@grepr/cli 1.14.13-15e1f25 → 1.15.0-48d68ee

package/build/dist/lib/job-graph-transformer.js

@@ -1,767 +0,0 @@
-/**
- * Job graph transformation utilities for converting production jobs to test configurations.
- *
- * This module provides functions to transform a production job definition into a test job
- * by replacing sources, sinks, and adding test-specific tagging. The transformations are
- * purely functional - they do not make any API calls or create resources.
- *
- * Key transformations:
- * - Source replacement: Replace vendor sources with dataset queries or keep originals with limits
- * - Sink replacement: Replace vendor sinks with synchronous test sinks or test dataset sinks
- * - Test tagging: Add test run ID and sink name tags to track test data
- *
- * The transformed job can then be executed using the standard job:create command.
- */
-import { DEFAULT_INPUT, DEFAULT_OUTPUT, JobExecution, JobProcessing, Vertex, DEFAULT_LIMIT } from '../types.js';
-import { LogsIcebergTableSourceType, LogsSynchronousSinkType, LogsIcebergTableSinkType, LogTransformActionType, DatadogQueryPredicateType, TagActionModification, TagActionType, LogsValuesSourceType, JsonLogProcessorType, GrokParserType, LogAttributesRemapperType, LogsFilterType, LogReducerType, LogsEventSamplerType, GreprUploadedLogFileSourceType, isSinkType, } from '../openapi/openApiTypes.js';
-import { canLimit, parseEdge, generateUUID, } from './job-graph-utils.js';
-import { RAW_JSON_PROCESSOR, RAW_PRE_EXCEPTIONS_FILTER, RAW_PRE_PARSER_FILTER, RAW_PRE_WAREHOUSE_FILTER, } from './job-graph-log-pipeline-constants.js';
-import fs from 'fs-extra';
-/**
- * Default `maxBurstLimit` for a draft logs-event-sampler. Matches the UI's
- * template-draft fallback (`logSampler?.maxBurstLimit ?? 1000`) so CLI raw-job
- * drafts sample at the same rate as the UI's template drafts. `maxAllowedRate`
- * has no CLI default: when unset we omit it and let the server default apply,
- * exactly as the UI does.
- */
-export const DEFAULT_DRAFT_SAMPLER_BURST = 1000;
-/**
- * Transforms a production job configuration into a test job configuration.
- *
- * This is the main entry point for job transformation. It applies the following changes:
- * 1. Renames the job (appends "_test" or uses custom name)
- * 2. Updates execution and processing types if specified
- * 3. Adds test run ID tag to job tags
- * 4. Transforms sources (replace with dataset query, sample data, or add limits)
- * 5. Transforms sinks (replace with sync sink or test dataset sinks)
- * 6. Adds test tagging operations before sinks
- *
- * The transformation is purely functional and does not make any API calls.
- *
- * @param originalJob - The production job to transform
- * @param options - Transformation options
- * @returns A new job configuration suitable for testing
- * @throws Error if job graph is missing or invalid
- *
- * @example
- * const testJob = transformJobToTest(prodJob, {
- *   execution: 'SYNCHRONOUS',
- *   processing: 'BATCH',
- *   datasetId: 'my_dataset',
- *   start: '2025-01-01T00:00:00Z',
- *   end: '2025-01-01T01:00:00Z',
- *   limitRecords: 100
- * });
- */
-export function transformJobToTest(originalJob, options) {
-    const job = { ...originalJob };
-    // Determine target execution and processing types (use options or keep original)
-    const targetExecution = options.execution || job.execution;
-    const targetProcessing = options.processing || job.processing;
-    // Create base transformed job with updated metadata
-    const transformedJob = {
-        ...job,
-        name: options.testName || `${job.name}_test`,
-        execution: targetExecution,
-        processing: targetProcessing,
-        tags: {
-            ...(job.tags || {}),
-            'grepr.test_run_id': options.testTag || generateUUID()
-        }
-    };
-    if (!transformedJob.jobGraph) {
-        throw new Error('Job graph is required');
-    }
-    const parsedGraph = parseGraph(transformedJob.jobGraph);
-    // Step 1: Add test tagging operations on all edges BEFORE transforming sources/sinks
-    // This captures the original source/sink names in the edge tags
-    addTestTagging(parsedGraph, transformedJob.tags?.['grepr.test_run_id'] || generateUUID());
-    // Step 2: Transform sources (replace with dataset, sample data, or add limits)
-    transformSources(parsedGraph, targetProcessing, options);
-    // Step 3: Transform sinks (replace with sync sink or test dataset sinks)
-    transformSinks(parsedGraph, targetExecution, options);
-    // Convert the parsed/modified graph back to SchemaGreprJobGraph format
-    const transformedVertices = [];
-    const transformedEdges = [];
-    for (const vertex of parsedGraph.values()) {
-        transformedVertices.push(vertex.operation);
-        for (const [outputPort, nextIos] of vertex.next) {
-            for (const nextIo of nextIos) {
-                const edgeStr = `${vertex.name}:${outputPort} -> ${nextIo.vertex.name}:${nextIo.name}`;
-                transformedEdges.push(edgeStr);
-            }
-        }
-    }
-    transformedJob.jobGraph = {
-        vertices: transformedVertices,
-        edges: transformedEdges
-    };
-    return transformedJob;
-}
-/**
- * Parses the job graph into a vertex-based representation, returning all vertices.
- */
-function parseGraph(jobGraph) {
-    const verticesMap = new Map();
-    // Create Vertex objects for each operation
-    for (const operation of jobGraph.vertices) {
-        verticesMap.set(operation.name, new Vertex(operation));
-    }
-    // Link vertices based on edges
-    for (const edge of jobGraph.edges) {
-        const { sourceVertex: sourceVertexName, targetVertex: targetVertexName, sourcePort, targetPort } = parseEdge(edge);
-        const sourceVertex = verticesMap.get(sourceVertexName);
-        const targetVertex = verticesMap.get(targetVertexName);
-        if (sourceVertex && targetVertex) {
-            sourceVertex.addNext(sourcePort, { name: targetPort, vertex: targetVertex });
-        }
-    }
-    return verticesMap;
-}
-/**
- * Transforms job graph sources based on the specified options.
- *
- * Three transformation strategies are supported:
- * 1. Sample data file: Replace all sources with a LogsValuesSource (not yet implemented)
- * 2. Dataset query: Replace all sources with a single IcebergTableSource
- * 3. In-place: Keep original sources but add record limits for batch processing
- *
- * @param jobGraph - The job graph to transform
- * @param processing - Target processing type
- * @param options - Source transformation options
- * @returns Transformed job graph with updated sources
- */
-function transformSources(jobGraph, processing, options) {
-    if (options.sampleDataFile) {
-        transformSourcesToSampleData(jobGraph, options.sampleDataFile);
-    }
-    else if (options.datasetId) {
-        transformSourcesToDatasetQuery(jobGraph, options.datasetId, options.query, options.start, options.end, processing, options.limitRecords);
-    }
-    else {
-        transformSourcesInPlace(jobGraph, processing, options.limitRecords);
-    }
-}
-/**
- * Transforms sources to use sample data from a file.
- *
- * This replaces all existing sources with a single LogsValuesSource that
- * emits the log events from the sample data file. The file should contain
- * a JSON array of log events matching the LogEvent schema.
- *
- * All edges from the old sources are rewired to connect from the new source
- * to the first downstream operation in the pipeline.
- *
- * @param jobGraph - The job graph to transform
- * @param sampleDataFile - Path to sample data file (JSON array of log events)
- * @throws Error if the file cannot be read or parsed
- */
-function transformSourcesToSampleData(jobGraph, sampleDataFile) {
-    if (!fs.existsSync(sampleDataFile)) {
-        throw new Error(`Sample data file not found: ${sampleDataFile}`);
-    }
-    // Read and parse the sample data file
-    let sampleLogs;
-    try {
-        const fileContent = fs.readFileSync(sampleDataFile, 'utf-8');
-        sampleLogs = JSON.parse(fileContent);
-        if (!Array.isArray(sampleLogs)) {
-            throw new Error('Sample data must be a JSON array of log events');
-        }
-    }
-    catch (error) {
-        if (error.message.includes('Sample data')) {
-            throw error;
-        }
-        throw new Error(`Failed to parse sample data file: ${error.message}`);
-    }
-    // Create a new LogsValuesSource with the sample data
-    const newSource = {
-        type: LogsValuesSourceType.logs_values_source,
-        name: 'test_sample_source',
-        values: sampleLogs,
-        indefinite: false
-    };
-    const newSourceVertex = new Vertex(newSource);
-    // Remove all old sources and connect them to the new source
-    for (const vertex of Array.from(jobGraph.values())) {
-        if (isSource(vertex)) {
-            // Rewire edges from old source's next vertices to new source
-            for (const [port, nextIos] of vertex.next) {
-                for (const nextIo of nextIos) {
-                    vertex.removeNext(port, nextIo);
-                    newSourceVertex.addNext(DEFAULT_OUTPUT, nextIo);
-                }
-            }
-            jobGraph.delete(vertex.name);
-        }
-    }
-    // Add the new source vertex to the graph
-    jobGraph.set(newSourceVertex.name, newSourceVertex);
-}
-/**
- * Transforms sources to use an Iceberg dataset as the data source.
- *
- * This replaces all existing sources with a single IcebergTableSource that:
- * - Reads from the specified dataset
- * - Optionally filters data using a query
- * - Reads data within the specified time range
- * - Limits records for batch processing if specified
- *
- * All edges from the old sources are rewired to connect from the new source
- * to the first downstream operation in the pipeline.
- *
- * @param jobGraph - The job graph to transform
- * @param datasetId - ID of the dataset to read from
- * @param query - Optional query to filter dataset data
- * @param start - Start time for data query
- * @param end - End time for data query
- * @param processing - Processing type (BATCH or STREAMING)
- * @param limitRecords - Optional record limit for batch processing
- * @returns Transformed job graph with dataset source
- */
-function transformSourcesToDatasetQuery(jobGraph, datasetId, query, start, end, processing, limitRecords) {
-    const newSource = {
-        type: LogsIcebergTableSourceType.logs_iceberg_table_source,
-        name: 'test_dataset_source',
-        datasetId,
-        query: {
-            type: DatadogQueryPredicateType.datadog_query,
-            query: query || ''
-        },
-        start: start || '',
-        end: end || '',
-        limit: limitRecords ?? DEFAULT_LIMIT
-    };
-    const newSourceVertex = new Vertex(newSource);
-    // Remove all old sources and connect them to the new source
-    for (const vertex of Array.from(jobGraph.values())) {
-        if (isSource(vertex)) {
-            // This is a source vertex since it has no inputs.
-            // Rewire edges from old source's next vertices to new source
-            for (const [port, nextIos] of vertex.next) {
-                for (const nextIo of nextIos) {
-                    vertex.removeNext(port, nextIo);
-                    newSourceVertex.addNext(DEFAULT_OUTPUT, nextIo);
-                }
-            }
-            jobGraph.delete(vertex.name);
-        }
-    }
-    // Add the new source vertex to the graph
-    jobGraph.set(newSourceVertex.name, newSourceVertex);
-}
-/**
- * Transforms sources in place by adding record limits for batch processing.
- *
- * This keeps the original sources but adds a 'limit' parameter to sources
- * that support it (e.g., IcebergTableSource). This is useful when you want
- * to test with the same data sources but limit the amount of data processed.
- *
- * @param jobGraph - The job graph to transform
- * @param processing - Processing type
- * @param limitRecords - Optional record limit to apply
- * @returns Transformed job graph with limited sources
- */
-function transformSourcesInPlace(jobGraph, processing, limitRecords) {
-    for (const vertex of jobGraph.values()) {
-        if (!isSource(vertex)) {
-            continue;
-        }
-        if (processing === JobProcessing.BATCH && limitRecords && canLimit(vertex.operation)) {
-            vertex.operation = {
-                ...vertex.operation,
-                limit: limitRecords
-            };
-        }
-    }
-}
-/**
- * Transforms job graph sinks based on the target execution type.
- *
- * Two transformation strategies:
- * 1. SYNCHRONOUS: Replace all sinks with a single synchronous sink that returns results
- * 2. ASYNCHRONOUS: Replace vendor sinks with test dataset sinks (if testDataset provided)
- *
- * @param jobGraph - The job graph to transform
- * @param execution - Target execution type
- * @param options - Sink transformation options
- * @returns Transformed job graph with updated sinks
- */
-function transformSinks(jobGraph, execution, options) {
-    if (execution === JobExecution.SYNCHRONOUS) {
-        return transformSinksToSynchronous(jobGraph);
-    }
-    else {
-        if (!options.testDataset) {
-            throw new Error('Test dataset ID is required for ASYNCHRONOUS test job');
-        }
-        return transformSinksToAsynchronous(jobGraph, options.testDataset);
-    }
-}
-/**
- * Transforms all sinks to a single synchronous sink.
- *
- * For synchronous testing, we replace all sink operations with a single
- * LogsSyncSink that returns the processed data in the API response.
- * This allows immediate verification of results.
- *
- * All operations that were feeding into the original sinks are rewired
- * to feed into the new synchronous sink.
- *
- * @param jobGraph - The job graph to transform
- */
-function transformSinksToSynchronous(jobGraph) {
-    // Create a single synchronous sink to replace them all
-    const syncSink = {
-        type: LogsSynchronousSinkType.logs_sync_sink,
-        name: 'test_synchronous_sink'
-    };
-    const sinkVertex = new Vertex(syncSink);
-    wireNewSink(jobGraph, sinkVertex);
-}
-/**
- * Transforms sinks for asynchronous testing with a test dataset.
- *
- * If a test dataset ID is provided, this replaces vendor sinks, raw data sinks,
- * and event dedup sinks with Iceberg table sinks that write to the test dataset.
- *
- * The original sink names are preserved to maintain the graph structure.
- * Test tagging (added separately) allows distinguishing which sink the data came from.
- *
- * @param jobGraph - The job graph to transform
- * @param testDatasetId - Optional test dataset ID for output
- */
-function transformSinksToAsynchronous(jobGraph, testDatasetId) {
-    // For each sink (no outputs), replace with test dataset sink.
-    const sinkOp = {
-        type: LogsIcebergTableSinkType.logs_iceberg_table_sink,
-        name: 'test_async_sink',
-        datasetId: testDatasetId
-    };
-    const sinkVertex = new Vertex(sinkOp);
-    wireNewSink(jobGraph, sinkVertex);
-}
-function wireNewSink(jobGraph, sinkVertex) {
-    // Find all vertices that feed into sinks and rewire them to the new sync sink
-    for (const vertex of Array.from(jobGraph.values())) {
-        if (isSink(vertex)) {
-            // This is a sink vertex since it has no outputs.
-            // Rewire edges from old sink's prev vertices to new sync sink
-            for (const [port, prevIos] of vertex.prev) {
-                for (const prevIo of prevIos) {
-                    vertex.removePrev(port, prevIo);
-                    prevIo.vertex.addNext(prevIo.name, { name: DEFAULT_INPUT, vertex: sinkVertex });
-                }
-            }
-            jobGraph.delete(vertex.name);
-        }
-    }
-    // Add the new sink vertex to the graph
-    jobGraph.set(sinkVertex.name, sinkVertex);
-}
-/**
- * Adds test tagging operations on all edges, with the id of the
- * original edge and the test run ID.
- *
- * @param jobGraph - The job graph to transform
- * @param testRunId - Unique test run identifier
- */
-function addTestTagging(jobGraph, testRunId) {
-    for (const vertex of Array.from(jobGraph.values())) {
-        for (const [output, nextIos] of Array.from(vertex.next)) {
-            for (const nextIo of nextIos) {
-                // insert a tagging operation before this io
-                const edge = `${vertex.name}_${output}_${nextIo.vertex.name}_${nextIo.name}`;
-                const tagOpName = `${edge}_test_tag`;
-                const tagOp = createTestTagOperation(tagOpName, testRunId, edge);
-                const tagVertex = new Vertex(tagOp);
-                // Rewire edges: vertex -> tagOp -> nextIo.vertex
-                vertex.removeNext(output, nextIo);
-                vertex.addNext(output, { name: DEFAULT_INPUT, vertex: tagVertex });
-                tagVertex.addNext(DEFAULT_OUTPUT, nextIo);
-                // Add the new tagging vertex to the graph
-                jobGraph.set(tagVertex.name, tagVertex);
-            }
-        }
-    }
-}
-/**
- * Creates a log transform operation that adds test tags.
- *
- * The operation adds two tags:
- * - grepr.test_run_id: Identifies which test run this data belongs to
- * - grepr.edge: Identifies edge this data was going through
- *
- * These tags are added using the ADD modification, which appends to existing tag arrays.
- *
- * @param name - Name for the tagging operation
- * @param testRunId - Unique test run identifier
- * @param edge - Name of the edge
- * @returns A LogTransformAction operation that adds the tags
- */
-function createTestTagOperation(name, testRunId, edge) {
-    return {
-        type: LogTransformActionType.log_transform,
-        name,
-        transforms: [
-            {
-                order: 0,
-                type: TagActionType.tag_action,
-                modification: TagActionModification.ADD,
-                tagKey: 'grepr.test_run_id',
-                values: [testRunId]
-            },
-            {
-                order: 1,
-                type: TagActionType.tag_action,
-                modification: TagActionModification.ADD,
-                tagKey: 'grepr.edge',
-                values: [edge]
-            }
-        ]
-    };
-}
-const isSource = (vertex) => vertex.prev.size === 0;
-const isSink = (vertex) => vertex.next.size === 0;
-/**
- * Vertex types tapped by type (the ones `applyPatchToJobGraph` mutates), so
- * legacy graphs with noncanonical names still tap the edited vertex. Pass-through
- * stages are matched by name instead (see {@link TAP_STAGE_NAME_TYPES}) because
- * linearizing parallel per-sink `logs-filter` vertices would change draft semantics.
- * Tap-eligible vertices fork their output to the shared sync sink (tagged for demux);
- * sources keep streaming live and sinks collapse into that one shared sync sink.
- */
-const TAP_MUTABLE_TYPES = new Set([
-    GrokParserType.grok_parser,
-    LogAttributesRemapperType.log_attributes_remapper,
-    LogReducerType.log_reducer,
-]);
-const TAP_STAGE_NAME_TYPES = new Map([
-    [RAW_PRE_PARSER_FILTER, LogsFilterType.logs_filter],
-    [RAW_JSON_PROCESSOR, JsonLogProcessorType.json_log_processor],
-    [RAW_PRE_WAREHOUSE_FILTER, LogsFilterType.logs_filter],
-    [RAW_PRE_EXCEPTIONS_FILTER, LogsFilterType.logs_filter],
-]);
-function isTapEligible(vertex) {
-    if (TAP_MUTABLE_TYPES.has(vertex.type)) {
-        return true;
-    }
-    return TAP_STAGE_NAME_TYPES.get(vertex.name) === vertex.type;
-}
-/**
- * Tap-eligible vertices in topological order (Kahn's algorithm). Uses edges,
- * not vertex-array order, since the array order isn't guaranteed topological.
- */
-function collectTapEligibleInOrder(jobGraph) {
-    const vertices = jobGraph.vertices ?? [];
-    const edges = jobGraph.edges ?? [];
-    const verticesByName = new Map();
-    for (const v of vertices) {
-        if (v.name)
-            verticesByName.set(v.name, v);
-    }
-    const downstream = new Map();
-    const inboundCount = new Map();
-    for (const v of vertices) {
-        if (v.name) {
-            downstream.set(v.name, []);
-            inboundCount.set(v.name, 0);
-        }
-    }
-    for (const edge of edges) {
-        const { sourceVertex, targetVertex } = parseEdge(edge);
-        downstream.get(sourceVertex)?.push(targetVertex);
-        inboundCount.set(targetVertex, (inboundCount.get(targetVertex) ?? 0) + 1);
-    }
-    const queue = [];
-    for (const [name, count] of inboundCount) {
-        if (count === 0)
-            queue.push(name);
-    }
-    const ordered = [];
-    const processed = new Set();
-    while (queue.length > 0) {
-        const name = queue.shift();
-        processed.add(name);
-        const vertex = verticesByName.get(name);
-        if (vertex && isTapEligible(vertex)) {
-            ordered.push(vertex);
-        }
-        for (const child of downstream.get(name) ?? []) {
-            const remaining = (inboundCount.get(child) ?? 0) - 1;
-            inboundCount.set(child, remaining);
-            if (remaining === 0)
-                queue.push(child);
-        }
-    }
-    if (processed.size < verticesByName.size) {
-        throw new Error(`Unsupported raw job graph shape: cycle detected while building tapped draft ` +
-            `(${processed.size}/${verticesByName.size} vertices were topologically reachable).`);
-    }
-    return ordered;
-}
-/**
- * Build a source-preserving draft from a job-graph pipeline: keep the proposed
- * source vertices streaming live (so add/remove-source edits are exercised),
- * interpose a logs-event-sampler after each live source to bound per-source
- * volume, strip production sinks, and fan output into a shared logs-sync-sink.
- * Mirrors the UI's template draft mode for raw (non-templated) jobs. Verifies
- * the graph runs, not external sink delivery.
- */
-export function transformJobGraphToSourcePreservingDraft(originalJob, options = {}) {
-    if (!originalJob.jobGraph) {
-        throw new Error('Job graph is required');
-    }
-    const graph = structuredClone(originalJob.jobGraph);
-    const existingNames = new Set();
-    for (const vertex of graph.vertices) {
-        if (vertex.name)
-            existingNames.add(vertex.name);
-    }
-    const targetProcessing = options.processing ?? originalJob.processing ?? JobProcessing.STREAMING;
-    const sinkNames = new Set(graph.vertices
-        .filter(isProductionSinkOperation)
-        .map(vertex => vertex.name)
-        .filter((name) => typeof name === 'string' && name.length > 0));
-    const originalParsedEdges = parseGraphEdges(graph.edges ?? []);
-    const sinkInboundEdges = originalParsedEdges.filter(edge => sinkNames.has(edge.targetVertex));
-    const retainedVertices = graph.vertices.filter(vertex => !sinkNames.has(vertex.name));
-    const retainedEdges = (graph.edges ?? []).filter(edge => !edgeTouchesAny(edge, sinkNames));
-    // Interpose a logs-event-sampler after each live source (mutates the retained
-    // vertices/edges in place) so the live stream is rate-limited, mirroring the
-    // UI's template draft mode. Done before the tap/sink wiring below so the
-    // sampler vertices are part of the graph the rest of this function reasons over.
-    insertSourceSamplers(retainedVertices, retainedEdges, existingNames, options.sampler);
-    const syncSinkName = uniqueDraftName(existingNames, 'draft_source_preserving_sink');
-    existingNames.add(syncSinkName);
-    const syncSink = makeOperation(LogsSynchronousSinkType.logs_sync_sink, syncSinkName);
-    const tapVertices = collectTapEligibleInOrder({ vertices: retainedVertices, edges: retainedEdges });
-    const tapEligibleNames = new Set(tapVertices
-        .map(vertex => vertex.name)
-        .filter((name) => typeof name === 'string' && name.length > 0));
-    const draftEdges = [...retainedEdges];
-    for (const inbound of sinkInboundEdges) {
-        // A sink chained to another sink was itself stripped, so it cannot source a
-        // draft edge. Its records are already observed one hop upstream.
-        if (sinkNames.has(inbound.sourceVertex))
-            continue;
-        // Tap-eligible sink-predecessors reach the sync sink through their tagger
-        // below; adding a direct edge here too would deliver their records twice
-        // (once untagged, once tagged), inflating counts the per-stage demux can't
-        // attribute. Route them only through the tagger.
-        if (tapEligibleNames.has(inbound.sourceVertex))
-            continue;
-        addDraftEdgeIfMissing(draftEdges, inbound.sourceVertex, inbound.sourcePort, syncSinkName, DEFAULT_INPUT);
-    }
-    // Fallback: no production sinks had recorded inbound edges; wire all terminal vertices to the sync sink.
-    if (sinkInboundEdges.length === 0) {
-        for (const terminal of terminalVertices(retainedVertices, retainedEdges)) {
-            if (tapEligibleNames.has(terminal.name))
-                continue; // same double-delivery reasoning as above
-            addDraftEdgeIfMissing(draftEdges, terminal.name, DEFAULT_OUTPUT, syncSinkName, DEFAULT_INPUT);
-        }
-    }
-    const taggerVertices = [];
-    for (const tapVertex of tapVertices) {
-        if (!tapVertex.name)
-            continue;
-        const taggerName = uniqueDraftName(existingNames, `tap_${tapVertex.name}`);
-        existingNames.add(taggerName);
-        const tagger = makeSinkSourceTagger(taggerName, tapVertex.name);
-        taggerVertices.push(tagger);
-        addDraftEdgeIfMissing(draftEdges, tapVertex.name, DEFAULT_OUTPUT, taggerName, DEFAULT_INPUT);
-        addDraftEdgeIfMissing(draftEdges, taggerName, DEFAULT_OUTPUT, syncSinkName, DEFAULT_INPUT);
-    }
-    return {
-        ...originalJob,
-        name: options.testName ?? `${originalJob.name}_draft`,
-        execution: JobExecution.SYNCHRONOUS,
-        processing: targetProcessing,
-        tags: {
-            ...(originalJob.tags ?? {}),
-        },
-        jobGraph: {
-            vertices: [...retainedVertices, ...taggerVertices, syncSink],
-            edges: draftEdges,
-        },
-    };
-}
-function parseGraphEdges(edges) {
-    return edges.map(edge => parseEdge(edge));
-}
-function edgeTouchesAny(edge, names) {
-    const parsed = parseEdge(edge);
-    return names.has(parsed.sourceVertex) || names.has(parsed.targetVertex);
-}
-function isProductionSinkOperation(operation) {
-    return typeof operation.type === 'string' && isSinkType(operation.type);
-}
-function makeOperation(type, name, extra = {}) {
-    return { type, name, ...extra };
-}
-/** A `log-transform` vertex that tags records with `sink-source=<sourceName>` for per-stage demux. */
-function makeSinkSourceTagger(taggerName, sourceName) {
-    return makeOperation(LogTransformActionType.log_transform, taggerName, {
-        transforms: [
-            { type: TagActionType.tag_action, modification: TagActionModification.ADD, tagKey: 'sink-source', values: [sourceName], order: 0 },
-        ],
-    });
-}
-/** A `logs-event-sampler` vertex tuned for a draft, mirroring the UI's template draft sampler. */
-function makeEventSampler(name, sampler) {
-    const extra = {
-        maxBurstLimit: sampler?.maxBurstLimit ?? DEFAULT_DRAFT_SAMPLER_BURST,
-    };
-    // Omit maxAllowedRate unless set so the server's sampler default applies, exactly as the UI does.
-    if (sampler?.maxAllowedRate !== undefined) {
-        extra.maxAllowedRate = sampler.maxAllowedRate;
-    }
-    return makeOperation(LogsEventSamplerType.logs_event_sampler, name, extra);
-}
-/**
- * Interpose a logs-event-sampler between each live source and its downstream
- * targets, rewiring `source -> target` into `source -> sampler -> target`.
- * Mutates `vertices` (appends samplers) and `edges` (rewires) in place.
- *
- * A source is any vertex with no inbound edge. Uploaded-log-file sources are
- * skipped, matching the UI, which doesn't rate-limit uploaded-file drafts.
- * Isolated sources (no outbound edge) get no sampler — there's nothing to bound.
- */
-function insertSourceSamplers(vertices, edges, existingNames, sampler) {
-    const targets = new Set(parseGraphEdges(edges).map(edge => edge.targetVertex));
-    const sources = vertices.filter((vertex) => typeof vertex.name === 'string' &&
-        vertex.name.length > 0 &&
-        !targets.has(vertex.name) &&
-        vertex.type !== GreprUploadedLogFileSourceType.grepr_uploaded_log_file_source);
-    for (const source of sources) {
-        const outbound = parseGraphEdges(edges).filter(edge => edge.sourceVertex === source.name);
-        if (outbound.length === 0)
-            continue;
-        const samplerName = uniqueDraftName(existingNames, `${source.name}_draft_sampler`);
-        existingNames.add(samplerName);
-        vertices.push(makeEventSampler(samplerName, sampler));
-        // Drop the source's outbound edges, then relink them through the sampler.
-        for (let i = edges.length - 1; i >= 0; i--) {
-            if (parseEdge(edges[i]).sourceVertex === source.name)
-                edges.splice(i, 1);
-        }
-        edges.push(formatDraftEdge(source.name, outbound[0]?.sourcePort ?? DEFAULT_OUTPUT, samplerName, DEFAULT_INPUT));
-        for (const edge of outbound) {
-            edges.push(formatDraftEdge(samplerName, DEFAULT_OUTPUT, edge.targetVertex, edge.targetPort));
-        }
-    }
-}
-function terminalVertices(vertices, edges) {
-    const sources = new Set();
-    const targets = new Set();
-    for (const edge of parseGraphEdges(edges)) {
-        sources.add(edge.sourceVertex);
-        targets.add(edge.targetVertex);
-    }
-    return vertices.filter(vertex => typeof vertex.name === 'string' &&
-        vertex.name.length > 0 &&
-        !sources.has(vertex.name) &&
-        targets.has(vertex.name));
-}
-function uniqueDraftName(existing, base) {
-    if (!existing.has(base))
-        return base;
-    let i = 1;
-    while (existing.has(`${base}_${i}`)) {
-        i += 1;
-    }
-    return `${base}_${i}`;
-}
-function formatDraftEdge(sourceVertex, sourcePort, targetVertex, targetPort) {
-    const source = sourcePort === DEFAULT_OUTPUT ? sourceVertex : `${sourceVertex}:${sourcePort}`;
-    const target = targetPort === DEFAULT_INPUT ? targetVertex : `${targetVertex}:${targetPort}`;
-    return `${source} -> ${target}`;
-}
-function addDraftEdgeIfMissing(edges, sourceVertex, sourcePort, targetVertex, targetPort) {
-    const exists = parseGraphEdges(edges).some(edge => edge.sourceVertex === sourceVertex &&
-        edge.sourcePort === sourcePort &&
-        edge.targetVertex === targetVertex &&
-        edge.targetPort === targetPort);
-    if (!exists) {
-        edges.push(formatDraftEdge(sourceVertex, sourcePort, targetVertex, targetPort));
-    }
-}
-/**
- * Displays a summary of the transformations applied to a job.
- *
- * This shows a before/after comparison including:
- * - Job name changes
- * - Execution type changes
- * - Processing type changes
- * - operation and edge changes
- * - Test run ID
- *
- * The output is formatted for console display.
- *
- * @param originalJob - The original production job
- * @param transformedJob - The transformed test job
- */
-export function showDiff(originalJob, transformedJob) {
-    console.log('\n=== TRANSFORMATION SUMMARY ===\n');
-    console.log(`Name: ${originalJob.name} → ${transformedJob.name}`);
-    console.log(`Execution: ${originalJob.execution} → ${transformedJob.execution}`);
-    console.log(`Processing: ${originalJob.processing} → ${transformedJob.processing}`);
-    if (originalJob.jobGraph && transformedJob.jobGraph) {
-        // Show added/removed operations and added/removed edges
-        // For each operation in original, check if in transformed, and if changed
-        const originalOps = new Set(originalJob.jobGraph.vertices.map(v => v.name));
-        const transformedOps = new Set(transformedJob.jobGraph.vertices.map(v => v.name));
-        const originalOpsMap = new Map(originalJob.jobGraph.vertices.map(v => [v.name, v]));
-        const transformedOpsMap = new Map(transformedJob.jobGraph.vertices.map(v => [v.name, v]));
-        const addedOps = Array.from(transformedOps).filter(op => !originalOps.has(op));
-        const removedOps = Array.from(originalOps).filter(op => !transformedOps.has(op));
-        const changedOps = Array.from(originalOps).filter(op => transformedOps.has(op) &&
-            JSON.stringify(originalOpsMap.get(op)) !== JSON.stringify(transformedOpsMap.get(op)));
-        if (addedOps.length > 0) {
-            console.log(`Added Operations:`);
-            for (const opName of addedOps) {
-                const op = transformedOpsMap.get(opName);
-                console.log(`${JSON.stringify(op, null, 2)}`);
-            }
-        }
-        if (removedOps.length > 0) {
-            console.log(`Removed Operations:`);
-            for (const opName of removedOps) {
-                const op = originalOpsMap.get(opName);
-                console.log(`${JSON.stringify(op, null, 2)}`);
-            }
-        }
-        if (changedOps.length > 0) {
-            console.log(`  Changed Operations:`);
-            for (const opName of changedOps) {
-                const originalOp = originalOpsMap.get(opName);
-                const transformedOp = transformedOpsMap.get(opName);
-                console.log(`- Original: ${JSON.stringify(originalOp, null, 2)}`);
-                console.log(`+ Transformed: ${JSON.stringify(transformedOp, null, 2)}`);
-            }
-        }
-        // Show edge changes
-        const originalEdges = new Set(originalJob.jobGraph.edges);
-        const transformedEdges = new Set(transformedJob.jobGraph.edges);
-        const addedEdges = Array.from(transformedEdges).filter(edge => !originalEdges.has(edge));
-        const removedEdges = Array.from(originalEdges).filter(edge => !transformedEdges.has(edge));
-        if (addedEdges.length > 0) {
-            console.log(`\n  Added Edges:`);
-            for (const edge of addedEdges) {
-                console.log(`+ ${edge}`);
-            }
-        }
-        if (removedEdges.length > 0) {
-            console.log(`\n  Removed Edges:`);
-            for (const edge of removedEdges) {
-                console.log(`- ${edge}`);
-            }
-        }
-        console.log('\nOperations:');
-    }
-    console.log('\nTest Tag:');
-    console.log(`  ${transformedJob.tags?.['grepr.test_run_id']}`);
-    console.log('\n=== TRANSFORMED JOB JSON ===\n');
-}
-//# sourceMappingURL=job-graph-transformer.js.map