Examples with SimpleDocument com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument used on opensource projects

Example 11 with SimpleDocument

use of com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument in project fdb-record-layer by FoundationDB.

the class TextIndexTest method saveSimpleDocumentsWithNoPositions.

@Test
public void saveSimpleDocumentsWithNoPositions() throws Exception {
    final SimpleDocument shakespeareDocument = SimpleDocument.newBuilder().setDocId(1623L).setText(TextSamples.ROMEO_AND_JULIET_PROLOGUE).build();
    final SimpleDocument germanDocument = SimpleDocument.newBuilder().setDocId(1066L).setText(TextSamples.GERMAN).build();
    // Save with positions
    final int shakespeareValueBytes;
    final int germanValueBytes;
    try (FDBRecordContext context = openContext()) {
        openRecordStore(context);
        recordStore.saveRecord(shakespeareDocument);
        shakespeareValueBytes = getSaveIndexValueBytes(recordStore);
        recordStore.saveRecord(germanDocument);
        germanValueBytes = getSaveIndexValueBytes(recordStore) - shakespeareValueBytes;
        commit(context);
    }
    // Save without positions
    try (FDBRecordContext context = openContext()) {
        openRecordStore(context, metaDataBuilder -> {
            metaDataBuilder.removeIndex(TextIndexTestUtils.SIMPLE_DEFAULT_NAME);
            metaDataBuilder.addIndex(SIMPLE_DOC, SIMPLE_TEXT_NO_POSITIONS);
        });
        recordStore.deleteAllRecords();
        recordStore.saveRecord(shakespeareDocument);
        assertThat(getSaveIndexValueBytes(recordStore), lessThan(shakespeareValueBytes));
        final int newShakespeareBytes = getSaveIndexValueBytes(recordStore);
        recordStore.saveRecord(germanDocument);
        assertThat(getSaveIndexValueBytes(recordStore) - newShakespeareBytes, lessThan(germanValueBytes));
        List<Map.Entry<Tuple, List<Integer>>> entryList = scanMapEntries(recordStore, SIMPLE_TEXT_NO_POSITIONS, Tuple.from("gewonnen"));
        assertEquals(Collections.singletonList(entryOf(Tuple.from(1066L), Collections.emptyList())), entryList);
        entryList = scanMapEntries(recordStore, SIMPLE_TEXT_NO_POSITIONS, Tuple.from("dignity"));
        assertEquals(Collections.singletonList(entryOf(Tuple.from(1623L), Collections.emptyList())), entryList);
        commit(context);
    }
}

Example 12 with SimpleDocument

use of com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument in project fdb-record-layer by FoundationDB.

the class TextIndexTest method textIndexPerf1000SerialInsertNoBatching.

@Tag(Tags.Performance)
@Test
public void textIndexPerf1000SerialInsertNoBatching() throws Exception {
    // Create 1000 records
    Random r = new Random();
    List<SimpleDocument> records = getRandomRecords(r, 1000);
    long startTime = System.nanoTime();
    for (int i = 0; i < records.size(); i++) {
        try (FDBRecordContext context = openContext()) {
            openRecordStore(context);
            recordStore.saveRecord(records.get(i));
            commit(context);
        }
    }
    long endTime = System.nanoTime();
    LOGGER.info("performed 1000 serial insertions in {} seconds.", (endTime - startTime) * 1e-9);
    printUsage();
}

Example 13 with SimpleDocument

use of com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument in project fdb-record-layer by FoundationDB.

the class TextIndexTest method queryScanEquivalence.

/**
 * Generate random documents and then make sure that querying them using the index
 * produces the same result as performing a full scan of all records.
 */
@MethodSource("indexArguments")
@ParameterizedTest
public void queryScanEquivalence(@Nonnull Index index) throws Exception {
    final Random r = new Random(0xba5eba1L + index.getName().hashCode());
    final int recordCount = 100;
    final int recordBatch = 25;
    final int queryCount = 25;
    final List<String> lexicon = getStandardLexicon();
    TextTokenizerRegistryImpl.instance().register(FILTERING_TOKENIZER);
    final TextTokenizer tokenizer = TextIndexMaintainer.getTokenizer(index);
    final RecordMetaDataHook hook = metaDataBuilder -> {
        metaDataBuilder.removeIndex(TextIndexTestUtils.SIMPLE_DEFAULT_NAME);
        metaDataBuilder.addIndex(SIMPLE_DOC, index);
    };
    long seed = r.nextLong();
    LOGGER.info(KeyValueLogMessage.of("initializing random number generator", TestLogMessageKeys.SEED, seed));
    r.setSeed(seed);
    for (int i = 0; i < recordCount; i += recordBatch) {
        List<SimpleDocument> records = getRandomRecords(r, recordBatch, lexicon);
        LOGGER.info(KeyValueLogMessage.of("creating and saving random records", TestLogMessageKeys.BATCH_SIZE, recordBatch));
        try (FDBRecordContext context = openContext()) {
            openRecordStore(context, hook);
            records.forEach(recordStore::saveRecord);
            commit(context);
        }
    }
    double[] proportions = getZipfProportions(lexicon);
    long totalScanningTime = 0;
    long totalQueryingTime = 0;
    long totalResults = 0;
    for (int i = 0; i < queryCount; i++) {
        // Generate a random text query
        List<String> tokens = getRandomWords(r, lexicon, proportions, 6, 3);
        String tokenString = String.join(" ", tokens);
        double filterChoice = r.nextDouble();
        final QueryComponent filter;
        if (filterChoice < 0.2) {
            filter = Query.field("text").text(tokenizer.getName()).containsAll(tokenString);
        } else if (filterChoice < 0.4) {
            filter = Query.field("text").text(tokenizer.getName()).containsAny(tokenString);
        } else if (filterChoice < 0.6) {
            filter = Query.field("text").text(tokenizer.getName()).containsPhrase(tokenString);
        } else if (filterChoice < 0.8) {
            int maxDistance = r.nextInt(10) + tokens.size();
            filter = Query.field("text").text(tokenizer.getName()).containsAll(tokenString, maxDistance);
        } else if (filterChoice < 0.9) {
            filter = Query.field("text").text(tokenizer.getName()).containsAnyPrefix(tokenString);
        } else if (filterChoice < 0.95) {
            filter = Query.field("text").text(tokenizer.getName()).containsAllPrefixes(tokenString);
        } else {
            if (tokens.isEmpty()) {
                continue;
            }
            // Choose the first non-empty token from the iterator
            Iterator<? extends CharSequence> tokenIterator = tokenizer.tokenize(tokenString, tokenizer.getMaxVersion(), TextTokenizer.TokenizerMode.QUERY);
            String firstToken = null;
            while (tokenIterator.hasNext()) {
                String nextToken = tokenIterator.next().toString();
                if (!nextToken.isEmpty()) {
                    firstToken = nextToken;
                    break;
                }
            }
            if (firstToken == null) {
                continue;
            }
            int prefixEnd;
            if (firstToken.length() > 1) {
                prefixEnd = r.nextInt(firstToken.length() - 1) + 1;
            } else {
                prefixEnd = 1;
            }
            filter = Query.field("text").text(tokenizer.getName()).containsPrefix(firstToken.substring(0, prefixEnd));
        }
        LOGGER.info(KeyValueLogMessage.of("generated random filter", TestLogMessageKeys.ITERATION, i, LogMessageKeys.FILTER, filter));
        // Manual scan all of the records
        long startTime = System.nanoTime();
        final Set<Long> manualRecordIds = performQueryWithRecordStoreScan(hook, filter);
        long endTime = System.nanoTime();
        LOGGER.info(KeyValueLogMessage.of("manual scan completed", TestLogMessageKeys.SCAN_MILLIS, TimeUnit.MILLISECONDS.convert(endTime - startTime, TimeUnit.NANOSECONDS)));
        totalScanningTime += endTime - startTime;
        // Generate a query and use the index
        startTime = System.nanoTime();
        final Set<Long> queryRecordIds = performQueryWithIndexScan(hook, index, filter);
        endTime = System.nanoTime();
        LOGGER.info(KeyValueLogMessage.of("query completed", TestLogMessageKeys.SCAN_MILLIS, TimeUnit.MILLISECONDS.convert(endTime - startTime, TimeUnit.NANOSECONDS)));
        totalQueryingTime += endTime - startTime;
        if (!manualRecordIds.equals(queryRecordIds)) {
            Set<Long> onlyManual = new HashSet<>(manualRecordIds);
            onlyManual.removeAll(queryRecordIds);
            Set<Long> onlyQuery = new HashSet<>(queryRecordIds);
            onlyManual.removeAll(manualRecordIds);
            LOGGER.warn(KeyValueLogMessage.of("results did not match", LogMessageKeys.FILTER, filter, TestLogMessageKeys.MANUAL_RESULT_COUNT, manualRecordIds.size(), TestLogMessageKeys.QUERY_RESULT_COUNT, queryRecordIds.size(), TestLogMessageKeys.ONLY_MANUAL_COUNT, onlyManual.size(), TestLogMessageKeys.ONLY_QUERY_COUNT, onlyQuery.size()));
        }
        assertEquals(manualRecordIds, queryRecordIds);
        LOGGER.info(KeyValueLogMessage.of("results matched", LogMessageKeys.FILTER, filter, TestLogMessageKeys.RESULT_COUNT, manualRecordIds.size()));
        totalResults += queryRecordIds.size();
    }
    LOGGER.info(KeyValueLogMessage.of("test completed", TestLogMessageKeys.TOTAL_SCAN_MILLIS, TimeUnit.MILLISECONDS.convert(totalScanningTime, TimeUnit.NANOSECONDS), TestLogMessageKeys.TOTAL_QUERY_MILLIS, TimeUnit.MILLISECONDS.convert(totalQueryingTime, TimeUnit.NANOSECONDS), TestLogMessageKeys.TOTAL_RESULT_COUNT, totalResults));
}

Example 14 with SimpleDocument

use of com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument in project fdb-record-layer by FoundationDB.

the class TextIndexTest method scan.

@Test
public void scan() throws Exception {
    final Random r = new Random(0x5ca1ab1e);
    List<SimpleDocument> records = getRandomRecords(r, 50);
    try (FDBRecordContext context = openContext()) {
        openRecordStore(context);
        records.forEach(recordStore::saveRecord);
        commit(context);
    }
    final Index index = recordStore.getRecordMetaData().getIndex(TextIndexTestUtils.SIMPLE_DEFAULT_NAME);
    // existing token
    scanWithZeroScanRecordLimit(index, "angstrom", false);
    // existing token
    scanWithZeroScanRecordLimit(index, "angstrom", true);
    // non-existent token
    scanWithZeroScanRecordLimit(index, "asdfasdf", false);
    // non-existent token
    scanWithZeroScanRecordLimit(index, "asdfasdf", true);
    final List<Integer> limits = Arrays.asList(0, 1, 2, Integer.MAX_VALUE);
    final List<Integer> skips = Arrays.asList(0, 1, 10, 1000);
    final List<String> tokens = Arrays.asList("angstrom", "the", "not_a_token_in_the_lexicon", "שפראך");
    for (int limit : limits) {
        scanMultipleWithScanRecordLimits(index, tokens, limit, false);
        scanMultipleWithScanRecordLimits(index, tokens, limit, true);
        scanWithContinuations(index, "достопримечательности", limit, false);
        scanWithContinuations(index, "достопримечательности", limit, false);
        for (int skip : skips) {
            scanWithSkip(index, "toil", skip, limit, false);
            scanWithSkip(index, "toil", skip, limit, true);
        }
    }
}

Example 15 with SimpleDocument

use of com.apple.foundationdb.record.TestRecordsTextProto.SimpleDocument in project fdb-record-layer by FoundationDB.

the class TextIndexTest method saveSimpleDocumentsWithFilter.

@Test
public void saveSimpleDocumentsWithFilter() throws Exception {
    final SimpleDocument russianDocument = SimpleDocument.newBuilder().setDocId(1547L).setText(TextSamples.RUSSIAN).build();
    try (FDBRecordContext context = openContext()) {
        // Because missing tokenizer
        assertThrows(MetaDataException.class, () -> openRecordStore(context, metaDataBuilder -> {
            metaDataBuilder.removeIndex(TextIndexTestUtils.SIMPLE_DEFAULT_NAME);
            metaDataBuilder.addIndex(SIMPLE_DOC, SIMPLE_TEXT_FILTERING);
        }));
    }
    TextTokenizerRegistryImpl.instance().register(FILTERING_TOKENIZER);
    try (FDBRecordContext context = openContext()) {
        openRecordStore(context, metaDataBuilder -> {
            metaDataBuilder.removeIndex(TextIndexTestUtils.SIMPLE_DEFAULT_NAME);
            metaDataBuilder.addIndex(SIMPLE_DOC, SIMPLE_TEXT_FILTERING);
        });
        recordStore.saveRecord(russianDocument);
        // Note that достопримечательности has been filtered out, so it's probably a
        // lot less interesting to visit.
        assertEquals(4, getSaveIndexKeyCount(recordStore));
        List<Map.Entry<Tuple, List<Integer>>> entryList = scanMapEntries(recordStore, SIMPLE_TEXT_FILTERING, Tuple.from("достопримечательности"));
        assertEquals(Collections.emptyList(), entryList);
        entryList = scanMapEntries(recordStore, SIMPLE_TEXT_FILTERING, Tuple.from("москвы"));
        assertEquals(Collections.singletonList(entryOf(Tuple.from(1547L), Collections.singletonList(4))), entryList);
        commit(context);
    }
}