* ExpressionSelectors: Add caching selectors.
- SingleLongInputCaching selector for expressions on the __time column,
using a similar optimization to SingleScanTimeDimSelector
- SingleStringInputDimensionSelector for expressions on string columns
that return strings, using a similar optimization to ExtractionFn
based DimensionSelectors.
- SingleStringInputCaching selector for expressions on string columns
that return primitives.
Also, in the SQL planner, prefer expressions for time operations
rather than extractionFns.
* Code review comments.
* Add flattenSpec support to the Avro parser.
Also:
- Refactor the JSONPathParser a bit so it can share flattening code
with Avro (see ObjectFlatteners).
- Remove the JSONParser. It was only used in two places: by
UriNamespaceExtractor, and as a base for JSONToLowerParser. Migrated
the former to JSONPathParser and made the latter a standalone.
- Move GenericRecordAsMap to the Parquet extension, since the Avro
extension no longer uses it.
* Fix indentation.
* Fix equals/hashCode.
* add jq expression in the flattenSpec
* more tests
* add benchmark
* fix style
* use JsonNode for both JSONPath and JQ
* clean up
* more clean up
* add documentation
* fix style
* move jackson-jq version to dependencyManagement section. remove commented code
* oops. revert wrong fix
* throw IllegalArgumentException for JQ syntax error
* remove e.printStackTrace() that is forbidden
* touch
* SQL: EXPLAIN improvements.
- Include query JSON in explain output.
- Fix a bug where semi-joins and nested groupBys were not fully explained.
- Fix a bug where limits were not included in "select" query explanations.
* Fix compile error.
* Fix compile error.
* Fix tests.
* Factor QueryableIndexColumnSelectorFactory and IncrementalIndexColumnSelectorFactory out of QueryableIndexStorageAdapter and IncrementalIndexStorageAdapter; Add Offset.getBaseReadableOffset(); Remove OffsetHolder interface; Replace Cursor extends ColumnSelectorFactory with composition; Reduce indirection in ColumnValueSelectors created by QueryableIndexColumnSelectorFactory
* Don't override clone() in FilteredOffset (the prev. implementation was broken); Some warnings fixed
* Simplify Cursors in QueryableIndexStorageAdapter
* Address comments
* Remove unused and unimplemented methods from GenericColumn interface
* Comments
* Add metrics to the native queries underpinning SQL.
This is done by factoring out the metrics and request log emitting
code from QueryResource into a new QueryLifecycle class. That class
is used by both QueryResource and the SQL DruidSchema and QueryMaker.
Also fixes a couple of bugs in QueryResource:
- RequestLogLine start time was set to `TimeUnit.NANOSECONDS.toMillis(startNs)`,
which is incorrect since absolute nanos cannot be converted to millis.
- DruidMetrics.makeRequestMetrics was called with null `query` on
unparseable queries, which led to spurious "Unable to log query"
errors.
Partial fix for #4047.
* Code style
* Remove unused imports.
* Fix tests.
* Remove unused import.
* Fixes and improvements to SQL metadata caching.
Also adds support for MultipleSpecificSegmentSpec to CachingClusteredClient.
SQL changes:
- Cache metadata on a per-segment level, in addition to per-dataSource, so
we don't need to re-query all segments whenever a single new one appears.
This should lower the load placed on the cluster by metadata queries.
- Fix race condition in DruidSchema that can cause us to miss metadata. It was
possible to notice new segments, then issue a query, and have that query
not actually hit those segments, and not notice that it didn't hit those segments.
Then, the metadata from those segments would be ignored.
- Fix assumption in DruidSchema that all segments are immutable. Now, mutable
segments are periodically re-queried.
- Fix inappropriate re-use of SchemaPlus. Now we create one for each planning
cycle, rather than sharing one. It caches table objects, which we want to
avoid, since it can cause stale metadata. We do the caching in DruidSchema
so we don't need the SchemaPlus caching.
Server changes:
- Add a TimelineCallback to TimelineServerView, for callers that want to get updates
when the timeline has been modified.
- Change CachingClusteredClient from a QueryRunner to a QuerySegmentWalker. This
allows it to accept queries that are segment-descriptor-based rather than
intervals-based. In particular it will now support MultipleSpecificSegmentSpec.
* Fix DruidSchema, and unused imports.
* Remove unused import.
* Fix SqlBenchmark.
* Remove some unnecessary use of boxed types.
* Fix some incorrect format strings.
* Enable IDEA's MalformedFormatString inspection.
* Add a Checkstyle check for finding uses of incorrect logging packages.
* Fix some incorrect usages of the metamx logger.
* Bypass incorrect logger Checkstyle check where using the correct logger is not simple.
* Fix some more places where the wrong number of arguments are provided to format strings.
* Suppress `MalformedFormatString` inspection on legacy logging test.
* Use @SuppressWarnings rather than a noinspection suppression comment.
* Fix some more incorrect format strings.
* Suppress some more incorrect format string warnings where the incorrect string is intentional.
* Log the aggregator when closing it fails.
* Remove some unneeded log lines.
* Avoid usages of Default system Locale and printing to System.out or System.err in production code
* Fix Charset in DruidKerberosUtil
* Remove redundant string format in GenericIndexed
* Rename StringUtils.safeFormat() to unimportantSafeFormat(); add StringUtils.format() which fails as well as String.format()
* Fix testSafeFormat()
* More fixes of redundant StringUtils.format() inside ISE
* Rename unimportantSafeFormat() to nonStrictFormat()
* Remove DruidProcessingModule, QueryableModule and QueryRunnerFactoryModule from DI for coordinator, overlord, middle-manager. Add RouterDruidProcessing not to allocate processing resources on router
* Fix examples
* Fixes
* Revert Peon configs and add comments
* Remove qualifier
* Remove ability to create segments in v8 format
* Fix IndexGeneratorJobTest
* Fix parameterized test name in IndexMergerTest
* Remove extra legacy merging stuff
* Remove legacy serializer builders
* Remove ConciseBitmapIndexMergerTest and RoaringBitmapIndexMergerTest
* Adding a flag to indicate when ObjectCachingColumnSelectorFactory need not be threadsafe.
* - Use of computeIfAbsent over putIfAbsent
- Replace Maps.newXXXMap() with normal instantiation
- Documentations on when is thread-safe required.
- Use Builders for On/OffheapIncrementalIndex
* - Optimization on computeIfAbsent
- Constant EMPTY DimensionsSpec
- Improvement on IncrementalIndexSchema.Builder
- Remove setting of default values
- Use var args for metrics
- Correction on On/OffheapIncrementalIndex Builders
- Combine On/OffheapIncrementalIndex Builders
* - Removing unused imports.
* - Helper method for testing with IncrementalIndex.Builder
* - Correction on javadoc.
* Style fix
* Expressions: Add ExprMacros, which have the same syntax as functions, but
can convert themselves to any kind of Expr at parse-time.
ExprMacroTable is an extension point for adding new ExprMacros. Anything
that might need to parse expressions needs an ExprMacroTable, which can
be injected through Guice.
* Address code review comments.
* Make using implicit system charset an error
* Use StringUtils.toUtf8() and fromUtf8() instead of String.getBytes() and new String()
* Use English locale in StringUtils.safeFormat()
* Restore comment
* Timeout and maxScatterGatherBytes handling for queries run by Druid SQL
* Address PR comments
* Fix contexts in CalciteQueryTest
* Fix contexts in QuantileSqlAggregatorTest
* Make Errorprone the default compiler
* Address comments
* Make Error Prone's ClassCanBeStatic rule a error
* Preconditions allow only %s pattern
* Fix DruidCoordinatorBalancerTester
* Try to give the compiler more memory
* Remove distribution module activation on jdk 1.8 because only jdk 1.8 is used now
* Don't show compiler warnings
* Try different travis script
* Fix travis.yml
* Make Error Prone optional again
* For error-prone compiler
* Increase compiler's maxmem
* Don't run Error Prone for benchmarks because of OOM
* Skip install step in Travis
* Remove MetricHolder.writeToChannel()
* In travis.yml, check compilation before tests, because it may fail faster
* Add QueryPlus. Add QueryRunner.run(QueryPlus, Map) method with default implementation, to replace QueryRunner.run(Query, Map).
* Fix GroupByMergingQueryRunnerV2
* Fix QueryResourceTest
* Expand the comment to Query.run(walker, context)
* Remove legacy version of BySegmentSkippingQueryRunner.doRun()
* Add LegacyApiQueryRunnerTest and be more specific about legacy API removal plans in Druid 0.11 in Javadocs
* Minor bug fixes in GenericIndexed; Refactor and optimize GenericIndexed; Remove some unnecessary ByteBuffer duplications in some deserialization paths; Add ZeroCopyByteArrayOutputStream
* Fixes
* Move GenericIndexedWriter.writeLongValueToOutputStream() and writeIntValueToOutputStream() to SerializerUtils
* Move constructors
* Add GenericIndexedBenchmark
* Comments
* Typo
* Note in Javadoc that IntermediateLongSupplierSerializer, LongColumnSerializer and LongMetricColumnSerializer are thread-unsafe
* Use primitive collections in IntermediateLongSupplierSerializer instead of BiMap
* Optimize TableLongEncodingWriter
* Add checks to SerializerUtils methods
* Don't restrict byte order in SerializerUtils.writeLongToOutputStream() and writeIntToOutputStream()
* Update GenericIndexedBenchmark
* SerializerUtils.writeIntToOutputStream() and writeLongToOutputStream() separate for big-endian and native-endian
* Add GenericIndexedBenchmark.indexOf()
* More checks in methods in SerializerUtils
* Use helperBuffer.arrayOffset()
* Optimizations in SerializerUtils
* No more singleton. Reduce iterations
* Granularities
* Fix the delay in the test
* Add license header
* Remove unused imports
* Lot more unused imports from all the rearranging
* CR feedback
* Move javadoc to constructor
* Refactor Segment Granularity
* Beginning of one granularity
* Copy the fix for custom periods in segment-grunalrity over here.
* Remove the custom serialization for now.
* Compilation cleanup
* Reformat code
* Fixing unit tests
* Unify to use a single iterable
* Backward compatibility for rolling upgrade
* Minor check style. Cosmetic changes.
* Rename length and millis to duration
* CR feedback
* Minor changes.
* Removing Integer.MAX column size limit.
* On demand creation of headerLong, use v2 instead of v3
* Avoid reusing the same object from a previous test.
* Avoid reusing the same object from a previous test part#2
* code formatting.
* GenericIndexed/Writer code review changes.
* GenericIndexed/writer code review requested changes.
* checkIndex() to static
* native endianess for genericIndexedV2, code review requested changes.
* Formatting
* Hll fix.
* use native endianess during bag size calculation.
* Code review requested changes.
* IOPeon close() changes.
* use different tmp directory path for testing.
* Code review requested changes.
* SQL: Add context and contextual functions to planner.
Added support for context parameters specified as JDBC connection properties
or a JSON object for SQL-over-JSON-over-HTTP.
Also added features that depend on context functionality:
- Added CURRENT_DATE, CURRENT_TIME, CURRENT_TIMESTAMP functions.
- Added support for time zones other than UTC via a "timeZone" context.
- Pass down query context to Druid queries too.
Also some bug fixes:
- Fix DATE handling, it was largely done incorrectly before.
- Fix CAST(__time TO DATE) which should do a floor-to-day.
- Fix non-equality comparisons to FLOOR(__time TO X).
- Fix maxQueryCount property.
* Pass down context to nested queries too.
* Less use of File.deleteOnExit()
* removed deleteOnExit from most of the tests/benchmarks/iopeon
* Made IOpeon closable
* Formatting.
* Revert DeterminePartitionsJobTest, remove cleanup method from IOPeon
* Add filter selectivity estimation for auto search strategy
* Addressed comments
* Lazy bitmap materialization for bitmap sampling and java docs
* Addressed comments.
- Fix wrong non-overlap ratio computation and added unit tests.
- Change Iterable<Integer> to IntIterable
- Remove unnecessary Iterable<Integer>
* Addressed comments
- Split a long ternary operation into if-else blocks
- Add IntListUtils.fromTo()
* Fix test failure and add a test for RangeIntList
* fix code style
* Diabled selectivity estimation for multi-valued dimensions
* Address comment
* SQL: Add resolution parameter to quantile agg, rename to APPROX_QUANTILE.
* Fix bug with re-use of filtered approximate histogram aggregators.
Also add APPROX_QUANTILE tests for filtering and running on complex columns.
Includes some slight refactoring to allow tests to make DruidTables that
include complex columns.
* Remove unused import
* SQL: Ditch CalciteConnection layer and add DruidMeta, extension aggregators.
Switched from CalciteConnection to Planner, bringing benefits:
- CalciteConnection's JDBC interface no longer sits between the SQL server
(HTTP/Avatica) and Druid's query layer. Instead, the SQL servers can use
Druid Sequence objects directly, reducing overhead in the query return path.
- Implemented our own Planner-based Avatica Meta, letting us control
connection timeouts and connection / statement limits. The previous
CalciteConnection-based implementation didn't have any limits or timeouts.
- The Planner interface lets us override the operator table, opening up
SQL language extensions. This patch includes two: APPROX_COUNT_DISTINCT
in core, and a QUANTILE aggregator in the druid-histogram extension.
Also:
- Added INFORMATION_SCHEMA metadata schema.
- Added tests for Unicode literals and escapes.
* Verify statement is actually open before closing it.
* More detailed INFORMATION_SCHEMA docs.
* SQL support for nested groupBys.
Allows, for example, doing exact count distinct by writing:
SELECT COUNT(*) FROM (SELECT DISTINCT col FROM druid.foo)
Contrast with approximate count distinct, which is:
SELECT COUNT(DISTINCT col) FROM druid.foo
* Add deeply-nested groupBy docs, tests, and maxQueryCount config.
* Extract magic constants into statics.
* Rework rules to put preconditions in the "matches" method.
* Add an option to SearchQuery to choose a search query execution strategy.
Supported strategies are
1) Index-only query execution
2) Cursor-based scan
3) Auto: choose an efficient strategy for a given query
* Add SearchStrategy and SearchQueryExecutor
* Address comments
* Rename strategies and set UseIndexesStrategy as the default strategy
* Add a cost-based planner for auto strategy
* Add document
* Fix code style
* apply code style
* apply comments
* Enable parallel test
* Remove unnecessary NotThreadSafe annocation
* Randomize the start port when finding available ports
* Fix test failure
* Change to handle all negatives
* GroupByBenchmark: Add serde, spilling, all-gran benchmarks.
Also use more iterations.
* groupBy v2: Ignore timestamp completely when granularity = all, except for the final merge.
Specifically:
- Remove timestamp from RowBasedKey when not needed
- Set timestamp to null in MapBasedRows that are not part of the final merge.
* Add "like" filter.
* Addressed some PR comments.
* Slight simplifications to LikeFilter.
* Additional simplifications.
* Fix comment in LikeFilter.
* Clarify comment in LikeFilter.
* Simplify LikeMatcher a bit.
* No use going through the optimized path if prefix is empty.
* Add more tests.
With the old code, all on-disk segments were the same. Now they're different.
This will end up altering benchmark results for queryMultiQueryableIndex,
likely making them slower (since values won't group as well as they used to).
The memory changes will help test with larger/more segments, since we won't
have to hold them all in memory at once.
* ability to not rollup at index time, make pre aggregation an option
* rename getRowIndexForRollup to getPriorIndex
* fix doc misspelling
* test query using no-rollup indexes
* fix benchmark fail due to jmh bug
* Add numeric StringComparator
* Only use direct long comparison for numeric ordering in BoundFilter, add time filtering benchmark query
* Address PR comments, add multithreaded BoundDimFilter test
* Add comment on strlen tie handling
* Add timeseries interval filter benchmark
* Adjust docs
* Use jackson for StringComparator, address PR comments
* Add new TopNMetricSpec and SearchSortSpec with tests (WIP)
* More TopNMetricSpec and SearchSortSpec tests
* Fix NewSearchSortSpec serde
* Update docs for new DimensionTopNMetricSpec
* Delete NumericDimensionTopNMetricSpec
* Delete old SearchSortSpec
* Rename NewSearchSortSpec to SearchSortSpec
* Add TopN numeric comparator benchmark, address PR comments
* Refactor OrderByColumnSpec
* Add null checks to NumericComparator and String->BigDecimal conversion function
* Add more OrderByColumnSpec serde tests
* Support filtering on __time column
* Rename DruidPredicate
* Add docs for ValueMatcherFactory, add comment on getColumnCapabilities
* Combine ValueMatcherFactory predicate methods to accept DruidCompositePredicate
* Address PR comments (support filter on all long columns)
* Use predicate factory instead of composite predicate
* Address PR comments
* Lazily initialize long handling in selector/in filter
* Move long value parsing from InFilter to InDimFilter, make long value parsing thread-safe
* Add multithreaded selector/in filter test
* Fix non-final lock object in SelectorDimFilter
Because timestamps at the end instant are not actually part of the interval. This
affected benchmark numbers, since it meant some data points would not be queried
(the interval for the query was based on getDataInterval) and also the
TimestampCheckingOffsets could not use the allWithinThreshold optimization.
This patch introduces a GroupByStrategy concept and two strategies: "v1"
is the current groupBy strategy and "v2" is a new one. It also introduces
a merge buffers concept in DruidProcessingModule, to try to better
manage memory used for merging.
Both of these are described in more detail in #2987.
There are two goals of this patch:
1. Make it possible for historical/realtime nodes to return larger groupBy
result sets, faster, with better memory management.
2. Make it possible for brokers to merge streams when there are no order-by
columns, avoiding materialization.
This patch does not do anything to help with memory management on the broker
when there are order-by columns or when there are nested queries. That could
potentially be done in a future patch.
* new interval based cost function
Addresses issues with balancing of segments in the existing cost function
- `gapPenalty` led to clusters of segments ~30 days apart
- `recencyPenalty` caused imbalance among recent segments
- size-based cost could be skewed by compression
New cost function is purely based on segment intervals:
- assumes each time-slice of a partition is a constant cost
- cost is additive, i.e. cost(A, B union C) = cost(A, B) + cost(A, C)
- cost decays exponentially based on distance between time-slices
* comments and formatting
* add more comments to explain the calculation
Roman Leventov
Jonathan Wei
Parag Jain
Gian Merlino
Jihoon Son
Akash Dwivedi
Kenji Noguchi
dgolitsyn
Slim
Chris Gavin
Goh Wei Xiang
Benedict Jin
Roman Leventov
praveev
DaimonPl
Himanshu
Navis Ryu
Dave Li
kaijianding
Hyukjin Kwon
Charles Allen
Xavier Léauté