
|
||||
return false;
|
||||
}
|
||||
|
||||
final private boolean jj_3R_11() {
|
||||
Token xsp;
|
||||
xsp = jj_scanpos;
|
||||
if (jj_3R_12()) {
|
||||
jj_scanpos = xsp;
|
||||
if (jj_3R_13()) return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
|
@ -1462,7 +1089,7 @@ public class Parser implements ParserConstants {
|
|||
public boolean lookingAhead = false;
|
||||
private boolean jj_semLA;
|
||||
private int jj_gen;
|
||||
final private int[] jj_la1 = new int[55];
|
||||
final private int[] jj_la1 = new int[40];
|
||||
static private int[] jj_la1_0;
|
||||
static private int[] jj_la1_1;
|
||||
static private int[] jj_la1_2;
|
||||
|
@ -1472,13 +1099,13 @@ public class Parser implements ParserConstants {
|
|||
jj_la1_2();
|
||||
}
|
||||
private static void jj_la1_0() {
|
||||
jj_la1_0 = new int[] {0x79ffe0,0x79ffe1,0x79ffe0,0x0,0x0,0x0,0x0,0x0,0x19fbc0,0x19fbc0,0x0,0x600,0x0,0x0,0x0,0x0,0x0,0x0,0x1000,0x0,0x0,0x1000000,0x1800000,0x1800000,0x20000000,0x40000000,0x80000000,0x0,0x40000,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x4000000,0x0,0x4000000,0x0,0x0,0x0,0x0,0x0,0x0,0x0,};
|
||||
jj_la1_0 = new int[] {0x79ffe0,0x79ffe1,0x79ffe0,0x0,0x0,0x0,0x0,0x0,0x19fbc0,0x19fbc0,0x0,0x600,0x0,0x0,0x0,0x0,0x0,0x0,0x1000,0x0,0x0,0x1000000,0x1800000,0x1800000,0x20000000,0x40000000,0x80000000,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,};
|
||||
}
|
||||
private static void jj_la1_1() {
|
||||
jj_la1_1 = new int[] {0x40000000,0x40000000,0x40000000,0x0,0x8,0x8,0x0,0x0,0xc0000000,0xc0000000,0x30000000,0x0,0x398e000,0x70000,0x700000,0x398e000,0x4,0x4,0xc000000,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x1,0x80000000,0x4,0x0,0x4,0x0,0x1000,0x0,0x1000,0x4,0x4,0x4,0x0,0x30000000,0x0,0x30000000,0x3,0x3,0x0,0x8,0x0,0x8,0x10,0x40,0x40,0x80,0x100,0xfc0,};
|
||||
jj_la1_1 = new int[] {0x0,0x0,0x0,0x8000000,0x38000000,0x38000000,0x38000000,0x38000000,0x8000000,0x8000000,0x88000000,0x0,0x1cc7000,0x38000,0x380000,0x1cc7000,0x4,0x4,0x6000000,0x80000000,0x80000000,0x0,0x0,0x0,0x0,0x0,0x0,0x88000004,0x88000000,0x88000004,0x80000000,0x18000800,0x80000000,0x98000800,0x4,0x4,0x4,0x18000000,0x80000000,0x88000000,};
|
||||
}
|
||||
private static void jj_la1_2() {
|
||||
jj_la1_2 = new int[] {0x2,0x82,0x2,0x2,0xe,0xe,0xe,0xe,0x2,0x2,0x62,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x60,0x60,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x2,0x62,0x62,0x62,0x60,0x6,0x60,0x66,0x0,0x0,0x0,0x6,0x2,0x60,0x62,0x0,0x0,0x2,0x0,0x2,0x0,0x46,0x0,0x0,0x0,0x0,0x0,};
|
||||
jj_la1_2 = new int[] {0x0,0x2,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x1,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x1,0x1,0x0,0x0,0x0,0x0,0x0,0x0,0x1,0x1,0x1,0x1,0x0,0x1,0x1,0x0,0x0,0x0,0x0,0x1,0x1,};
|
||||
}
|
||||
final private JJCalls[] jj_2_rtns = new JJCalls[1];
|
||||
private boolean jj_rescan = false;
|
||||
|
@ -1493,7 +1120,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1506,7 +1133,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1516,7 +1143,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1526,7 +1153,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1535,7 +1162,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1544,7 +1171,7 @@ public class Parser implements ParserConstants {
|
|||
token = new Token();
|
||||
jj_ntk = -1;
|
||||
jj_gen = 0;
|
||||
for (int i = 0; i < 55; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < 40; i++) jj_la1[i] = -1;
|
||||
for (int i = 0; i < jj_2_rtns.length; i++) jj_2_rtns[i] = new JJCalls();
|
||||
}
|
||||
|
||||
|
@ -1655,15 +1282,15 @@ public class Parser implements ParserConstants {
|
|||
|
||||
public ParseException generateParseException() {
|
||||
jj_expentries.removeAllElements();
|
||||
boolean[] la1tokens = new boolean[72];
|
||||
for (int i = 0; i < 72; i++) {
|
||||
boolean[] la1tokens = new boolean[66];
|
||||
for (int i = 0; i < 66; i++) {
|
||||
la1tokens[i] = false;
|
||||
}
|
||||
if (jj_kind >= 0) {
|
||||
la1tokens[jj_kind] = true;
|
||||
jj_kind = -1;
|
||||
}
|
||||
for (int i = 0; i < 55; i++) {
|
||||
for (int i = 0; i < 40; i++) {
|
||||
if (jj_la1[i] == jj_gen) {
|
||||
for (int j = 0; j < 32; j++) {
|
||||
if ((jj_la1_0[i] & (1<<j)) != 0) {
|
||||
|
@ -1678,7 +1305,7 @@ public class Parser implements ParserConstants {
|
|||
}
|
||||
}
|
||||
}
|
||||
for (int i = 0; i < 72; i++) {
|
||||
for (int i = 0; i < 66; i++) {
|
||||
if (la1tokens[i]) {
|
||||
jj_expentry = new int[1];
|
||||
jj_expentry[0] = i;
|
||||
|
|
|
@ -34,42 +34,36 @@ public interface ParserConstants {
|
|||
int AND = 32;
|
||||
int OR = 33;
|
||||
int COMMA = 34;
|
||||
int DOT = 35;
|
||||
int LPAREN = 36;
|
||||
int RPAREN = 37;
|
||||
int EQUALS = 38;
|
||||
int LCOMP = 39;
|
||||
int RCOMP = 40;
|
||||
int NOT = 41;
|
||||
int IN = 42;
|
||||
int NOTEQUAL = 43;
|
||||
int ASTERISK = 44;
|
||||
int MAX_VERSIONS = 45;
|
||||
int MAX_LENGTH = 46;
|
||||
int COMPRESSION = 47;
|
||||
int NONE = 48;
|
||||
int BLOCK = 49;
|
||||
int RECORD = 50;
|
||||
int IN_MEMORY = 51;
|
||||
int BLOOMFILTER = 52;
|
||||
int COUNTING_BLOOMFILTER = 53;
|
||||
int RETOUCHED_BLOOMFILTER = 54;
|
||||
int VECTOR_SIZE = 55;
|
||||
int NUM_HASH = 56;
|
||||
int NUM_ENTRIES = 57;
|
||||
int ADD = 58;
|
||||
int CHANGE = 59;
|
||||
int META_TABLE = 60;
|
||||
int ROOT_TABLE = 61;
|
||||
int SAVE = 62;
|
||||
int GROUP = 63;
|
||||
int BY = 64;
|
||||
int ID = 65;
|
||||
int INTEGER_LITERAL = 66;
|
||||
int FLOATING_POINT_LITERAL = 67;
|
||||
int EXPONENT = 68;
|
||||
int QUOTED_IDENTIFIER = 69;
|
||||
int STRING_LITERAL = 70;
|
||||
int LPAREN = 35;
|
||||
int RPAREN = 36;
|
||||
int EQUALS = 37;
|
||||
int LCOMP = 38;
|
||||
int RCOMP = 39;
|
||||
int NOT = 40;
|
||||
int IN = 41;
|
||||
int NOTEQUAL = 42;
|
||||
int ASTERISK = 43;
|
||||
int MAX_VERSIONS = 44;
|
||||
int MAX_LENGTH = 45;
|
||||
int COMPRESSION = 46;
|
||||
int NONE = 47;
|
||||
int BLOCK = 48;
|
||||
int RECORD = 49;
|
||||
int IN_MEMORY = 50;
|
||||
int BLOOMFILTER = 51;
|
||||
int COUNTING_BLOOMFILTER = 52;
|
||||
int RETOUCHED_BLOOMFILTER = 53;
|
||||
int VECTOR_SIZE = 54;
|
||||
int NUM_HASH = 55;
|
||||
int NUM_ENTRIES = 56;
|
||||
int ADD = 57;
|
||||
int CHANGE = 58;
|
||||
int ID = 59;
|
||||
int INTEGER_LITERAL = 60;
|
||||
int FLOATING_POINT_LITERAL = 61;
|
||||
int EXPONENT = 62;
|
||||
int QUOTED_IDENTIFIER = 63;
|
||||
int STRING_LITERAL = 64;
|
||||
|
||||
int DEFAULT = 0;
|
||||
|
||||
|
@ -109,7 +103,6 @@ public interface ParserConstants {
|
|||
"\"and\"",
|
||||
"\"or\"",
|
||||
"\",\"",
|
||||
"\".\"",
|
||||
"\"(\"",
|
||||
"\")\"",
|
||||
"\"=\"",
|
||||
|
@ -134,11 +127,6 @@ public interface ParserConstants {
|
|||
"\"num_entries\"",
|
||||
"\"add\"",
|
||||
"\"change\"",
|
||||
"\".META.\"",
|
||||
"\"-ROOT-\"",
|
||||
"\"save\"",
|
||||
"\"group\"",
|
||||
"\"by\"",
|
||||
"<ID>",
|
||||
"<INTEGER_LITERAL>",
|
||||
"<FLOATING_POINT_LITERAL>",
|
||||
|
|
File diff suppressed because it is too large
Load Diff
|
@ -1,74 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell;
|
||||
|
||||
import java.io.OutputStreamWriter;
|
||||
import java.io.UnsupportedEncodingException;
|
||||
import java.io.Writer;
|
||||
|
||||
import junit.framework.TestCase;
|
||||
import junit.framework.TestSuite;
|
||||
import junit.textui.TestRunner;
|
||||
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.shell.algebra.Constants;
|
||||
import org.apache.hadoop.hbase.shell.algebra.TestBooleanCondition;
|
||||
|
||||
/**
|
||||
* Binding variables, substitution variables test
|
||||
*/
|
||||
public class TestSubstitutionVariables extends TestCase {
|
||||
|
||||
private String TABLE_NAME = "table_name";
|
||||
private String SUBSTITUTION_VARIABLE = "A";
|
||||
static HBaseConfiguration conf = new HBaseConfiguration();
|
||||
|
||||
public void testSubstitution() {
|
||||
SubstituteCommand substitute = new SubstituteCommand(null);
|
||||
|
||||
substitute.setKey(SUBSTITUTION_VARIABLE);
|
||||
substitute.setInput(TABLE_NAME);
|
||||
substitute.execute(conf);
|
||||
|
||||
VariableRef ref = VariablesPool.get(SUBSTITUTION_VARIABLE).get(null);
|
||||
assertTrue(ref.getArgument().equals(TABLE_NAME));
|
||||
}
|
||||
|
||||
public void testCombinedQueries() throws UnsupportedEncodingException {
|
||||
Writer out = new OutputStreamWriter(System.out, "UTF-8");
|
||||
SubstituteCommand substitute = new SubstituteCommand(out);
|
||||
|
||||
substitute.setKey(SUBSTITUTION_VARIABLE);
|
||||
substitute.setInput(TABLE_NAME);
|
||||
substitute.execute(conf);
|
||||
|
||||
substitute = new SubstituteCommand(out);
|
||||
substitute.setKey("B");
|
||||
substitute.setChainKey(SUBSTITUTION_VARIABLE);
|
||||
substitute.setOperation(Constants.RELATIONAL_SELECTION);
|
||||
substitute.setCondition(TestBooleanCondition.EXPRESSION_OR);
|
||||
substitute.execute(conf);
|
||||
}
|
||||
|
||||
public static void main(String[] args) {
|
||||
TestRunner.run(new TestSuite(TestSubstitutionVariables.class));
|
||||
}
|
||||
|
||||
}
|
|
@ -1,84 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.UnsupportedEncodingException;
|
||||
|
||||
import junit.framework.TestCase;
|
||||
import junit.framework.TestSuite;
|
||||
import junit.textui.TestRunner;
|
||||
|
||||
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ExpressionParser;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ParseException;
|
||||
import org.apache.hadoop.io.MapWritable;
|
||||
import org.apache.hadoop.io.Text;
|
||||
|
||||
/**
|
||||
* Test boolean expression
|
||||
*/
|
||||
public class TestBooleanCondition extends TestCase {
|
||||
ExpressionParser expression;
|
||||
public static String EXPRESSION_OR = "key:2 = value2 OR key:3 = value1";
|
||||
public static String EXPRESSION_AND = "key:2 = value2 AND key:3 = value1";
|
||||
public static String EXPRESSION = "key:3 > 1";
|
||||
public static String TERM_FILTER = "key:4 > 100 AND key:4 !! 110|120|130|140|150";
|
||||
|
||||
Text[] keys = { new Text("key:1"), new Text("key:2"), new Text("key:3"), new Text("key:4") };
|
||||
|
||||
ImmutableBytesWritable[] values = {
|
||||
new ImmutableBytesWritable("value1".getBytes()),
|
||||
new ImmutableBytesWritable("value2".getBytes()),
|
||||
new ImmutableBytesWritable("3".getBytes()),
|
||||
new ImmutableBytesWritable("150".getBytes())
|
||||
};
|
||||
|
||||
public void testCheckConstraints() throws UnsupportedEncodingException {
|
||||
MapWritable data = new MapWritable();
|
||||
for (int i = 0; i < keys.length; i++) {
|
||||
data.put(keys[i], values[i]);
|
||||
}
|
||||
|
||||
try {
|
||||
expression = new ExpressionParser(EXPRESSION_OR);
|
||||
expression.booleanExpressionParse();
|
||||
assertTrue(expression.checkConstraints(data));
|
||||
|
||||
expression = new ExpressionParser(EXPRESSION_AND);
|
||||
expression.booleanExpressionParse();
|
||||
assertFalse(expression.checkConstraints(data));
|
||||
|
||||
expression = new ExpressionParser(EXPRESSION);
|
||||
expression.booleanExpressionParse();
|
||||
assertTrue(expression.checkConstraints(data));
|
||||
|
||||
expression = new ExpressionParser(TERM_FILTER);
|
||||
expression.booleanExpressionParse();
|
||||
assertFalse(expression.checkConstraints(data));
|
||||
} catch (ParseException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
}
|
||||
|
||||
public static void main(String[] args) {
|
||||
TestRunner.run(new TestSuite(TestBooleanCondition.class));
|
||||
}
|
||||
|
||||
}
|
|
@ -1,250 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.Map;
|
||||
import java.util.Random;
|
||||
import java.util.TreeMap;
|
||||
|
||||
import org.apache.commons.logging.Log;
|
||||
import org.apache.commons.logging.LogFactory;
|
||||
import org.apache.hadoop.dfs.MiniDFSCluster;
|
||||
import org.apache.hadoop.fs.FileSystem;
|
||||
import org.apache.hadoop.fs.Path;
|
||||
import org.apache.hadoop.hbase.HBaseAdmin;
|
||||
import org.apache.hadoop.hbase.HBaseConfiguration;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.hbase.HConstants;
|
||||
import org.apache.hadoop.hbase.HScannerInterface;
|
||||
import org.apache.hadoop.hbase.HStoreKey;
|
||||
import org.apache.hadoop.hbase.HTable;
|
||||
import org.apache.hadoop.hbase.HTableDescriptor;
|
||||
import org.apache.hadoop.hbase.MasterNotRunningException;
|
||||
import org.apache.hadoop.hbase.MiniHBaseCluster;
|
||||
import org.apache.hadoop.hbase.MultiRegionTable;
|
||||
import org.apache.hadoop.hbase.mapred.IdentityTableReduce;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobClient;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.MiniMRCluster;
|
||||
|
||||
public class TestBooleanTermFilter extends MultiRegionTable {
|
||||
@SuppressWarnings("hiding")
|
||||
private static final Log LOG = LogFactory.getLog(TestBooleanTermFilter.class
|
||||
.getName());
|
||||
|
||||
static final String INPUT_TABLE = "test_table";
|
||||
static final String OUTPUT_TABLE = "result_table";
|
||||
static final Text COUNT_COLUMNFAMILY = new Text("count:");
|
||||
static final Text RANDOMINT_COLUMNFAMILY = new Text("randomInt:");
|
||||
static final String GROUP_COLUMN_FAMILIES = "count: randomInt:";
|
||||
static final String BOOLEAN_TERM = "randomInt: > 100 AND count: <= 100 AND randomInt: !! 110|120|130|140|150";
|
||||
private MiniDFSCluster dfsCluster = null;
|
||||
private FileSystem fs;
|
||||
private Path dir;
|
||||
private MiniHBaseCluster hCluster = null;
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public void setUp() throws Exception {
|
||||
super.setUp();
|
||||
conf.setLong("hbase.hregion.max.filesize", 256 * 1024);
|
||||
dfsCluster = new MiniDFSCluster(conf, 1, true, (String[]) null);
|
||||
try {
|
||||
fs = dfsCluster.getFileSystem();
|
||||
dir = new Path("/hbase");
|
||||
fs.mkdirs(dir);
|
||||
// Start up HBase cluster
|
||||
hCluster = new MiniHBaseCluster(conf, 1, dfsCluster);
|
||||
} catch (Exception e) {
|
||||
if (dfsCluster != null) {
|
||||
dfsCluster.shutdown();
|
||||
dfsCluster = null;
|
||||
}
|
||||
throw e;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public void tearDown() throws Exception {
|
||||
super.tearDown();
|
||||
if (hCluster != null) {
|
||||
hCluster.shutdown();
|
||||
}
|
||||
|
||||
if (dfsCluster != null) {
|
||||
dfsCluster.shutdown();
|
||||
}
|
||||
|
||||
if (fs != null) {
|
||||
try {
|
||||
fs.close();
|
||||
} catch (IOException e) {
|
||||
LOG.info("During tear down got a " + e.getMessage());
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
public void testBooleanFilterMapReduce() {
|
||||
try {
|
||||
HTableDescriptor desc = new HTableDescriptor(INPUT_TABLE);
|
||||
String[] columns = GROUP_COLUMN_FAMILIES.split(" ");
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
desc.addFamily(new HColumnDescriptor(columns[i]));
|
||||
}
|
||||
HBaseAdmin admin = new HBaseAdmin(this.conf);
|
||||
admin.createTable(desc);
|
||||
|
||||
// insert random data into the input table
|
||||
HTable table = new HTable(conf, new Text(INPUT_TABLE));
|
||||
Random oRandom = new Random();
|
||||
|
||||
for (int j = 0; j < 200; j++) {
|
||||
int i = oRandom.nextInt(200) + 1;
|
||||
|
||||
long lockid = table.startUpdate(new Text("rowKey" + j));
|
||||
table.put(lockid, COUNT_COLUMNFAMILY, Integer.toString(j).getBytes(
|
||||
HConstants.UTF8_ENCODING));
|
||||
table.put(lockid, RANDOMINT_COLUMNFAMILY, Integer.toString(i).getBytes(
|
||||
HConstants.UTF8_ENCODING));
|
||||
table.commit(lockid, System.currentTimeMillis());
|
||||
}
|
||||
|
||||
long lockid = table.startUpdate(new Text("rowKey2001"));
|
||||
table.put(lockid, COUNT_COLUMNFAMILY, "12"
|
||||
.getBytes(HConstants.UTF8_ENCODING));
|
||||
table.put(lockid, RANDOMINT_COLUMNFAMILY, "120"
|
||||
.getBytes(HConstants.UTF8_ENCODING));
|
||||
table.commit(lockid, System.currentTimeMillis());
|
||||
|
||||
} catch (MasterNotRunningException e) {
|
||||
e.printStackTrace();
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
try {
|
||||
HTableDescriptor output = new HTableDescriptor(OUTPUT_TABLE);
|
||||
String[] columns = GROUP_COLUMN_FAMILIES.split(" ");
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
output.addFamily(new HColumnDescriptor(columns[i]));
|
||||
}
|
||||
// create output table
|
||||
HBaseAdmin admin = new HBaseAdmin(this.conf);
|
||||
admin.createTable(output);
|
||||
} catch (MasterNotRunningException e) {
|
||||
e.printStackTrace();
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
MiniMRCluster mrCluster = null;
|
||||
try {
|
||||
mrCluster = new MiniMRCluster(2, fs.getUri().toString(), 1);
|
||||
|
||||
JobConf jobConf = new JobConf(conf, TestBooleanTermFilter.class);
|
||||
jobConf.setJobName("process boolean term filter mapreduce");
|
||||
jobConf.setNumMapTasks(2);
|
||||
jobConf.setNumReduceTasks(1);
|
||||
|
||||
IdentityFilterMap.initJob(INPUT_TABLE, GROUP_COLUMN_FAMILIES,
|
||||
BOOLEAN_TERM, IdentityFilterMap.class, jobConf);
|
||||
|
||||
IdentityTableReduce.initJob(OUTPUT_TABLE, IdentityTableReduce.class,
|
||||
jobConf);
|
||||
|
||||
JobClient.runJob(jobConf);
|
||||
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
} finally {
|
||||
mrCluster.shutdown();
|
||||
}
|
||||
|
||||
try {
|
||||
verify(conf, OUTPUT_TABLE);
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Check the filtered value
|
||||
*
|
||||
* @param conf
|
||||
* @param outputTable
|
||||
* @throws IOException
|
||||
*/
|
||||
private void verify(HBaseConfiguration conf, String outputTable)
|
||||
throws IOException {
|
||||
HTable table = new HTable(conf, new Text(outputTable));
|
||||
Text[] columns = { COUNT_COLUMNFAMILY, RANDOMINT_COLUMNFAMILY };
|
||||
HScannerInterface scanner = table.obtainScanner(columns,
|
||||
HConstants.EMPTY_START_ROW);
|
||||
|
||||
try {
|
||||
HStoreKey key = new HStoreKey();
|
||||
TreeMap<Text, byte[]> results = new TreeMap<Text, byte[]>();
|
||||
|
||||
int i = 0;
|
||||
while (scanner.next(key, results)) {
|
||||
for (Map.Entry<Text, byte[]> e : results.entrySet()) {
|
||||
if (e.getKey().equals(COUNT_COLUMNFAMILY)) {
|
||||
if (Integer.parseInt(new String(e.getValue())) > 100) {
|
||||
LOG.info("result_table.count: " + new String(e.getValue()));
|
||||
}
|
||||
assertTrue((Integer.parseInt(new String(e.getValue())) <= 100));
|
||||
} else {
|
||||
if (Integer.parseInt(new String(e.getValue())) <= 100
|
||||
|| !checkNotInList(Integer.parseInt(new String(e.getValue())))) {
|
||||
LOG.info("result_table.randomInt: " + new String(e.getValue()));
|
||||
}
|
||||
assertTrue((Integer.parseInt(new String(e.getValue())) > 100 && checkNotInList(Integer
|
||||
.parseInt(new String(e.getValue())))));
|
||||
}
|
||||
i++;
|
||||
}
|
||||
}
|
||||
|
||||
assertTrue(i > 0);
|
||||
if (i <= 0) {
|
||||
LOG.info("result_table.rowNumber: " + i);
|
||||
}
|
||||
|
||||
} finally {
|
||||
scanner.close();
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
/**
|
||||
* Check 'NOT IN' filter-list
|
||||
*/
|
||||
private boolean checkNotInList(int parseInt) {
|
||||
return (parseInt != 110 && parseInt != 120 && parseInt != 130
|
||||
&& parseInt != 140 && parseInt != 150) ? true : false;
|
||||
}
|
||||
}
|
|
@ -1,45 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import junit.framework.TestCase;
|
||||
import junit.framework.TestSuite;
|
||||
import junit.textui.TestRunner;
|
||||
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ExpressionParser;
|
||||
import org.apache.hadoop.hbase.shell.algebra.generated.ParseException;
|
||||
|
||||
public class TestJoinCondition extends TestCase {
|
||||
ExpressionParser expressionParser;
|
||||
public static String EXPRESSION = "a.ROW = b.size BOOL a.length = b.length AND a.name = edward";
|
||||
|
||||
public void testJoinCondition() {
|
||||
expressionParser = new ExpressionParser(EXPRESSION);
|
||||
try {
|
||||
expressionParser.joinExpressionParse();
|
||||
} catch (ParseException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
}
|
||||
|
||||
public static void main(String[] args) {
|
||||
TestRunner.run(new TestSuite(TestJoinCondition.class));
|
||||
}
|
||||
}
|
|
@ -1,250 +0,0 @@
|
|||
/**
|
||||
* Copyright 2007 The Apache Software Foundation
|
||||
*
|
||||
* Licensed to the Apache Software Foundation (ASF) under one
|
||||
* or more contributor license agreements. See the NOTICE file
|
||||
* distributed with this work for additional information
|
||||
* regarding copyright ownership. The ASF licenses this file
|
||||
* to you under the Apache License, Version 2.0 (the
|
||||
* "License"); you may not use this file except in compliance
|
||||
* with the License. You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
package org.apache.hadoop.hbase.shell.algebra;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.TreeMap;
|
||||
|
||||
import org.apache.commons.logging.Log;
|
||||
import org.apache.commons.logging.LogFactory;
|
||||
import org.apache.hadoop.dfs.MiniDFSCluster;
|
||||
import org.apache.hadoop.fs.FileSystem;
|
||||
import org.apache.hadoop.fs.Path;
|
||||
import org.apache.hadoop.hbase.HBaseAdmin;
|
||||
import org.apache.hadoop.hbase.HColumnDescriptor;
|
||||
import org.apache.hadoop.hbase.HConstants;
|
||||
import org.apache.hadoop.hbase.HScannerInterface;
|
||||
import org.apache.hadoop.hbase.HStoreKey;
|
||||
import org.apache.hadoop.hbase.HTable;
|
||||
import org.apache.hadoop.hbase.HTableDescriptor;
|
||||
import org.apache.hadoop.hbase.StaticTestEnvironment;
|
||||
import org.apache.hadoop.hbase.MasterNotRunningException;
|
||||
import org.apache.hadoop.hbase.MiniHBaseCluster;
|
||||
import org.apache.hadoop.hbase.MultiRegionTable;
|
||||
import org.apache.hadoop.hbase.mapred.TableReduce;
|
||||
import org.apache.hadoop.io.Text;
|
||||
import org.apache.hadoop.mapred.JobClient;
|
||||
import org.apache.hadoop.mapred.JobConf;
|
||||
import org.apache.hadoop.mapred.MiniMRCluster;
|
||||
|
||||
/**
|
||||
* HBase shell join test
|
||||
*/
|
||||
public class TestTableJoinMapReduce extends MultiRegionTable {
|
||||
@SuppressWarnings("hiding")
|
||||
private static final Log LOG = LogFactory.getLog(TestTableJoinMapReduce.class
|
||||
.getName());
|
||||
static final String FIRST_RELATION = "r1";
|
||||
static final String SECOND_RELATION = "r2";
|
||||
static final String JOIN_EXPRESSION = "r1.c: = r2.ROW BOOL ";
|
||||
static final String FIRST_COLUMNS = "a: b: c:";
|
||||
static final String SECOND_COLUMNS = "d: e:";
|
||||
static final String OUTPUT_TABLE = "result_table";
|
||||
private MiniDFSCluster dfsCluster = null;
|
||||
private FileSystem fs;
|
||||
private Path dir;
|
||||
private MiniHBaseCluster hCluster = null;
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public void setUp() throws Exception {
|
||||
super.setUp();
|
||||
dfsCluster = new MiniDFSCluster(conf, 1, true, (String[]) null);
|
||||
try {
|
||||
fs = dfsCluster.getFileSystem();
|
||||
dir = new Path("/hbase");
|
||||
fs.mkdirs(dir);
|
||||
// Start up HBase cluster
|
||||
hCluster = new MiniHBaseCluster(conf, 1, dfsCluster);
|
||||
} catch (Exception e) {
|
||||
StaticTestEnvironment.shutdownDfs(dfsCluster);
|
||||
throw e;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public void tearDown() throws Exception {
|
||||
super.tearDown();
|
||||
if (hCluster != null) {
|
||||
hCluster.shutdown();
|
||||
}
|
||||
StaticTestEnvironment.shutdownDfs(dfsCluster);
|
||||
}
|
||||
|
||||
/**
|
||||
* @throws Exception
|
||||
*/
|
||||
public void testTableJoinMapReduce() throws Exception {
|
||||
HTable table = null;
|
||||
try {
|
||||
HTableDescriptor desc = new HTableDescriptor(FIRST_RELATION);
|
||||
String[] columns = FIRST_COLUMNS.split(" ");
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
desc.addFamily(new HColumnDescriptor(columns[i]));
|
||||
}
|
||||
HBaseAdmin admin = new HBaseAdmin(this.conf);
|
||||
admin.createTable(desc);
|
||||
|
||||
// insert random data into the input table
|
||||
table = new HTable(conf, new Text(FIRST_RELATION));
|
||||
for (int j = 0; j < 5; j++) {
|
||||
long lockid = table.startUpdate(new Text("rowKey" + j));
|
||||
table.put(lockid, new Text("a:"), Integer.toString(j).getBytes(
|
||||
HConstants.UTF8_ENCODING));
|
||||
table.put(lockid, new Text("b:"), Integer.toString(j).getBytes(
|
||||
HConstants.UTF8_ENCODING));
|
||||
table.put(lockid, new Text("c:"), ("joinKey-" + Integer.toString(j))
|
||||
.getBytes(HConstants.UTF8_ENCODING));
|
||||
table.commit(lockid, System.currentTimeMillis());
|
||||
}
|
||||
|
||||
} catch (MasterNotRunningException e) {
|
||||
e.printStackTrace();
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
} finally {
|
||||
if (table != null) {
|
||||
table.close();
|
||||
}
|
||||
}
|
||||
|
||||
try {
|
||||
HTableDescriptor desc = new HTableDescriptor(SECOND_RELATION);
|
||||
String[] columns = SECOND_COLUMNS.split(" ");
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
desc.addFamily(new HColumnDescriptor(columns[i]));
|
||||
}
|
||||
HBaseAdmin admin = new HBaseAdmin(this.conf);
|
||||
admin.createTable(desc);
|
||||
|
||||
// insert random data into the input table
|
||||
table = new HTable(conf, new Text(SECOND_RELATION));
|
||||
for (int j = 0; j < 3; j++) {
|
||||
long lockid = table.startUpdate(new Text("joinKey-" + j));
|
||||
table.put(lockid, new Text("d:"), ("s-" + Integer.toString(j))
|
||||
.getBytes(HConstants.UTF8_ENCODING));
|
||||
table.put(lockid, new Text("e:"), ("s-" + Integer.toString(j))
|
||||
.getBytes(HConstants.UTF8_ENCODING));
|
||||
table.commit(lockid, System.currentTimeMillis());
|
||||
}
|
||||
|
||||
} catch (MasterNotRunningException e) {
|
||||
e.printStackTrace();
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
} finally {
|
||||
if (table != null) {
|
||||
table.close();
|
||||
}
|
||||
}
|
||||
|
||||
try {
|
||||
HTableDescriptor output = new HTableDescriptor(OUTPUT_TABLE);
|
||||
String[] columns = (FIRST_COLUMNS + " " + SECOND_COLUMNS).split(" ");
|
||||
for (int i = 0; i < columns.length; i++) {
|
||||
output.addFamily(new HColumnDescriptor(columns[i]));
|
||||
}
|
||||
// create output table
|
||||
HBaseAdmin admin = new HBaseAdmin(this.conf);
|
||||
admin.createTable(output);
|
||||
} catch (MasterNotRunningException e) {
|
||||
e.printStackTrace();
|
||||
} catch (IOException e) {
|
||||
e.printStackTrace();
|
||||
}
|
||||
|
||||
MiniMRCluster mrCluster = null;
|
||||
try {
|
||||
mrCluster = new MiniMRCluster(2, fs.getUri().toString(), 1);
|
||||
|
||||
JobConf jobConf = new JobConf(conf, TestTableJoinMapReduce.class);
|
||||
jobConf.setJobName("process table join mapreduce");
|
||||
jobConf.setNumMapTasks(2);
|
||||
jobConf.setNumReduceTasks(1);
|
||||
|
||||
IndexJoinMap.initJob(FIRST_RELATION, SECOND_RELATION, FIRST_COLUMNS,
|
||||
SECOND_COLUMNS, JOIN_EXPRESSION, IndexJoinMap.class, jobConf);
|
||||
TableReduce.initJob(OUTPUT_TABLE, IndexJoinReduce.class, jobConf);
|
||||
|
||||
JobClient.runJob(jobConf);
|
||||
|
||||
} catch (Exception e) {
|
||||
e.printStackTrace();
|
||||
throw e;
|
||||
} finally {
|
||||
if (mrCluster != null) {
|
||||
mrCluster.shutdown();
|
||||
}
|
||||
}
|
||||
|
||||
try {
|
||||
verify(OUTPUT_TABLE);
|
||||
} catch (Exception e) {
|
||||
e.printStackTrace();
|
||||
throw e;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Check the result table lattice size.
|
||||
*
|
||||
* @param conf
|
||||
* @param outputTable
|
||||
* @throws IOException
|
||||
*/
|
||||
private void verify(String outputTable) throws IOException {
|
||||
// Sleep before we start the verify to ensure that when the scanner takes
|
||||
// its snapshot, all the updates have made it into the cache.
|
||||
try {
|
||||
Thread.sleep(conf.getLong("hbase.regionserver.optionalcacheflushinterval",
|
||||
60L * 1000L));
|
||||
} catch (InterruptedException e) {
|
||||
// ignore
|
||||
}
|
||||
|
||||
HTable table = new HTable(conf, new Text(outputTable));
|
||||
Text[] columns = { new Text("a:"), new Text("b:"), new Text("c:"),
|
||||
new Text("d:"), new Text("e:") };
|
||||
HScannerInterface scanner = table.obtainScanner(columns,
|
||||
HConstants.EMPTY_START_ROW);
|
||||
|
||||
try {
|
||||
HStoreKey key = new HStoreKey();
|
||||
TreeMap<Text, byte[]> results = new TreeMap<Text, byte[]>();
|
||||
|
||||
int i = 0;
|
||||
while (scanner.next(key, results)) {
|
||||
LOG.info("result_table.column.size: " + results.keySet().size());
|
||||
assertEquals(5, results.keySet().size());
|
||||
i++;
|
||||
}
|
||||
assertTrue(i == 3);
|
||||
LOG.info("result_table.row.count: " + i);
|
||||
} finally {
|
||||
scanner.close();
|
||||
}
|
||||
|
||||
}
|
||||
}
|
Loading…
Reference in New Issue