Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Speed up TestIndexOrDocValuesQuery. (#12672) 

This changes the following:
 - fewer docs indexed in non-nightly runs,
 - `QueryUtils#checkFirstSkipTo` uses the `ScorerSupplier` API to convey it
   will only check one doc,
 - `QueryUtils#checkFirstSkipTo` no longer relies on timing to run in a
   reasonably amount of time.
This commit is contained in:
Adrien Grand 2023-10-17 14:13:21 +02:00 committed by GitHub
parent 8961dda2a4
commit f055ac6e16
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 11 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
lucene/core/src/test/org/apache/lucene/search/TestIndexOrDocValuesQuery.java
View File

@ -112,14 +112,15 @@ public class TestIndexOrDocValuesQuery extends LuceneTestCase {
newIndexWriterConfig() newIndexWriterConfig()
// relies on costs and PointValues.estimateCost so we need the default codec // relies on costs and PointValues.estimateCost so we need the default codec
.setCodec(TestUtil.getDefaultCodec())); .setCodec(TestUtil.getDefaultCodec()));
for (int i = 0; i < 2000; ++i) { final int numDocs = atLeast(1000);
for (int i = 0; i < numDocs; ++i) {
Document doc = new Document(); Document doc = new Document();
if (i < 1000) { if (i < numDocs / 2) {
doc.add(new StringField("f1", "bar", Store.NO)); doc.add(new StringField("f1", "bar", Store.NO));
for (int j = 0; j < 500; j++) { for (int j = 0; j < 500; j++) {
doc.add(new LongField("f2", 42L, Store.NO)); doc.add(new LongField("f2", 42L, Store.NO));
} }
} else if (i == 1001) { } else if (i == numDocs / 2) {
doc.add(new StringField("f1", "foo", Store.NO)); doc.add(new StringField("f1", "foo", Store.NO));
doc.add(new LongField("f2", 2L, Store.NO)); doc.add(new LongField("f2", 2L, Store.NO));
} else { } else {

17
lucene/test-framework/src/java/org/apache/lucene/tests/search/QueryUtils.java
View File

@ -540,8 +540,10 @@ public class QueryUtils {
s.search( s.search(
q, q,
new SimpleCollector() { new SimpleCollector() {
private final Weight w = s.createWeight(rewritten, ScoreMode.COMPLETE, 1);
private Scorable scorer; private Scorable scorer;
private int leafPtr; private int leafPtr;
private long intervalTimes32 = 1 * 32;
@Override @Override
public void setScorer(Scorable scorer) { public void setScorer(Scorable scorer) {
@ -552,10 +554,11 @@ public class QueryUtils {
public void collect(int doc) throws IOException { public void collect(int doc) throws IOException {
float score = scorer.score(); float score = scorer.score();
try { try {
long startNS = System.nanoTime(); // The intervalTimes32 trick helps contain the runtime of this check: first we check
for (int i = lastDoc[0] + 1; i <= doc; i++) { // every single doc in the interval, then after 32 docs we check every 2 docs, etc.
Weight w = s.createWeight(rewritten, ScoreMode.COMPLETE, 1); for (int i = lastDoc[0] + 1; i <= doc; i += intervalTimes32++ / 1024) {
Scorer scorer = w.scorer(context.get(leafPtr)); ScorerSupplier supplier = w.scorerSupplier(context.get(leafPtr));
Scorer scorer = supplier.get(1L); // only checking one doc, so leadCost = 1
assertTrue( assertTrue(
"query collected " + doc + " but advance(" + i + ") says no more docs!", "query collected " + doc + " but advance(" + i + ") says no more docs!",
scorer.iterator().advance(i) != DocIdSetIterator.NO_MORE_DOCS); scorer.iterator().advance(i) != DocIdSetIterator.NO_MORE_DOCS);
@ -579,12 +582,6 @@ public class QueryUtils {
score, score,
advanceScore, advanceScore,
maxDiff); maxDiff);
// Hurry things along if they are going slow (eg
// if you got SimpleText codec this will kick in):
if (i < doc && System.nanoTime() - startNS > 5_000_000) {
i = doc - 1;
}
} }
lastDoc[0] = doc; lastDoc[0] = doc;
} catch (IOException e) { } catch (IOException e) {