jaketf commented on a change in pull request #11339: [BEAM-9468] [WIP] Fhir io URL: https://github.com/apache/beam/pull/11339#discussion_r405824473
########## File path: sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/healthcare/FhirIO.java ########## @@ -0,0 +1,662 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.beam.sdk.io.gcp.healthcare; + +import com.fasterxml.jackson.databind.ObjectMapper; +import com.google.api.services.healthcare.v1beta1.model.HttpBody; +import com.google.api.services.healthcare.v1beta1.model.Operation; +import com.google.auto.value.AutoValue; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.channels.WritableByteChannel; +import java.nio.charset.StandardCharsets; +import java.util.Collection; +import java.util.Collections; +import java.util.Map; +import java.util.Optional; +import java.util.UUID; +import javax.annotation.Nullable; +import org.apache.beam.sdk.Pipeline; +import org.apache.beam.sdk.annotations.Experimental; +import org.apache.beam.sdk.coders.StringUtf8Coder; +import org.apache.beam.sdk.io.FileSystems; +import org.apache.beam.sdk.io.fs.ResourceId; +import org.apache.beam.sdk.io.gcp.healthcare.FhirIO.Import.ContentStructure; +import org.apache.beam.sdk.io.gcp.healthcare.FhirIO.Write.Result; +import org.apache.beam.sdk.io.gcp.pubsub.PubsubIO; +import org.apache.beam.sdk.metrics.Counter; +import org.apache.beam.sdk.metrics.Metrics; +import org.apache.beam.sdk.transforms.DoFn; +import org.apache.beam.sdk.transforms.PTransform; +import org.apache.beam.sdk.transforms.ParDo; +import org.apache.beam.sdk.util.Sleeper; +import org.apache.beam.sdk.values.PCollection; +import org.apache.beam.sdk.values.PCollectionTuple; +import org.apache.beam.sdk.values.PInput; +import org.apache.beam.sdk.values.POutput; +import org.apache.beam.sdk.values.PValue; +import org.apache.beam.sdk.values.TupleTag; +import org.apache.beam.sdk.values.TupleTagList; +import org.apache.beam.vendor.guava.v26_0_jre.com.google.common.base.Throwables; +import org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.ImmutableMap; +import org.codehaus.jackson.JsonProcessingException; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * {@link FhirIO} provides an API for reading and writing resources to <a + * href="https://cloud.google.com/healthcare/docs/concepts/fhir";>Google Cloud Healthcare Fhir API. + * </a> + * + * <p>Read + * + * <p>FHIR resources can be read with {@link FhirIO.Read} supports use cases where you have a + * ${@link PCollection} of message IDS. This is appropriate for reading the Fhir notifications from + * a Pub/Sub subscription with {@link PubsubIO#readStrings()} or in cases where you have a manually + * prepared list of messages that you need to process (e.g. in a text file read with {@link + * org.apache.beam.sdk.io.TextIO}) . + * + * <p>Fetch Resource contents from Fhir Store based on the {@link PCollection} of message ID strings + * {@link FhirIO.Read.Result} where one can call {@link Read.Result#getResources()} to retrieved a + * {@link PCollection} containing the successfully fetched {@link HttpBody}s and/or {@link + * FhirIO.Read.Result#getFailedReads()} to retrieve a {@link PCollection} of {@link + * HealthcareIOError} containing the resource ID that could not be fetched and the exception as a + * {@link HealthcareIOError}, this can be used to write to the dead letter storage system of your + * choosing. This error handling is mainly to transparently surface errors where the upstream {@link + * PCollection} contains IDs that are not valid or are not reachable due to permissions issues. + * + * <p>Write Resources can be written to FHIR with two different methods: Import or Execute Bundle. + * + * <p>Execute Bundle This is best for use cases where you are writing to a non-empty FHIR store with + * other clients or otherwise need referential integrity (e.g. A Streaming HL7v2 to FHIR ETL + * pipeline). + * + * @see <a + * href=>https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores.fhir/executeBundle></a> + * <p>Import This is best for use cases where you are populating an empty FHIR store with no + * other clients. It is faster than the execute bundles method but does not respect referential + * integrity and the resources are not written transactionally (e.g. a historicaly backfill on a + * new FHIR store) This requires each resource to contain a client provided ID. It is important + * that when using import you give the appropriate permissions to the Google Cloud Healthcare + * Service Agent + * @see <a + * href=>https://cloud.google.com/healthcare/docs/how-tos/permissions-healthcare-api-gcp-products#fhir_store_cloud_storage_permissions></a> + * @see <a + * href=>https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores/import></a> + * A {@link PCollection} of {@link HttpBody} can be ingested into an Fhir store using {@link + * FhirIO.Write#fhirStoresImport(String, String, String, ContentStructure)} This will return a + * {@link FhirIO.Write.Result} on which you can call {@link + * FhirIO.Write.Result#getFailedInsertsWithErr()} to retrieve a {@link PCollection} of {@link + * HealthcareIOError} containing the {@link HttpBody} that failed to be ingested and the + * exception. + */ +public class FhirIO { + + @Experimental + public static class Read extends PTransform<PCollection<String>, FhirIO.Read.Result> { + private static final Logger LOG = LoggerFactory.getLogger(Read.class); + + public Read() { + LOG.warn("FhirIO.Read is experimental and has not been tested."); + } + + public static class Result implements POutput, PInput { + private PCollection<HttpBody> resources; + + private PCollection<HealthcareIOError<String>> failedReads; + PCollectionTuple pct; + + public static FhirIO.Read.Result of(PCollectionTuple pct) throws IllegalArgumentException { + if (pct.getAll() + .keySet() + .containsAll((Collection<?>) TupleTagList.of(OUT).and(DEAD_LETTER))) { + return new FhirIO.Read.Result(pct); + } else { + throw new IllegalArgumentException( + "The PCollection tuple must have the FhirIO.Read.OUT " + + "and FhirIO.Read.DEAD_LETTER tuple tags"); + } + } + + private Result(PCollectionTuple pct) { + this.pct = pct; + this.resources = pct.get(OUT).setCoder(new HttpBodyCoder()); + this.failedReads = + pct.get(DEAD_LETTER).setCoder(new HealthcareIOErrorCoder<>(StringUtf8Coder.of())); + } + + public PCollection<HealthcareIOError<String>> getFailedReads() { + return failedReads; + } + + public PCollection<HttpBody> getResources() { + return resources; + } + + @Override + public Pipeline getPipeline() { + return this.pct.getPipeline(); + } + + @Override + public Map<TupleTag<?>, PValue> expand() { + return ImmutableMap.of(OUT, resources); + } + + @Override + public void finishSpecifyingOutput( + String transformName, PInput input, PTransform<?, ?> transform) {} + } + + /** The tag for the main output of Fhir Messages. */ + public static final TupleTag<HttpBody> OUT = new TupleTag<HttpBody>() {}; + /** The tag for the deadletter output of Fhir Messages. */ + public static final TupleTag<HealthcareIOError<String>> DEAD_LETTER = + new TupleTag<HealthcareIOError<String>>() {}; + + @Override + public FhirIO.Read.Result expand(PCollection<String> input) { + return input.apply("Fetch Fhir messages", new FhirIO.Read.FetchHttpBody()); + } + + /** + * DoFn to fetch a resource from an Google Cloud Healthcare FHIR store based on resourceID + * + * <p>This DoFn consumes a {@link PCollection} of notifications {@link String}s from the FHIR + * store, and fetches the actual {@link HttpBody} object based on the id in the notification and + * will output a {@link PCollectionTuple} which contains the output and dead-letter {@link + * PCollection}. + * + * <p>The {@link PCollectionTuple} output will contain the following {@link PCollection}: + * + * <ul> + * <li>{@link FhirIO.Read#OUT} - Contains all {@link PCollection} records successfully read + * from the Fhir store. + * <li>{@link FhirIO.Read#DEAD_LETTER} - Contains all {@link PCollection} of {@link + * HealthcareIOError} of message IDs which failed to be fetched from the Fhir store, with + * error message and stacktrace. + * </ul> + */ + public static class FetchHttpBody extends PTransform<PCollection<String>, FhirIO.Read.Result> { + + /** Instantiates a new Fetch Fhir message DoFn. */ + public FetchHttpBody() {} + + @Override + public FhirIO.Read.Result expand(PCollection<String> resourceIds) { + return new FhirIO.Read.Result( + resourceIds.apply( + ParDo.of(new FhirIO.Read.FetchHttpBody.HttpBodyGetFn()) + .withOutputTags(FhirIO.Read.OUT, TupleTagList.of(FhirIO.Read.DEAD_LETTER)))); + } + + /** DoFn for fetching messages from the Fhir store with error handling. */ + public static class HttpBodyGetFn extends DoFn<String, HttpBody> { + + private Counter failedMessageGets = + Metrics.counter(FhirIO.Read.FetchHttpBody.HttpBodyGetFn.class, "failed-message-reads"); + private static final Logger LOG = + LoggerFactory.getLogger(FhirIO.Read.FetchHttpBody.HttpBodyGetFn.class); + private final Counter successfulHttpBodyGets = + Metrics.counter( + FhirIO.Read.FetchHttpBody.HttpBodyGetFn.class, "successful-hl7v2-message-gets"); + private HealthcareApiClient client; + + /** Instantiates a new Hl 7 v 2 message get fn. */ + HttpBodyGetFn() {} + + /** + * Instantiate healthcare client. + * + * @throws IOException the io exception + */ + @Setup + public void instantiateHealthcareClient() throws IOException { + this.client = new HttpHealthcareApiClient(); + } + + /** + * Process element. + * + * @param context the context + */ + @ProcessElement + public void processElement(ProcessContext context) { + String resourceId = context.element(); + try { + context.output(fetchResource(this.client, resourceId)); + } catch (Exception e) { + failedMessageGets.inc(); + LOG.warn( + String.format( + "Error fetching Fhir message with ID %s writing to Dead Letter " + + "Queue. Cause: %s Stack Trace: %s", + resourceId, e.getMessage(), Throwables.getStackTraceAsString(e))); + context.output(FhirIO.Read.DEAD_LETTER, HealthcareIOError.of(resourceId, e)); + } + } + + private HttpBody fetchResource(HealthcareApiClient client, String resourceId) + throws IOException, IllegalArgumentException, InterruptedException { + long startTime = System.currentTimeMillis(); + Sleeper sleeper = Sleeper.DEFAULT; + + com.google.api.services.healthcare.v1beta1.model.HttpBody resource = + client.readFhirResource(resourceId); + + if (resource == null) { + throw new IOException(String.format("GET request for %s returned null", resourceId)); + } + this.successfulHttpBodyGets.inc(); + return resource; + } + } + } + } + + /** The type Write. */ + @AutoValue + public abstract static class Write extends PTransform<PCollection<HttpBody>, Write.Result> { + + /** The tag for the failed writes to FHIR store`. */ + public static final TupleTag<HealthcareIOError<HttpBody>> FAILED_BODY = + new TupleTag<HealthcareIOError<HttpBody>>() {}; + + /** The enum Write method. */ + public enum WriteMethod { + /** + * Execute Bundle Method executes a batch of requests as a single transaction @see <a + * href=https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores.fhir/executeBundle></a>. + */ + EXECUTE_BUNDLE, + /** + * Import Method bulk imports resources from GCS. This is ideal for initial loads to empty + * FHIR stores. <a + * href=https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores/import></a>. + */ + IMPORT + } + + public static class Result implements POutput { + private final Pipeline pipeline; + private final PCollection<HealthcareIOError<HttpBody>> failedInsertsWithErr; + + /** Creates a {@link FhirIO.Write.Result} in the given {@link Pipeline}. */ + static Result in(Pipeline pipeline, PCollection<HealthcareIOError<HttpBody>> failedInserts) { + return new Result(pipeline, failedInserts); + } + + public PCollection<HealthcareIOError<HttpBody>> getFailedInsertsWithErr() { + return this.failedInsertsWithErr; + } + + @Override + public Pipeline getPipeline() { + return this.pipeline; + } + + @Override + public Map<TupleTag<?>, PValue> expand() { + failedInsertsWithErr.setCoder(new HealthcareIOErrorCoder<HttpBody>(new HttpBodyCoder())); + return ImmutableMap.of(Write.FAILED_BODY, failedInsertsWithErr); + } + + @Override + public void finishSpecifyingOutput( + String transformName, PInput input, PTransform<?, ?> transform) {} + + private Result( + Pipeline pipeline, PCollection<HealthcareIOError<HttpBody>> failedInsertsWithErr) { + this.pipeline = pipeline; + this.failedInsertsWithErr = failedInsertsWithErr; + } + } + + /** + * Gets Fhir store. + * + * @return the Fhir store + */ + abstract String getFhirStore(); + + /** + * Gets write method. + * + * @return the write method + */ + abstract WriteMethod getWriteMethod(); + + abstract Optional<ContentStructure> getContentStructure(); + + abstract Optional<String> getImportGcsTempPath(); + + abstract Optional<String> getImportGcsDeadLetterPath(); + + /** The type Builder. */ + @AutoValue.Builder + abstract static class Builder { + + /** + * Sets Fhir store. + * + * @param fhirStore the Fhir store + * @return the Fhir store + */ + abstract Builder setFhirStore(String fhirStore); + + /** + * Sets write method. + * + * @param writeMethod the write method + * @return the write method + */ + abstract Builder setWriteMethod(WriteMethod writeMethod); + + abstract Builder setContentStructure(ContentStructure contentStructure); + + abstract Builder setImportGcsTempPath(String gcsTempPath); + + abstract Builder setImportGcsDeadLetterPath(String gcsDeadLetterPath); + + /** + * Build write. + * + * @return the write + */ + abstract Write build(); + } + + private static Write.Builder write(String fhirStore) { + return new AutoValue_FhirIO_Write.Builder().setFhirStore(fhirStore); + } + + /** + * Create Method creates a single FHIR resource. @see <a + * href=https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores.fhir/create></a> + * + * @param fhirStore the hl 7 v 2 store + * @return the write + */ + public static Write fhirStoresImport( + String fhirStore, + String gcsTempPath, + String gcsDeadLetterPath, + @Nullable ContentStructure contentStructure) { + return new AutoValue_FhirIO_Write.Builder() + .setFhirStore(fhirStore) + .setWriteMethod(Write.WriteMethod.IMPORT) + .setContentStructure(contentStructure) + .setImportGcsTempPath(gcsTempPath) + .setImportGcsDeadLetterPath(gcsDeadLetterPath) + .build(); + } + + /** + * Execute Bundle Method executes a batch of requests as a single transaction @see <a + * href=https://cloud.google.com/healthcare/docs/reference/rest/v1beta1/projects.locations.datasets.fhirStores.fhir/executeBundle></a>. + * + * @param fhirStore the hl 7 v 2 store + * @return the write + */ + public static Write executeBundles(String fhirStore) { + return new AutoValue_FhirIO_Write.Builder() + .setFhirStore(fhirStore) + .setWriteMethod(WriteMethod.EXECUTE_BUNDLE) + .build(); + } + + private static final Logger LOG = LoggerFactory.getLogger(Write.class); + + @Override + public Result expand(PCollection<HttpBody> input) { + PCollection<HealthcareIOError<HttpBody>> failedBundles; + PCollection<HealthcareIOError<String>> failedImports; + switch (this.getWriteMethod()) { + case IMPORT: + LOG.warn( + "Make sure the Cloud Healthcare Service Agent has permissions when using import:" + + " https://cloud.google.com/healthcare/docs/how-tos/permissions-healthcare-api-gcp-products#fhir_store_cloud_storage_permissions";); + String tempPath = getImportGcsTempPath().orElseThrow(IllegalArgumentException::new); + String deadPath = getImportGcsDeadLetterPath().orElseThrow(IllegalArgumentException::new); + ContentStructure contentStructure = + getContentStructure().orElseThrow(IllegalArgumentException::new); + + failedBundles = + input + .apply(new Import(getFhirStore(), tempPath, deadPath, contentStructure)) + .setCoder(new HealthcareIOErrorCoder<>(new HttpBodyCoder())); + // fall through + case EXECUTE_BUNDLE: + default: + failedBundles = + input.apply( + "Execute FHIR Bundles", + ParDo.of(new ExecuteBundles.ExecuteBundlesFn(this.getFhirStore()))); + } + return Result.in(input.getPipeline(), failedBundles); + } + } + + /** + * Writes each bundle of elements to a new-line delimited JSON file on GCS and issues a + * fhirStores.import Request for that file. + */ + public static class Import + extends PTransform<PCollection<HttpBody>, PCollection<HealthcareIOError<HttpBody>>> { + + private final String fhirStore; + private final String tempGcsPath; + private final String deadLetterGcsPath; + private final ContentStructure contentStructure; + + Import( + String fhirStore, + String tempGcsPath, + String deadLetterGcsPath, + @Nullable ContentStructure contentStructure) { + this.fhirStore = fhirStore; + this.tempGcsPath = tempGcsPath; + this.deadLetterGcsPath = deadLetterGcsPath; + if (contentStructure == null) { + this.contentStructure = ContentStructure.CONTENT_STRUCTURE_UNSPECIFIED; + } else { + this.contentStructure = contentStructure; + } + } + + @Override + public PCollection<HealthcareIOError<HttpBody>> expand(PCollection<HttpBody> input) { + return input.apply( + ParDo.of(new ImportFn(fhirStore, tempGcsPath, deadLetterGcsPath, contentStructure))); + } + + public enum ContentStructure { + /** If the content structure is not specified, the default value BUNDLE will be used. */ + CONTENT_STRUCTURE_UNSPECIFIED, + /** + * The source file contains one or more lines of newline-delimited JSON (ndjson). Each line is + * a bundle, which contains one or more resources. Set the bundle type to history to import + * resource versions. + */ + BUNDLE, + /** + * The source file contains one or more lines of newline-delimited JSON (ndjson). Each line is + * a single resource. + */ + RESOURCE, + /** The entire file is one JSON bundle. The JSON can span multiple lines. */ + BUNDLE_PRETTY, + /** The entire file is one JSON resource. The JSON can span multiple lines. */ + RESOURCE_PRETTY + } + + static class ImportFn extends DoFn<HttpBody, HealthcareIOError<HttpBody>> { + private final String fhirStore; + private final String tempGcsPath; + private final String deadLetterGcsPath; + private final ContentStructure contentStructure; + private ObjectMapper mapper; + private ResourceId resourceId; + private ResourceId deadLetterResourceId; + private WritableByteChannel ndJsonChannel; + + private transient HealthcareApiClient client; + private static final Logger LOG = LoggerFactory.getLogger(ImportFn.class); + + ImportFn( + String fhirStore, + String tempGcsPath, + String deadLetterGcsPath, + @Nullable ContentStructure contentStructure) { + this.fhirStore = fhirStore; + this.tempGcsPath = tempGcsPath; + this.deadLetterGcsPath = deadLetterGcsPath; + if (contentStructure == null) { + this.contentStructure = ContentStructure.CONTENT_STRUCTURE_UNSPECIFIED; + } else { + this.contentStructure = contentStructure; + } + } + + @Setup + public void initClient() throws IOException { + this.client = new HttpHealthcareApiClient(); + } + + @StartBundle + public void initBatch() throws IOException { Review comment: Research: - Based on [this SO post](https://stackoverflow.com/questions/44255924/error-message-too-many-sources-provided-15285-limit-is-10000) We might be able to `GroupByKey` single dummy key to encourage runner to do less bundle splitting, though I'm not sure that will really do the trick. - From Beam docs: > Specifically, it makes it impossible to batch any operations, such as writing elements to a sink or checkpointing progress during processing. >The division of the collection into bundles is arbitrary and selected by the runner. This allows the runner to choose an appropriate middle-ground between persisting results after every element, and having to retry everything if there is a failure. For example, a streaming runner may prefer to process and commit small bundles, and a batch runner may prefer to process larger bundles. ([source](https://beam.apache.org/documentation/runtime/model/#bundling-and-persistence)) It seems to me that our predominant use case for Import is a batch pipeline and this sentence implies that bundles are larger in batch (not sure if we can find internal details on Dataflow Runner). - BigQueryIO has a [batch loads](https://github.com/apache/beam/blob/master/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigquery/BatchLoads.java) pattern which seems similar to what we are aiming to achieve here. This seems quite complex and I'm wondering if there are simpler approaches elsewhere in the beam code base. Do we have any guidance on how many parallel import operations would be acceptable? Perhaps we can run some benchmarks with 500, 1000, 10000 resources with this approach / adding group by single key and see if it's acceptable. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
