Examples with IncrementalExecutorsInitialiser io.siddhi.core.aggregation.IncrementalExecutorsInitialiser used on opensource projects

Example 1 with IncrementalExecutorsInitialiser

use of io.siddhi.core.aggregation.IncrementalExecutorsInitialiser in project siddhi by wso2.

the class AggregationParser method parse.

public static AggregationRuntime parse(AggregationDefinition aggregationDefinition, SiddhiAppContext siddhiAppContext, Map<String, AbstractDefinition> streamDefinitionMap, Map<String, AbstractDefinition> tableDefinitionMap, Map<String, AbstractDefinition> windowDefinitionMap, Map<String, AbstractDefinition> aggregationDefinitionMap, Map<String, Table> tableMap, Map<String, Window> windowMap, Map<String, AggregationRuntime> aggregationMap, SiddhiAppRuntimeBuilder siddhiAppRuntimeBuilder) {
    // set timeZone for aggregation
    String timeZone = getTimeZone(siddhiAppContext);
    boolean isDebugEnabled = log.isDebugEnabled();
    boolean isPersistedAggregation = false;
    boolean isReadOnly = false;
    if (!validateTimeZone(timeZone)) {
        throw new SiddhiAppCreationException("Given timeZone '" + timeZone + "' for aggregations is invalid. Please provide a valid time zone.");
    }
    // get aggregation name
    String aggregatorName = aggregationDefinition.getId();
    if (isDebugEnabled) {
        log.debug("Incremental aggregation initialization process started for aggregation " + aggregatorName);
    }
    Annotation aggregationProperties = AnnotationHelper.getAnnotation(ANNOTATION_PERSISTED_AGGREGATION, aggregationDefinition.getAnnotations());
    if (aggregationProperties != null) {
        String persistedAggregationMode = aggregationProperties.getElement(ANNOTATION_ELEMENT_ENABLE);
        isPersistedAggregation = persistedAggregationMode == null || Boolean.parseBoolean(persistedAggregationMode);
        String readOnlyMode = aggregationProperties.getElement(ANNOTATION_ELEMENT_IS_READ_ONLY);
        isReadOnly = Boolean.parseBoolean(readOnlyMode);
    }
    if (isPersistedAggregation) {
        aggregationDefinition.getSelector().getSelectionList().stream().forEach(outputAttribute -> {
            if (outputAttribute.getExpression() instanceof AttributeFunction && ((AttributeFunction) outputAttribute.getExpression()).getName().equals("distinctCount")) {
                throw new SiddhiAppCreationException("Aggregation function 'distinctCount' does not supported " + "with persisted aggregation type please use default incremental aggregation");
            }
        });
    }
    if (isDebugEnabled) {
        log.debug("Aggregation mode is defined as " + (isPersistedAggregation ? "persisted" : "inMemory") + " for aggregation " + aggregatorName);
    }
    if (aggregationDefinition.getTimePeriod() == null) {
        throw new SiddhiAppCreationException("Aggregation Definition '" + aggregationDefinition.getId() + "'s timePeriod is null. " + "Hence, can't create the siddhi app '" + siddhiAppContext.getName() + "'", aggregationDefinition.getQueryContextStartIndex(), aggregationDefinition.getQueryContextEndIndex());
    }
    if (aggregationDefinition.getSelector() == null) {
        throw new SiddhiAppCreationException("Aggregation Definition '" + aggregationDefinition.getId() + "'s selection is not defined. " + "Hence, can't create the siddhi app '" + siddhiAppContext.getName() + "'", aggregationDefinition.getQueryContextStartIndex(), aggregationDefinition.getQueryContextEndIndex());
    }
    if (streamDefinitionMap.get(aggregationDefinition.getBasicSingleInputStream().getStreamId()) == null) {
        throw new SiddhiAppCreationException("Stream " + aggregationDefinition.getBasicSingleInputStream().getStreamId() + " has not been defined");
    }
    // Check if the user defined primary keys exists for @store annotation on aggregation if yes, an error is
    // is thrown
    Element userDefinedPrimaryKey = AnnotationHelper.getAnnotationElement(SiddhiConstants.ANNOTATION_PRIMARY_KEY, null, aggregationDefinition.getAnnotations());
    if (userDefinedPrimaryKey != null) {
        throw new SiddhiAppCreationException("Aggregation Tables have predefined primary key, but found '" + userDefinedPrimaryKey.getValue() + "' primary key defined though annotation.");
    }
    try {
        List<VariableExpressionExecutor> incomingVariableExpressionExecutors = new ArrayList<>();
        SiddhiQueryContext siddhiQueryContext = new SiddhiQueryContext(siddhiAppContext, aggregatorName);
        StreamRuntime streamRuntime = InputStreamParser.parse(aggregationDefinition.getBasicSingleInputStream(), null, streamDefinitionMap, tableDefinitionMap, windowDefinitionMap, aggregationDefinitionMap, tableMap, windowMap, aggregationMap, incomingVariableExpressionExecutors, false, siddhiQueryContext);
        // Get original meta for later use.
        MetaStreamEvent incomingMetaStreamEvent = (MetaStreamEvent) streamRuntime.getMetaComplexEvent();
        // Create new meta stream event.
        // This must hold the timestamp, group by attributes (if given) and the incremental attributes, in
        // onAfterWindowData array
        // Example format: AGG_TIMESTAMP, groupByAttribute1, groupByAttribute2, AGG_incAttribute1, AGG_incAttribute2
        // AGG_incAttribute1, AGG_incAttribute2 would have the same attribute names as in
        // finalListOfIncrementalAttributes
        // To enter data as onAfterWindowData
        incomingMetaStreamEvent.initializeOnAfterWindowData();
        // List of all aggregationDurations
        List<TimePeriod.Duration> aggregationDurations = getSortedPeriods(aggregationDefinition.getTimePeriod(), isPersistedAggregation);
        // Incoming executors will be executors for timestamp, externalTimestamp(if used),
        List<ExpressionExecutor> incomingExpressionExecutors = new ArrayList<>();
        List<IncrementalAttributeAggregator> incrementalAttributeAggregators = new ArrayList<>();
        // group by attributes (if given) and the incremental attributes expression executors
        List<Variable> groupByVariableList = aggregationDefinition.getSelector().getGroupByList();
        // Expressions to get final aggregate outputs. e.g for avg the expression is Divide expression with
        // AGG_SUM/ AGG_COUNT
        List<Expression> outputExpressions = new ArrayList<>();
        boolean isProcessingOnExternalTime = aggregationDefinition.getAggregateAttribute() != null;
        boolean isGroupBy = aggregationDefinition.getSelector().getGroupByList().size() != 0;
        final boolean isDistributed;
        ConfigManager configManager = siddhiAppContext.getSiddhiContext().getConfigManager();
        final String shardId = configManager.extractProperty("shardId");
        boolean enablePartitioning = false;
        // check if the setup is Active Active(distributed deployment) by checking availability of partitionById
        // config
        Annotation partitionById = AnnotationHelper.getAnnotation(ANNOTATION_PARTITION_BY_ID, aggregationDefinition.getAnnotations());
        if (partitionById != null) {
            String enableElement = partitionById.getElement(ANNOTATION_ELEMENT_ENABLE);
            enablePartitioning = enableElement == null || Boolean.parseBoolean(enableElement);
        }
        boolean shouldPartitionById = Boolean.parseBoolean(configManager.extractProperty("partitionById"));
        if (enablePartitioning || shouldPartitionById) {
            if (shardId == null) {
                throw new SiddhiAppCreationException("Configuration 'shardId' not provided for @partitionById " + "annotation");
            }
            isDistributed = true;
        } else {
            isDistributed = false;
        }
        if (isDebugEnabled) {
            log.debug("Distributed aggregation processing is " + (isDistributed ? "enabled" : "disabled") + " in " + aggregatorName + " aggregation ");
        }
        populateIncomingAggregatorsAndExecutors(aggregationDefinition, siddhiQueryContext, tableMap, incomingVariableExpressionExecutors, incomingMetaStreamEvent, incomingExpressionExecutors, incrementalAttributeAggregators, groupByVariableList, outputExpressions, isProcessingOnExternalTime, isDistributed, shardId);
        // check if the populateIncomingAggregatorsAndExecutors process has been completed successfully
        boolean isLatestEventColAdded = incomingMetaStreamEvent.getOutputData().get(incomingMetaStreamEvent.getOutputData().size() - 1).getName().equals(AGG_LAST_TIMESTAMP_COL);
        int baseAggregatorBeginIndex = incomingMetaStreamEvent.getOutputData().size();
        List<Expression> finalBaseExpressions = new ArrayList<>();
        boolean isOptimisedLookup = populateFinalBaseAggregators(tableMap, incomingVariableExpressionExecutors, incomingMetaStreamEvent, incomingExpressionExecutors, incrementalAttributeAggregators, siddhiQueryContext, finalBaseExpressions);
        if (isDebugEnabled) {
            log.debug("Optimised lookup mode is " + (isOptimisedLookup ? "enabled" : "disabled") + " for " + "aggregation " + aggregatorName);
        }
        // Creating an intermediate stream with aggregated stream and above extracted output variables
        StreamDefinition incomingOutputStreamDefinition = StreamDefinition.id(aggregatorName + "_intermediate");
        incomingOutputStreamDefinition.setQueryContextStartIndex(aggregationDefinition.getQueryContextStartIndex());
        incomingOutputStreamDefinition.setQueryContextEndIndex(aggregationDefinition.getQueryContextEndIndex());
        MetaStreamEvent processedMetaStreamEvent = new MetaStreamEvent();
        for (Attribute attribute : incomingMetaStreamEvent.getOutputData()) {
            incomingOutputStreamDefinition.attribute(attribute.getName(), attribute.getType());
            processedMetaStreamEvent.addOutputData(attribute);
        }
        incomingMetaStreamEvent.setOutputDefinition(incomingOutputStreamDefinition);
        processedMetaStreamEvent.addInputDefinition(incomingOutputStreamDefinition);
        processedMetaStreamEvent.setOutputDefinition(incomingOutputStreamDefinition);
        // Executors of processing meta
        List<VariableExpressionExecutor> processVariableExpressionExecutors = new ArrayList<>();
        Map<TimePeriod.Duration, List<ExpressionExecutor>> processExpressionExecutorsMap = new HashMap<>();
        Map<TimePeriod.Duration, List<ExpressionExecutor>> processExpressionExecutorsMapForFind = new HashMap<>();
        aggregationDurations.forEach(incrementalDuration -> {
            processExpressionExecutorsMap.put(incrementalDuration, constructProcessExpressionExecutors(siddhiQueryContext, tableMap, baseAggregatorBeginIndex, finalBaseExpressions, incomingOutputStreamDefinition, processedMetaStreamEvent, processVariableExpressionExecutors, isProcessingOnExternalTime, incrementalDuration, isDistributed, shardId, isLatestEventColAdded));
            processExpressionExecutorsMapForFind.put(incrementalDuration, constructProcessExpressionExecutors(siddhiQueryContext, tableMap, baseAggregatorBeginIndex, finalBaseExpressions, incomingOutputStreamDefinition, processedMetaStreamEvent, processVariableExpressionExecutors, isProcessingOnExternalTime, incrementalDuration, isDistributed, shardId, isLatestEventColAdded));
        });
        ExpressionExecutor shouldUpdateTimestamp = null;
        if (isLatestEventColAdded) {
            Expression shouldUpdateTimestampExp = new Variable(AGG_LAST_TIMESTAMP_COL);
            shouldUpdateTimestamp = ExpressionParser.parseExpression(shouldUpdateTimestampExp, processedMetaStreamEvent, 0, tableMap, processVariableExpressionExecutors, false, 0, ProcessingMode.BATCH, false, siddhiQueryContext);
        }
        List<ExpressionExecutor> outputExpressionExecutors = outputExpressions.stream().map(expression -> ExpressionParser.parseExpression(expression, processedMetaStreamEvent, 0, tableMap, processVariableExpressionExecutors, isGroupBy, 0, ProcessingMode.BATCH, false, siddhiQueryContext)).collect(Collectors.toList());
        // Create group by key generator
        Map<TimePeriod.Duration, GroupByKeyGenerator> groupByKeyGeneratorMap = new HashMap<>();
        aggregationDurations.forEach(incrementalDuration -> {
            GroupByKeyGenerator groupByKeyGenerator = null;
            if (isProcessingOnExternalTime || isGroupBy) {
                List<Expression> groupByExpressionList = new ArrayList<>();
                if (isProcessingOnExternalTime) {
                    Expression externalTimestampExpression = AttributeFunction.function("incrementalAggregator", "getAggregationStartTime", new Variable(AGG_EXTERNAL_TIMESTAMP_COL), new StringConstant(incrementalDuration.name()));
                    groupByExpressionList.add(externalTimestampExpression);
                }
                groupByExpressionList.addAll(groupByVariableList.stream().map(groupByVariable -> (Expression) groupByVariable).collect(Collectors.toList()));
                groupByKeyGenerator = new GroupByKeyGenerator(groupByExpressionList, processedMetaStreamEvent, SiddhiConstants.UNKNOWN_STATE, tableMap, processVariableExpressionExecutors, siddhiQueryContext);
            }
            groupByKeyGeneratorMap.put(incrementalDuration, groupByKeyGenerator);
        });
        // GroupBy for reading
        Map<TimePeriod.Duration, GroupByKeyGenerator> groupByKeyGeneratorMapForReading = new HashMap<>();
        if (isDistributed && !isProcessingOnExternalTime) {
            aggregationDurations.forEach(incrementalDuration -> {
                List<Expression> groupByExpressionList = new ArrayList<>();
                Expression timestampExpression = AttributeFunction.function("incrementalAggregator", "getAggregationStartTime", new Variable(AGG_START_TIMESTAMP_COL), new StringConstant(incrementalDuration.name()));
                groupByExpressionList.add(timestampExpression);
                if (isGroupBy) {
                    groupByExpressionList.addAll(groupByVariableList.stream().map(groupByVariable -> (Expression) groupByVariable).collect(Collectors.toList()));
                }
                GroupByKeyGenerator groupByKeyGenerator = new GroupByKeyGenerator(groupByExpressionList, processedMetaStreamEvent, SiddhiConstants.UNKNOWN_STATE, tableMap, processVariableExpressionExecutors, siddhiQueryContext);
                groupByKeyGeneratorMapForReading.put(incrementalDuration, groupByKeyGenerator);
            });
        } else {
            groupByKeyGeneratorMapForReading.putAll(groupByKeyGeneratorMap);
        }
        // Create new scheduler
        EntryValveExecutor entryValveExecutor = new EntryValveExecutor(siddhiAppContext);
        LockWrapper lockWrapper = new LockWrapper(aggregatorName);
        lockWrapper.setLock(new ReentrantLock());
        Scheduler scheduler = SchedulerParser.parse(entryValveExecutor, siddhiQueryContext);
        scheduler.init(lockWrapper, aggregatorName);
        scheduler.setStreamEventFactory(new StreamEventFactory(processedMetaStreamEvent));
        QueryParserHelper.reduceMetaComplexEvent(incomingMetaStreamEvent);
        QueryParserHelper.reduceMetaComplexEvent(processedMetaStreamEvent);
        QueryParserHelper.updateVariablePosition(incomingMetaStreamEvent, incomingVariableExpressionExecutors);
        QueryParserHelper.updateVariablePosition(processedMetaStreamEvent, processVariableExpressionExecutors);
        Map<TimePeriod.Duration, Table> aggregationTables = initDefaultTables(aggregatorName, aggregationDurations, processedMetaStreamEvent.getOutputStreamDefinition(), siddhiAppRuntimeBuilder, aggregationDefinition.getAnnotations(), groupByVariableList, isProcessingOnExternalTime, isDistributed);
        Map<TimePeriod.Duration, Executor> incrementalExecutorMap = buildIncrementalExecutors(processedMetaStreamEvent, processExpressionExecutorsMap, groupByKeyGeneratorMap, aggregationDurations, aggregationTables, siddhiQueryContext, aggregatorName, shouldUpdateTimestamp, timeZone, isPersistedAggregation, incomingOutputStreamDefinition, isDistributed, shardId, isProcessingOnExternalTime, aggregationDefinition, configManager, groupByVariableList, isReadOnly);
        isOptimisedLookup = isOptimisedLookup && aggregationTables.get(aggregationDurations.get(0)) instanceof QueryableProcessor;
        List<String> groupByVariablesList = groupByVariableList.stream().map(Variable::getAttributeName).collect(Collectors.toList());
        List<OutputAttribute> defaultSelectorList = new ArrayList<>();
        if (isOptimisedLookup) {
            defaultSelectorList = incomingOutputStreamDefinition.getAttributeList().stream().map((attribute) -> new OutputAttribute(new Variable(attribute.getName()))).collect(Collectors.toList());
        }
        IncrementalDataPurger incrementalDataPurger = new IncrementalDataPurger();
        incrementalDataPurger.init(aggregationDefinition, new StreamEventFactory(processedMetaStreamEvent), aggregationTables, isProcessingOnExternalTime, siddhiQueryContext, aggregationDurations, timeZone, windowMap, aggregationMap);
        // Recreate in-memory data from tables
        IncrementalExecutorsInitialiser incrementalExecutorsInitialiser = new IncrementalExecutorsInitialiser(aggregationDurations, aggregationTables, incrementalExecutorMap, isDistributed, shardId, siddhiAppContext, processedMetaStreamEvent, tableMap, windowMap, aggregationMap, timeZone, isReadOnly, isPersistedAggregation);
        IncrementalExecutor rootIncrementalExecutor = (IncrementalExecutor) incrementalExecutorMap.get(aggregationDurations.get(0));
        rootIncrementalExecutor.setScheduler(scheduler);
        // Connect entry valve to root incremental executor
        entryValveExecutor.setNextExecutor(rootIncrementalExecutor);
        QueryParserHelper.initStreamRuntime(streamRuntime, incomingMetaStreamEvent, lockWrapper, aggregatorName);
        LatencyTracker latencyTrackerFind = null;
        LatencyTracker latencyTrackerInsert = null;
        ThroughputTracker throughputTrackerFind = null;
        ThroughputTracker throughputTrackerInsert = null;
        if (siddhiAppContext.getStatisticsManager() != null) {
            latencyTrackerFind = QueryParserHelper.createLatencyTracker(siddhiAppContext, aggregationDefinition.getId(), METRIC_INFIX_AGGREGATIONS, METRIC_TYPE_FIND);
            latencyTrackerInsert = QueryParserHelper.createLatencyTracker(siddhiAppContext, aggregationDefinition.getId(), METRIC_INFIX_AGGREGATIONS, METRIC_TYPE_INSERT);
            throughputTrackerFind = QueryParserHelper.createThroughputTracker(siddhiAppContext, aggregationDefinition.getId(), METRIC_INFIX_AGGREGATIONS, METRIC_TYPE_FIND);
            throughputTrackerInsert = QueryParserHelper.createThroughputTracker(siddhiAppContext, aggregationDefinition.getId(), METRIC_INFIX_AGGREGATIONS, METRIC_TYPE_INSERT);
        }
        AggregationRuntime aggregationRuntime = new AggregationRuntime(aggregationDefinition, isProcessingOnExternalTime, isDistributed, aggregationDurations, incrementalExecutorMap, aggregationTables, outputExpressionExecutors, processExpressionExecutorsMapForFind, shouldUpdateTimestamp, groupByKeyGeneratorMapForReading, isOptimisedLookup, defaultSelectorList, groupByVariablesList, isLatestEventColAdded, baseAggregatorBeginIndex, finalBaseExpressions, incrementalDataPurger, incrementalExecutorsInitialiser, ((SingleStreamRuntime) streamRuntime), processedMetaStreamEvent, latencyTrackerFind, throughputTrackerFind, timeZone);
        streamRuntime.setCommonProcessor(new IncrementalAggregationProcessor(aggregationRuntime, incomingExpressionExecutors, processedMetaStreamEvent, latencyTrackerInsert, throughputTrackerInsert, siddhiAppContext));
        return aggregationRuntime;
    } catch (Throwable t) {
        ExceptionUtil.populateQueryContext(t, aggregationDefinition, siddhiAppContext);
        throw t;
    }
}