Examples with RecordCursorResult com.apple.foundationdb.record.RecordCursorResult used on opensource projects

Example 1 with RecordCursorResult

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

the class IndexingScrubMissing method getRecordIfMissingIndex.

@Nullable
private CompletableFuture<FDBStoredRecord<Message>> getRecordIfMissingIndex(FDBRecordStore store, final RecordCursorResult<FDBStoredRecord<Message>> currResult) {
    final FDBStoredRecord<Message> rec = currResult.get();
    // return true if an index is missing and updated
    if (!common.getAllRecordTypes().contains(rec.getRecordType())) {
        return CompletableFuture.completedFuture(null);
    }
    final Index index = common.getIndex();
    final IndexMaintainer maintainer = store.getIndexMaintainer(index);
    List<IndexEntry> indexEntryNoPKs = maintainer.filteredIndexEntries(rec);
    if (indexEntryNoPKs == null) {
        return CompletableFuture.completedFuture(null);
    }
    return AsyncUtil.getAll(indexEntryNoPKs.stream().map(entry -> {
        // should I convert it to a single nested statement?
        final IndexEntry indexEntry = new IndexEntry(index, FDBRecordStoreBase.indexEntryKey(index, entry.getKey(), rec.getPrimaryKey()), entry.getValue());
        final Tuple valueKey = indexEntry.getKey();
        final byte[] keyBytes = maintainer.getIndexSubspace().pack(valueKey);
        return maintainer.state.transaction.get(keyBytes).thenApply(indexVal -> indexVal == null ? valueKey : null);
    }).collect(Collectors.toList())).thenApply(list -> {
        List<Tuple> missingIndexesKeys = list.stream().filter(Objects::nonNull).collect(Collectors.toList());
        if (missingIndexesKeys.isEmpty()) {
            return null;
        }
        // (Maybe) report an error and (maybe) return this record to be index
        if (LOGGER.isWarnEnabled() && logWarningCounter > 0) {
            logWarningCounter--;
            LOGGER.warn(KeyValueLogMessage.build("Scrubber: missing index entry", LogMessageKeys.KEY, rec.getPrimaryKey().toString(), LogMessageKeys.INDEX_KEY, missingIndexesKeys.toString()).addKeysAndValues(common.indexLogMessageKeyValues()).toString());
        }
        missingCount.incrementAndGet();
        final FDBStoreTimer timer = getRunner().getTimer();
        timerIncrement(timer, FDBStoreTimer.Counts.INDEX_SCRUBBER_MISSING_ENTRIES);
        if (scrubbingPolicy.allowRepair()) {
            // record to be indexed
            return rec;
        }
        // report only mode
        return null;
    });
}

Example 2 with RecordCursorResult

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

the class IndexingScrubMissing method scrubRecordsRangeOnly.

@Nonnull
private CompletableFuture<Boolean> scrubRecordsRangeOnly(@Nonnull FDBRecordStore store, byte[] startBytes, byte[] endBytes, @Nonnull AtomicLong recordsScanned) {
    // return false when done
    Index index = common.getIndex();
    final RecordMetaData metaData = store.getRecordMetaData();
    final RecordMetaDataProvider recordMetaDataProvider = common.getRecordStoreBuilder().getMetaDataProvider();
    if (recordMetaDataProvider == null || !metaData.equals(recordMetaDataProvider.getRecordMetaData())) {
        throw new MetaDataException("Store does not have the same metadata");
    }
    final IndexMaintainer maintainer = store.getIndexMaintainer(index);
    // scrubbing only readable, VALUE, idempotence indexes (at least for now)
    validateOrThrowEx(maintainer.isIdempotent(), "scrubbed index is not idempotent");
    validateOrThrowEx(index.getType().equals(IndexTypes.VALUE) || scrubbingPolicy.ignoreIndexTypeCheck(), "scrubbed index is not a VALUE index");
    validateOrThrowEx(store.getIndexState(index) == IndexState.READABLE, "scrubbed index is not readable");
    RangeSet rangeSet = new RangeSet(indexScrubRecordsRangeSubspace(store, index));
    AsyncIterator<Range> ranges = rangeSet.missingRanges(store.ensureContextActive(), startBytes, endBytes).iterator();
    final ExecuteProperties.Builder executeProperties = ExecuteProperties.newBuilder().setIsolationLevel(IsolationLevel.SNAPSHOT).setReturnedRowLimit(// always respectLimit in this path; +1 allows a continuation item
    getLimit() + 1);
    final ScanProperties scanProperties = new ScanProperties(executeProperties.build());
    return ranges.onHasNext().thenCompose(hasNext -> {
        if (Boolean.FALSE.equals(hasNext)) {
            // Here: no more missing ranges - all done
            // To avoid stale metadata, we'll keep the scrubbed-ranges indicator empty until the next scrub call.
            Transaction tr = store.getContext().ensureActive();
            tr.clear(indexScrubRecordsRangeSubspace(store, index).range());
            return AsyncUtil.READY_FALSE;
        }
        final Range range = ranges.next();
        final Tuple rangeStart = RangeSet.isFirstKey(range.begin) ? null : Tuple.fromBytes(range.begin);
        final Tuple rangeEnd = RangeSet.isFinalKey(range.end) ? null : Tuple.fromBytes(range.end);
        final TupleRange tupleRange = TupleRange.between(rangeStart, rangeEnd);
        final RecordCursor<FDBStoredRecord<Message>> cursor = store.scanRecords(tupleRange, null, scanProperties);
        final AtomicBoolean hasMore = new AtomicBoolean(true);
        final AtomicReference<RecordCursorResult<FDBStoredRecord<Message>>> lastResult = new AtomicReference<>(RecordCursorResult.exhausted());
        final long scanLimit = scrubbingPolicy.getEntriesScanLimit();
        // Note that currently we only scrub idempotent indexes
        final boolean isIdempotent = true;
        return iterateRangeOnly(store, cursor, this::getRecordIfMissingIndex, lastResult, hasMore, recordsScanned, isIdempotent).thenApply(vignore -> hasMore.get() ? lastResult.get().get().getPrimaryKey() : rangeEnd).thenCompose(cont -> rangeSet.insertRange(store.ensureContextActive(), packOrNull(rangeStart), packOrNull(cont), true).thenApply(ignore -> {
            if (scanLimit > 0) {
                scanCounter += recordsScanned.get();
                if (scanLimit <= scanCounter) {
                    return false;
                }
            }
            return !Objects.equals(cont, rangeEnd);
        }));
    });
}

Example 3 with RecordCursorResult

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

the class Main method main.

public static void main(String[] args) {
    // Get a database connection.
    FDBDatabase fdb = FDBDatabaseFactory.instance().getDatabase();
    // Create a subspace using the key space API to create a subspace within
    // the cluster used by this record store. The key space API in general
    // allows the user to specify a hierarchical structure of named sub-paths.
    // Each record store can then fill in the named entries within the path
    // with values relevant to that store. If the key space includes a directory
    // layer directory, then the value supplied by the user will be replaced
    // by a short prefix supplied by the the directory layer. The results from
    // the directory layer are cached locally by the Record Layer to avoid excessive
    // database reads.
    // 
    // In this case, the key space implies that there are multiple "applications"
    // that might be defined to run on the same FoundationDB cluster, and then
    // each "application" might have multiple "environments". This could be used,
    // for example, to connect to either the "prod" or "qa" environment for the same
    // application from within the same code base.
    final KeySpace keySpace = new KeySpace(new DirectoryLayerDirectory("application").addSubdirectory(new KeySpaceDirectory("environment", KeySpaceDirectory.KeyType.STRING)));
    // Create a path for the "record-layer-sample" application's demo environment.
    // Clear all existing data and then return the subspace associated with the key space path.
    final KeySpacePath path = keySpace.path("application", "record-layer-sample").add("environment", "demo");
    // Clear out any data that may be in the record store.
    LOGGER.info("Clearing the Record Store ...");
    fdb.run(context -> {
        path.deleteAllData(context);
        return null;
    });
    // Build the metadata. This simple approach only works for primary
    // keys and secondary indexes defined in the Protobuf message types.
    RecordMetaData rmd = RecordMetaData.build(SampleProto.getDescriptor());
    FDBRecordStore.Builder recordStoreBuilder = FDBRecordStore.newBuilder().setMetaDataProvider(rmd).setKeySpacePath(path);
    // Write records for Vendor and Item.
    LOGGER.info("Writing Vendor and Item record ...");
    fdb.run((FDBRecordContext cx) -> {
        FDBRecordStore store = recordStoreBuilder.copyBuilder().setContext(cx).create();
        store.saveRecord(SampleProto.Vendor.newBuilder().setVendorId(9375L).setVendorName("Acme").build());
        store.saveRecord(SampleProto.Vendor.newBuilder().setVendorId(1066L).setVendorName("Buy n Large").build());
        store.saveRecord(SampleProto.Item.newBuilder().setItemId(4836L).setItemName("GPS").setVendorId(9375L).build());
        store.saveRecord(SampleProto.Item.newBuilder().setItemId(9970L).setItemName("Personal Transport").setVendorId(1066L).build());
        store.saveRecord(SampleProto.Item.newBuilder().setItemId(8380L).setItemName("Piles of Garbage").setVendorId(1066L).build());
        return null;
    });
    // Use the primary key declared in the Vendor message type to read a
    // record.
    LOGGER.info("Reading Vendor record with primary key 9375L ...");
    SampleProto.Vendor.Builder readBuilder = fdb.run((FDBRecordContext cx) -> {
        FDBRecordStore store = recordStoreBuilder.copyBuilder().setContext(cx).open();
        return SampleProto.Vendor.newBuilder().mergeFrom(store.loadRecord(Key.Evaluated.scalar(9375L).toTuple()).getRecord());
    });
    LOGGER.info("    Result -> Id: {}, Name: {}", readBuilder.getVendorId(), readBuilder.getVendorName());
    // Using the secondary index declared in the message type, query
    // Item by vendor ID, then look up the item ID.
    LOGGER.info("Looking for item IDs with vendor ID 9375L ...");
    ArrayList<Long> ids = fdb.run((FDBRecordContext cx) -> {
        ArrayList<Long> itemIDs = new ArrayList<>();
        FDBRecordStore store = recordStoreBuilder.copyBuilder().setContext(cx).open();
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Item").setFilter(Query.field("vendor_id").equalsValue(9375L)).build();
        try (RecordCursor<FDBQueriedRecord<Message>> cursor = store.executeQuery(query)) {
            RecordCursorResult<FDBQueriedRecord<Message>> result;
            do {
                result = cursor.getNext();
                if (result.hasNext()) {
                    itemIDs.add(SampleProto.Item.newBuilder().mergeFrom(result.get().getRecord()).getItemId());
                }
            } while (result.hasNext());
        }
        return itemIDs;
    });
    ids.forEach((Long res) -> LOGGER.info("    Result -> Vendor ID: 9375, Item ID: {}", res));
    // A kind of hand-crafted "cross-table join" (in some sense). This returns a list
    // linking the name of each vendor to the names of the products they sell.
    // Note that this query is entirely non-blocking until the end.
    // In SQL, this might look something like:
    // 
    // SELECT Vendor.name, Item.name FROM Item JOIN Vendor ON Vendor.vendor_id = Item.vid
    // 
    // One difference is that the above SQL query will flatten the results out so that there
    // is exactly one returned row per item name (per vendor) where as the map returned by
    // this RecordLayer query will feature exactly one entry per vendor where the key is the
    // vendor name and the value is the vendor's items.
    // 
    // Note that this query is not particularly efficient as is. To make this efficient, one
    // might consider an index on vendor name. This could scan the index to get the vendor
    // name of the Vendor record type and a second index on item by vendor ID, perhaps with
    // the item name in the value portion of the index definition. This would allow the
    // query to be satisfied with one scan of the vendor name index and another scan of the
    // item's vendor ID index (one scan per vendor).
    LOGGER.info("Grouping items by vendor ...");
    Map<String, List<String>> namesToItems = fdb.run((FDBRecordContext cx) -> cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(store -> {
        // Outer plan gets all of the vendors
        RecordQueryPlan outerPlan = store.planQuery(RecordQuery.newBuilder().setRecordType("Vendor").setRequiredResults(Arrays.asList(field("vendor_id"), field("vendor_name"))).build());
        // Inner plan gets all items for the given vendor ID.
        // Using "equalsParameter" does the plan once and re-uses the plan for each vendor ID.
        RecordQueryPlan innerPlan = store.planQuery(RecordQuery.newBuilder().setRecordType("Item").setRequiredResults(Collections.singletonList(field("item_name"))).setFilter(Query.field("vendor_id").equalsParameter("vid")).build());
        return store.executeQuery(outerPlan).mapPipelined(record -> {
            SampleProto.Vendor vendor = SampleProto.Vendor.newBuilder().mergeFrom(record.getRecord()).build();
            return innerPlan.execute(store, EvaluationContext.forBinding("vid", vendor.getVendorId())).map(innerRecord -> SampleProto.Item.newBuilder().mergeFrom(innerRecord.getRecord()).getItemName()).asList().thenApply(list -> Pair.of(vendor.getVendorName(), list));
        }, 10).asList().thenApply((List<Pair<String, List<String>>> list) -> list.stream().collect(Collectors.toMap(Pair::getKey, Pair::getValue)));
    })));
    namesToItems.forEach((String name, List<String> items) -> LOGGER.info("    Result -> Vendor Name: {}, Item names: {}", name, items));
    // Richer indexes:
    // To build richer primary keys or secondary indexes (than those definable in the protobuf
    // message types), you need to use the more verbose and powerful RecordMetaDataBuilder.
    RecordMetaDataBuilder rmdBuilder = RecordMetaData.newBuilder().setRecords(SampleProto.getDescriptor());
    // Order customers by last name, then first name, then their ID if otherwise equal.
    // NOTE: This operation is dangerous if you have existing data! Existing records are *not*
    // automatically migrated.
    rmdBuilder.getRecordType("Customer").setPrimaryKey(concatenateFields("last_name", "first_name", "customer_id"));
    // Add a global count index. Most record stores should probably add this index as it allows
    // the database to make intelligent decisions based on the current size of the record store.
    rmdBuilder.addUniversalIndex(new Index("globalCount", new GroupingKeyExpression(EmptyKeyExpression.EMPTY, 0), IndexTypes.COUNT));
    // Add a FanType.FanOut secondary index for email_address, so that
    // each value for email_address generates its own key in the index.
    rmdBuilder.addIndex("Customer", new Index("email_address", field("email_address", FanType.FanOut), IndexTypes.VALUE));
    // Add a FanType.Concatenate secondary index for preference_tag, so
    // that all values for preference_tag generate a single key in the index.
    rmdBuilder.addIndex("Customer", new Index("preference_tag", field("preference_tag", FanType.Concatenate), IndexTypes.VALUE));
    // Add an index on the count of each preference tag. This allows us to
    // quickly get the number of users for each preference tag. The key
    // provided will create a separate "count" field for each value of the
    // preference_tag field and keep track of the number of customer
    // records with each value.
    rmdBuilder.addIndex("Customer", new Index("preference_tag_count", new GroupingKeyExpression(field("preference_tag", FanType.FanOut), 0), IndexTypes.COUNT));
    // Add a nested secondary index for order such that each value for
    // quantity in Order generates a single key in the index.
    rmdBuilder.addIndex("Customer", new Index("order", field("order", FanType.FanOut).nest("quantity"), IndexTypes.VALUE));
    // Add an index on the sum of the quantity of each item in each
    // order. This can be used to know how many of each item have been ordered across
    // all customers. The grouping key here is a little hairy, but it
    // specifies that the "item_id" column should be used as a grouping key
    // and the quantity used as the sum value, so it will keep track of the
    // quantity ordered of each item.
    rmdBuilder.addIndex("Customer", new Index("item_quantity_sum", new GroupingKeyExpression(field("order", FanType.FanOut).nest(concatenateFields("item_id", "quantity")), 1), IndexTypes.SUM));
    // Rebuild the metadata for the newly added indexes before reading or
    // writing more data.
    RecordMetaData rmd2 = rmdBuilder.getRecordMetaData();
    recordStoreBuilder.setMetaDataProvider(rmd2);
    // Calling "open" on an existing record store with new meta-data will
    // create the index and place them in a "disabled" mode that means that
    // they cannot yet be used for queries. (In particular, the query planner
    // will ignore this index and any attempt to read from the index will
    // throw an error.) To enable querying, one must invoke the online index
    // builder. This will scan through the record store across multiple
    // transactions and populate the new indexes with data from the existing
    // entries. During the build job, the record store remains available for
    // reading and writing, but there may be additional conflicts if the index
    // build job and normal operations happen to mutate the same records.
    RecordStoreState storeState = fdb.run(cx -> {
        FDBRecordStore store = recordStoreBuilder.copyBuilder().setContext(cx).open();
        return store.getRecordStoreState();
    });
    LOGGER.info("Running index builds of new indexes:");
    // Build all of the indexes in parallel by firing off a future for each and
    // then wait for all of them.
    fdb.asyncToSync(null, FDBStoreTimer.Waits.WAIT_ONLINE_BUILD_INDEX, AsyncUtil.whenAll(storeState.getDisabledIndexNames().stream().map(indexName -> {
        // Build this index. It will begin the background job and return a future
        // that will complete when the index is ready for querying.
        OnlineIndexer indexBuilder = OnlineIndexer.newBuilder().setDatabase(fdb).setRecordStoreBuilder(recordStoreBuilder).setIndex(indexName).build();
        return indexBuilder.buildIndexAsync().thenRun(() -> LOGGER.info("  Index build of {} is complete.", indexName)).whenComplete((vignore, eignore) -> indexBuilder.close());
    }).collect(Collectors.toList())));
    // Write larger records for Customer (and Order).
    LOGGER.info("Adding records with new secondary indexes ...");
    fdb.run((FDBRecordContext cx) -> {
        FDBRecordStore store = recordStoreBuilder.copyBuilder().setContext(cx).open();
        store.saveRecord(SampleProto.Customer.newBuilder().setCustomerId(9264L).setFirstName("John").setLastName("Smith").addEmailAddress("jsmith@example.com").addEmailAddress("john_smith@example.com").addPreferenceTag("books").addPreferenceTag("movies").addOrder(SampleProto.Order.newBuilder().setOrderId(3875L).setItemId(9374L).setQuantity(2)).addOrder(SampleProto.Order.newBuilder().setOrderId(4828L).setItemId(2740L).setQuantity(1)).setPhoneNumber("(703) 555-8255").build());
        store.saveRecord(SampleProto.Customer.newBuilder().setCustomerId(8365L).setFirstName("Jane").setLastName("Doe").addEmailAddress("jdoe@example.com").addEmailAddress("jane_doe@example.com").addPreferenceTag("games").addPreferenceTag("lawn").addPreferenceTag("books").addOrder(SampleProto.Order.newBuilder().setOrderId(9280L).setItemId(2740L).setQuantity(3)).setPhoneNumber("(408) 555-0248").build());
        return null;
    });
    // Get the record count. This uses the global count index to get the
    // full number of records in the store.
    Long recordCount = fdb.run((FDBRecordContext cx) -> cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(FDBRecordStore::getSnapshotRecordCount)));
    LOGGER.info("Store contains {} records.", recordCount);
    // Query all records with the first name "Jane".
    // Performs a full scan of the primary key index.
    LOGGER.info("Retrieving all customers with first name \"Jane\"...");
    List<String> names = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.field("first_name").equalsValue("Jane")).build();
        return readNames(recordStoreBuilder, cx, query);
    });
    names.forEach((String res) -> LOGGER.info("    Result -> {}", res));
    // Query all records with last name "Doe".
    // Scans only the customers from the primary key index.
    LOGGER.info("Retrieving all customers with last name \"Doe\"...");
    names = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.field("last_name").equalsValue("Doe")).build();
        return readNames(recordStoreBuilder, cx, query);
    });
    names.forEach((String res) -> LOGGER.info("    Result -> {}", res));
    // Query all records with first_name "Jane" and last_name "Doe"
    // Scans only the customers from the primary key index.
    LOGGER.info("Retrieving all customers with name \"Jane Doe\"...");
    names = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.and(Query.field("first_name").equalsValue("Jane"), Query.field("last_name").equalsValue("Doe"))).build();
        return readNames(recordStoreBuilder, cx, query);
    });
    names.forEach((String res) -> LOGGER.info("    Result -> {}", res));
    // Query all records with an email address beginning with "john".
    // Uses FanType.FanOut secondary index.
    LOGGER.info("Retrieving all customers with an email address beginning with \"john\"...");
    Map<String, List<String>> addresses = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.field("email_address").oneOfThem().startsWith("john")).build();
        return cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(store -> {
            Map<String, List<String>> addressMap = new HashMap<>();
            return store.executeQuery(query).forEach((FDBQueriedRecord<Message> record) -> {
                SampleProto.Customer.Builder builder = SampleProto.Customer.newBuilder().mergeFrom(record.getRecord());
                addressMap.put(builder.getFirstName() + " " + builder.getLastName(), builder.getEmailAddressList());
            }).thenApply(v -> addressMap);
        }));
    });
    addresses.forEach((String k, List<String> vals) -> LOGGER.info("    Result -> {} with emails {}", k, vals));
    // Query all records with preference_tags "books" and "movies".
    // Uses FanType.Concatenate secondary index.
    LOGGER.info("Retrieving all customers with preference tags \"books\" and \"movies\"...");
    names = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.and(Query.field("preference_tag").oneOfThem().equalsValue("books"), Query.field("preference_tag").oneOfThem().equalsValue("movies"))).build();
        return readNames(recordStoreBuilder, cx, query);
    });
    names.forEach((String res) -> LOGGER.info("    Result -> {}", res));
    // Get the number of customers who have "books" listed as one of their preference tags
    Long bookPreferenceCount = fdb.run((FDBRecordContext cx) -> cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(store -> {
        Index index = store.getRecordMetaData().getIndex("preference_tag_count");
        IndexAggregateFunction function = new IndexAggregateFunction(FunctionNames.COUNT, index.getRootExpression(), index.getName());
        return store.evaluateAggregateFunction(Collections.singletonList("Customer"), function, Key.Evaluated.scalar("books"), IsolationLevel.SERIALIZABLE).thenApply(tuple -> tuple.getLong(0));
    })));
    LOGGER.info("Number of customers with the \"books\" preference tag: {}", bookPreferenceCount);
    // Query all customers with an order of quantity greater than 2.
    // Uses nested secondary index.
    LOGGER.info("Retrieving all customers with an order of quantity greater than 2 ...");
    names = fdb.run((FDBRecordContext cx) -> {
        RecordQuery query = RecordQuery.newBuilder().setRecordType("Customer").setFilter(Query.field("order").oneOfThem().matches(Query.field("quantity").greaterThan(2))).build();
        return readNames(recordStoreBuilder, cx, query);
    });
    names.forEach((String res) -> LOGGER.info("    Result -> {}", res));
    // Get the sum of the quantity of items ordered for item ID 2740.
    // Using the index, it can determine this by reading a single
    // key in the database.
    Long itemQuantitySum = fdb.run((FDBRecordContext cx) -> cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(store -> {
        Index index = store.getRecordMetaData().getIndex("item_quantity_sum");
        IndexAggregateFunction function = new IndexAggregateFunction(FunctionNames.SUM, index.getRootExpression(), index.getName());
        return store.evaluateAggregateFunction(Collections.singletonList("Customer"), function, Key.Evaluated.scalar(2740L), IsolationLevel.SERIALIZABLE).thenApply(tuple -> tuple.getLong(0));
    })));
    LOGGER.info("Total quantity ordered of item 2740L: {}", itemQuantitySum);
    // Get the sum of the quantity of all items ordered.
    // Using the index, it will do a scan that will hit one key
    // for each unique item id with a single range scan.
    Long allItemsQuantitySum = fdb.run((FDBRecordContext cx) -> cx.asyncToSync(FDBStoreTimer.Waits.WAIT_EXECUTE_QUERY, recordStoreBuilder.copyBuilder().setContext(cx).openAsync().thenCompose(store -> {
        Index index = store.getRecordMetaData().getIndex("item_quantity_sum");
        IndexAggregateFunction function = new IndexAggregateFunction(FunctionNames.SUM, index.getRootExpression(), index.getName());
        return store.evaluateAggregateFunction(Collections.singletonList("Customer"), function, TupleRange.ALL, IsolationLevel.SERIALIZABLE).thenApply(tuple -> tuple.getLong(0));
    })));
    LOGGER.info("Total quantity ordered of all items: {}", allItemsQuantitySum);
}

Example 4 with RecordCursorResult

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

the class IndexingByRecords method buildRangeOnly.

@Nonnull
private CompletableFuture<Tuple> buildRangeOnly(@Nonnull FDBRecordStore store, @Nonnull TupleRange range, boolean respectLimit, @Nullable AtomicLong recordsScanned) {
    validateSameMetadataOrThrow(store);
    Index index = common.getIndex();
    int limit = getLimit();
    final IndexMaintainer maintainer = store.getIndexMaintainer(index);
    final boolean isIdempotent = maintainer.isIdempotent();
    final ExecuteProperties.Builder executeProperties = ExecuteProperties.newBuilder().setIsolationLevel(isIdempotent ? IsolationLevel.SNAPSHOT : IsolationLevel.SERIALIZABLE);
    if (respectLimit) {
        // +1 allows continuation item
        executeProperties.setReturnedRowLimit(limit + 1);
    }
    final ScanProperties scanProperties = new ScanProperties(executeProperties.build());
    final RecordCursor<FDBStoredRecord<Message>> cursor = store.scanRecords(range, null, scanProperties);
    final AtomicBoolean hasMore = new AtomicBoolean(true);
    // Note: This runs all of the updates serially in order to avoid invoking a race condition
    // in the rank code that was causing incorrect results. If everything were thread safe,
    // a larger pipeline size would be possible.
    final AtomicReference<RecordCursorResult<FDBStoredRecord<Message>>> lastResult = new AtomicReference<>(RecordCursorResult.exhausted());
    return iterateRangeOnly(store, cursor, this::getRecordIfTypeMatch, lastResult, hasMore, recordsScanned, isIdempotent).thenApply(vignore -> hasMore.get() ? lastResult.get().get().getPrimaryKey() : range.getHigh());
}

Example 5 with RecordCursorResult

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

the class MapPipelinedCursor method tryToFillPipeline.

/**
 * Take items from inner cursor and put in pipeline until no more or a mapping result is available.
 * @return a future that will complete with {@code false} if an item is available or none will ever be, or with {@code true} if this method should be called to try again
 */
protected CompletableFuture<Boolean> tryToFillPipeline() {
    while (!innerExhausted && pipeline.size() < pipelineSize) {
        // try to add a future to the pipeline
        if (waitInnerFuture == null) {
            waitInnerFuture = inner.onNext();
        }
        if (!waitInnerFuture.isDone()) {
            // still waiting for inner future, check back once something has finished
            CompletableFuture<RecordCursorResult<V>> nextEntry = pipeline.peek();
            if (nextEntry == null) {
                // loop back to process inner result
                return waitInnerFuture.thenApply(vignore -> true);
            } else {
                // keep looping unless the next entry is done
                return CompletableFuture.anyOf(waitInnerFuture, nextEntry).thenApply(vignore -> !nextEntry.isDone());
            }
        }
        // future is ready, doesn't block
        final RecordCursorResult<T> innerResult = waitInnerFuture.join();
        pipeline.add(innerResult.mapAsync(func));
        if (innerResult.hasNext()) {
            // just added something to the pipeline, so pipeline will contain an entry
            // done with this future, should advanced cursor next time
            waitInnerFuture = null;
            if (pipeline.peek().isDone()) {
                // next entry ready, don't loop
                return AsyncUtil.READY_FALSE;
            }
        // otherwise, keep looping
        } else {
            // don't have next, and won't ever with this cursor
            innerExhausted = true;
            if (innerResult.getNoNextReason() == NoNextReason.TIME_LIMIT_REACHED && nextResult != null) {
                // Under time pressure, do not want to wait for any futures to complete.
                // For other out-of-band reasons, still return results from the futures that were
                // already started.
                // Cannot do this for the very first entry, because do not have a continuation before that.
                RecordCursorContinuation lastFinishedContinuation = cancelPendingFutures();
                pipeline.add(CompletableFuture.completedFuture(RecordCursorResult.withoutNextValue(lastFinishedContinuation, NoNextReason.TIME_LIMIT_REACHED)));
            }
            // Wait for next entry, as if pipeline were full
            break;
        }
    }
    // the next result is ready
    return pipeline.peek().thenApply(vignore -> false);
}