ALS-6511: Initial commit for PFB response type

client-api/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/data/query/Query.java

@@ -179,7 +179,6 @@ public String toString() {
 			writePartFormat("Observation Count Fields", fields, builder, true);
 			break;
 		case DATAFRAME:
-		case DATAFRAME_MERGED:
 		case SECRET_ADMIN_DATAFRAME:
 			writePartFormat("Data Export Fields", fields, builder, true);
 			break;

client-api/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/data/query/ResultType.java

@@ -49,12 +49,7 @@ public enum ResultType {
 	 * Return the number of observations for included patients and
 	 * included fields, broken up across the included cross count fields.
 	 */
-	OBSERVATION_CROSS_COUNT, 
-	/**
-	 * This was developed for UDN, but is completely useless and should
-	 * be deleted.
-	 */
-	DATAFRAME_MERGED, 
+	OBSERVATION_CROSS_COUNT,
 	/**
 	 * Not completely implemented and currently dead code. Someone with 
 	 * statistics experience needs to develop a p-value based filter for
@@ -94,5 +89,10 @@ public enum ResultType {
 	 * is suitable to time series analysis and/or loading into another 
 	 * instance of HPDS.
 	 */
-	DATAFRAME_TIMESERIES
+	DATAFRAME_TIMESERIES,
+	/**
+     * Exports data as PFB, using avro
+     * <a href="https://uc-cdis.github.io/pypfb/">https://uc-cdis.github.io/pypfb/</a>
+     */
+	DATAFRAME_PFB
 }

pom.xml

@@ -310,7 +310,11 @@
 				<version>1.18.30</version>
 				<scope>provided</scope>
 			</dependency>
-
+			<dependency>
+				<groupId>org.apache.avro</groupId>
+				<artifactId>avro</artifactId>
+				<version>1.11.3</version>
+			</dependency>
 
 		</dependencies>
 	</dependencyManagement>

processing/pom.xml

@@ -33,5 +33,9 @@
 			<groupId>io.projectreactor.netty</groupId>
 			<artifactId>reactor-netty</artifactId>
 		</dependency>
+		<dependency>
+			<groupId>org.apache.avro</groupId>
+			<artifactId>avro</artifactId>
+		</dependency>
 	</dependencies>
 </project>

processing/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/processing/AsyncResult.java

@@ -1,6 +1,9 @@
 package edu.harvard.hms.dbmi.avillach.hpds.processing;
 
 import java.io.IOException;
+import java.io.UncheckedIOException;
+import java.util.ArrayList;
+import java.util.List;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.RejectedExecutionException;
 
@@ -17,7 +20,15 @@
 public class AsyncResult implements Runnable, Comparable<AsyncResult>{
 
 	private static Logger log = LoggerFactory.getLogger(AsyncResult.class);
-
+
+	public byte[] readAllBytes() {
+        try {
+            return stream.readAllBytes();
+        } catch (IOException e) {
+            throw new UncheckedIOException(e);
+        }
+    }
+
 	public static enum Status{
 		SUCCESS {
 			@Override
@@ -52,29 +63,82 @@ public PicSureStatus toPicSureStatus() {
 		public abstract PicSureStatus toPicSureStatus();
 	}
 
-	public Query query;
-
-	public Status status;
-
-	public long queuedTime;
-
-	public long completedTime;
-
-	public int retryCount;
-
-	public int queueDepth;
-
-	public int positionInQueue;
-
-	public int numRows;
+	private Query query;
+
+	public Query getQuery() {
+		return query;
+	}
+
+	private Status status;
+
+	public Status getStatus() {
+		return status;
+	}
+
+	public AsyncResult setStatus(Status status) {
+		this.status = status;
+		return this;
+	}
+
+	private long queuedTime;
+
+	public long getQueuedTime() {
+		return queuedTime;
+	}
+
+	public AsyncResult setQueuedTime(long queuedTime) {
+		this.queuedTime = queuedTime;
+		return this;
+	}
+
+	private long completedTime;
+
+	public long getCompletedTime() {
+		return completedTime;
+	}
+
+	private int retryCount;
 
-	public int numColumns;
+	private int queueDepth;
+
+	public int getQueueDepth() {
+		return queueDepth;
+	}
+
+	public AsyncResult setQueueDepth(int queueDepth) {
+		this.queueDepth = queueDepth;
+		return this;
+	}
+
+	private int positionInQueue;
+
+	public AsyncResult setPositionInQueue(int positionInQueue) {
+		this.positionInQueue = positionInQueue;
+		return this;
+	}
+
+	private int numRows;
 
-	public String id;
+	private int numColumns;
 
+	private String id;
+
+	public String getId() {
+		return id;
+	}
+
+	public AsyncResult setId(String id) {
+		this.id = id;
+		return this;
+	}
+
 	@JsonIgnore
-	public ResultStoreStream stream;
-
+	private ResultStoreStream stream;
+
+	public ResultStoreStream getStream() {
+		return stream;
+	}
+
 	@JsonIgnore
 	private String[] headerRow;
 
@@ -86,21 +150,48 @@ public PicSureStatus toPicSureStatus() {
 	 * The actual exception is thrown in @see ResultStore#constructor
 	 */
 	@JsonIgnore
-	public ExecutorService jobQueue;
+	private ExecutorService jobQueue;
+
+	public ExecutorService getJobQueue() {
+		return jobQueue;
+	}
+
+	public AsyncResult setJobQueue(ExecutorService jobQueue) {
+		this.jobQueue = jobQueue;
+		return this;
+	}
 
 	@JsonIgnore
-	public HpdsProcessor processor;
+	private HpdsProcessor processor;
+
+	public HpdsProcessor getProcessor() {
+		return processor;
+	}
+
+	public AsyncResult setProcessor(HpdsProcessor processor) {
+		this.processor = processor;
+		return this;
+	}
 
 	public AsyncResult(Query query, String[] headerRow) {
 		this.query = query;
 		this.headerRow = headerRow;
 		try {
-			stream = new ResultStoreStream(headerRow, query.getExpectedResultType() == ResultType.DATAFRAME_MERGED);
+			stream = new ResultStoreStream(headerRow);
 		} catch (IOException e) {
 			log.error("Exception creating result stream", e);
 		}
 	}
 
+	public void appendResults(List<String[]> dataEntries) {
+		stream.appendResults(dataEntries);
+	}
+
+	public void appendResultStore(ResultStore resultStore) {
+		stream.appendResultStore(resultStore);
+	}
+
+
 	@Override
 	public void run() {
 		status = AsyncResult.Status.RUNNING;
@@ -127,9 +218,15 @@ public void enqueue() {
 		}
 	}
 
+	public void open() {
+		stream.open();
+	}
+
 	@Override
 	public int compareTo(AsyncResult o) {
 		return this.query.getId().compareTo(o.query.getId());
 	}
-
+
+
+
 }

processing/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/processing/PfbProcessor.java

@@ -0,0 +1,112 @@
+package edu.harvard.hms.dbmi.avillach.hpds.processing;
+
+import com.google.common.collect.Lists;
+import edu.harvard.hms.dbmi.avillach.hpds.data.phenotype.ColumnMeta;
+import edu.harvard.hms.dbmi.avillach.hpds.data.phenotype.KeyAndValue;
+import edu.harvard.hms.dbmi.avillach.hpds.data.phenotype.PhenoCube;
+import edu.harvard.hms.dbmi.avillach.hpds.data.query.Query;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.util.*;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.stream.Collectors;
+
+public class PfbProcessor implements HpdsProcessor {
+
+    public static final String PATIENT_ID_FIELD_NAME = "patient_id";
+    private final int ID_BATCH_SIZE;
+    private final AbstractProcessor abstractProcessor;
+
+    private Logger log = LoggerFactory.getLogger(PfbProcessor.class);
+
+
+    public PfbProcessor(AbstractProcessor abstractProcessor) {
+        this.abstractProcessor = abstractProcessor;
+        ID_BATCH_SIZE = Integer.parseInt(System.getProperty("ID_BATCH_SIZE", "0"));
+    }
+
+    @Override
+    public String[] getHeaderRow(Query query) {
+        String[] header = new String[query.getFields().size()+1];
+        header[0] = PATIENT_ID_FIELD_NAME;
+        System.arraycopy(query.getFields().toArray(), 0, header, 1, query.getFields().size());
+        return header;
+    }
+
+    @Override
+    public void runQuery(Query query, AsyncResult result) {
+        Set<Integer> idList = abstractProcessor.getPatientSubsetForQuery(query);
+        log.info("Processing " + idList.size() + " rows for result " + result.getId());
+        Lists.partition(new ArrayList<>(idList), ID_BATCH_SIZE).stream()
+                .forEach(patientIds -> {
+                    Map<String, Map<Integer, String>> pathToPatientToValueMap = buildResult(result, query, new TreeSet<>(patientIds));
+                    List<String[]> fieldValuesPerPatient = patientIds.stream().map(patientId -> {
+                        return Arrays.stream(getHeaderRow(query)).map(field -> {
+                            if (PATIENT_ID_FIELD_NAME.equals(field)) {
+                                return patientId.toString();
+                            } else {
+                                return pathToPatientToValueMap.get(field).get(patientId);
+                            }
+                        }).toArray(String[]::new);
+                    }).collect(Collectors.toList());
+                    result.appendResults(fieldValuesPerPatient);
+                });
+    }
+
+    private Map<String, Map<Integer, String>> buildResult(AsyncResult result, Query query, TreeSet<Integer> ids) {
+        ConcurrentHashMap<String, Map<Integer, String>> pathToPatientToValueMap = new ConcurrentHashMap<>();
+        List<ColumnMeta> columns = query.getFields().stream()
+                .map(abstractProcessor.getDictionary()::get)
+                .filter(Objects::nonNull)
+                .collect(Collectors.toList());
+        List<String> paths = columns.stream()
+                .map(ColumnMeta::getName)
+                .collect(Collectors.toList());
+        int columnCount = paths.size() + 1;
+
+        ArrayList<Integer> columnIndex = abstractProcessor.useResidentCubesFirst(paths, columnCount);
+        ResultStore results = new ResultStore(result.getId(), columns, ids);
+
+        // todo: investigate if the parallel stream will thrash the cache if the number of executors is > number of resident cubes
+        columnIndex.parallelStream().forEach((columnId)->{
+            String columnPath = paths.get(columnId-1);
+            Map<Integer, String> patientIdToValueMap = processColumn(ids, columnPath);
+            pathToPatientToValueMap.put(columnPath, patientIdToValueMap);
+        });
+
+        return pathToPatientToValueMap;
+    }
+
+    private Map<Integer, String> processColumn(TreeSet<Integer> patientIds, String path) {
+
+        Map<Integer, String> patientIdToValueMap = new HashMap<>();
+        PhenoCube<?> cube = abstractProcessor.getCube(path);
+
+        KeyAndValue<?>[] cubeValues = cube.sortedByKey();
+
+        int idPointer = 0;
+        for(int patientId : patientIds) {
+            while(idPointer < cubeValues.length) {
+                int key = cubeValues[idPointer].getKey();
+                if(key < patientId) {
+                    idPointer++;
+                } else if(key == patientId){
+                    String value = getResultField(cube, cubeValues, idPointer);
+                    patientIdToValueMap.put(patientId, value);
+                    idPointer++;
+                    break;
+                } else {
+                    break;
+                }
+            }
+        }
+        return patientIdToValueMap;
+    }
+
+    private String getResultField(PhenoCube<?> cube, KeyAndValue<?>[] cubeValues,
+                                 int idPointer) {
+        Comparable<?> value = cubeValues[idPointer].getValue();
+        return value.toString();
+    }
+}

processing/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/processing/QueryProcessor.java

@@ -53,11 +53,11 @@ public String[] getHeaderRow(Query query) {
 
 	public void runQuery(Query query, AsyncResult result) {
 		Set<Integer> idList = abstractProcessor.getPatientSubsetForQuery(query);
-		log.info("Processing " + idList.size() + " rows for result " + result.id);
+		log.info("Processing " + idList.size() + " rows for result " + result.getId());
 		Lists.partition(new ArrayList<>(idList), ID_BATCH_SIZE).parallelStream()
 			.map(list -> buildResult(result, query, new TreeSet<>(list)))
 			.sequential()
-			.forEach(result.stream::appendResultStore);
+			.forEach(result::appendResultStore);
 	}
 
 
@@ -72,7 +72,7 @@ private ResultStore buildResult(AsyncResult result, Query query, TreeSet<Integer
 		int columnCount = paths.size() + 1;
 
 		ArrayList<Integer> columnIndex = abstractProcessor.useResidentCubesFirst(paths, columnCount);
-		ResultStore results = new ResultStore(result.id, columns, ids);
+		ResultStore results = new ResultStore(result.getId(), columns, ids);
 
 		columnIndex.parallelStream().forEach((column)->{
 			clearColumn(paths, ids, results, column);

processing/src/main/java/edu/harvard/hms/dbmi/avillach/hpds/processing/ResultStore.java

@@ -124,7 +124,7 @@ private int getFieldOffset(int row, int column) {
 	 * @param row
 	 * @throws IOException
 	 */
-	public void readRowIntoStringArray(int rowNumber, int[] columnWidths, String[] row) throws IOException {
+	public void readRowIntoStringArray(int rowNumber, int[] columnWidths, String[] row) {
 		if(wrappedResultArray == null) {
 			wrappedResultArray = ByteBuffer.wrap(resultArray);
 		}