Search in sources :

Example 1 with LongFrameToLongWritableFrameFunction

use of org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction in project incubator-systemml by apache.

the class WriteSPInstruction method processFrameWriteInstruction.

@SuppressWarnings("unchecked")
protected void processFrameWriteInstruction(SparkExecutionContext sec, String fname, OutputInfo oi, ValueType[] schema) throws DMLRuntimeException, IOException {
    //get input rdd
    JavaPairRDD<Long, FrameBlock> in1 = (JavaPairRDD<Long, FrameBlock>) sec.getRDDHandleForVariable(input1.getName(), InputInfo.BinaryBlockInputInfo);
    MatrixCharacteristics mc = sec.getMatrixCharacteristics(input1.getName());
    if (oi == OutputInfo.TextCellOutputInfo) {
        JavaRDD<String> out = FrameRDDConverterUtils.binaryBlockToTextCell(in1, mc);
        customSaveTextFile(out, fname, false);
    } else if (oi == OutputInfo.CSVOutputInfo) {
        CSVFileFormatProperties props = (formatProperties != null) ? (CSVFileFormatProperties) formatProperties : null;
        JavaRDD<String> out = FrameRDDConverterUtils.binaryBlockToCsv(in1, mc, props, true);
        customSaveTextFile(out, fname, false);
    } else if (oi == OutputInfo.BinaryBlockOutputInfo) {
        JavaPairRDD<LongWritable, FrameBlock> out = in1.mapToPair(new LongFrameToLongWritableFrameFunction());
        out.saveAsHadoopFile(fname, LongWritable.class, FrameBlock.class, SequenceFileOutputFormat.class);
    } else {
        //unsupported formats: binarycell (not externalized)
        throw new DMLRuntimeException("Unexpected data format: " + OutputInfo.outputInfoToString(oi));
    }
    // write meta data file
    MapReduceTool.writeMetaDataFile(fname + ".mtd", input1.getValueType(), schema, DataType.FRAME, mc, oi, formatProperties);
}
Also used : CSVFileFormatProperties(org.apache.sysml.runtime.matrix.data.CSVFileFormatProperties) MatrixCharacteristics(org.apache.sysml.runtime.matrix.MatrixCharacteristics) JavaRDD(org.apache.spark.api.java.JavaRDD) DMLRuntimeException(org.apache.sysml.runtime.DMLRuntimeException) FrameBlock(org.apache.sysml.runtime.matrix.data.FrameBlock) JavaPairRDD(org.apache.spark.api.java.JavaPairRDD) LongWritable(org.apache.hadoop.io.LongWritable) LongFrameToLongWritableFrameFunction(org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction)

Example 2 with LongFrameToLongWritableFrameFunction

use of org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction in project incubator-systemml by apache.

the class FrameConverterTest method runConverter.

/**
	 * @param oinfo 
	 * @param frame1
	 * @param frame2
	 * @param fprop
	 * @param schema
	 * @return 
	 * @throws DMLRuntimeException, IOException
	 */
@SuppressWarnings("unchecked")
private void runConverter(ConvType type, MatrixCharacteristics mc, MatrixCharacteristics mcMatrix, List<ValueType> schema, String fnameIn, String fnameOut) throws DMLRuntimeException, IOException {
    SparkExecutionContext sec = (SparkExecutionContext) ExecutionContextFactory.createContext();
    JavaSparkContext sc = sec.getSparkContext();
    ValueType[] lschema = schema.toArray(new ValueType[0]);
    MapReduceTool.deleteFileIfExistOnHDFS(fnameOut);
    switch(type) {
        case CSV2BIN:
            {
                InputInfo iinfo = InputInfo.CSVInputInfo;
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                JavaPairRDD<LongWritable, Text> rddIn = (JavaPairRDD<LongWritable, Text>) sc.hadoopFile(fnameIn, iinfo.inputFormatClass, iinfo.inputKeyClass, iinfo.inputValueClass);
                JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.csvToBinaryBlock(sc, rddIn, mc, null, false, separator, false, 0).mapToPair(new LongFrameToLongWritableFrameFunction());
                rddOut.saveAsHadoopFile(fnameOut, LongWritable.class, FrameBlock.class, oinfo.outputFormatClass);
                break;
            }
        case BIN2CSV:
            {
                InputInfo iinfo = InputInfo.BinaryBlockInputInfo;
                JavaPairRDD<LongWritable, FrameBlock> rddIn = sc.hadoopFile(fnameIn, iinfo.inputFormatClass, LongWritable.class, FrameBlock.class);
                JavaPairRDD<Long, FrameBlock> rddIn2 = rddIn.mapToPair(new CopyFrameBlockPairFunction(false));
                CSVFileFormatProperties fprop = new CSVFileFormatProperties();
                JavaRDD<String> rddOut = FrameRDDConverterUtils.binaryBlockToCsv(rddIn2, mc, fprop, true);
                rddOut.saveAsTextFile(fnameOut);
                break;
            }
        case TXTCELL2BIN:
            {
                InputInfo iinfo = InputInfo.TextCellInputInfo;
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                JavaPairRDD<LongWritable, Text> rddIn = (JavaPairRDD<LongWritable, Text>) sc.hadoopFile(fnameIn, iinfo.inputFormatClass, iinfo.inputKeyClass, iinfo.inputValueClass);
                JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.textCellToBinaryBlock(sc, rddIn, mc, lschema).mapToPair(new LongFrameToLongWritableFrameFunction());
                rddOut.saveAsHadoopFile(fnameOut, LongWritable.class, FrameBlock.class, oinfo.outputFormatClass);
                break;
            }
        case BIN2TXTCELL:
            {
                InputInfo iinfo = InputInfo.BinaryBlockInputInfo;
                JavaPairRDD<LongWritable, FrameBlock> rddIn = sc.hadoopFile(fnameIn, iinfo.inputFormatClass, LongWritable.class, FrameBlock.class);
                JavaPairRDD<Long, FrameBlock> rddIn2 = rddIn.mapToPair(new CopyFrameBlockPairFunction(false));
                JavaRDD<String> rddOut = FrameRDDConverterUtils.binaryBlockToTextCell(rddIn2, mc);
                rddOut.saveAsTextFile(fnameOut);
                break;
            }
        case MAT2BIN:
            {
                InputInfo iinfo = InputInfo.BinaryBlockInputInfo;
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                JavaPairRDD<MatrixIndexes, MatrixBlock> rddIn = (JavaPairRDD<MatrixIndexes, MatrixBlock>) sc.hadoopFile(fnameIn, iinfo.inputFormatClass, iinfo.inputKeyClass, iinfo.inputValueClass);
                JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.matrixBlockToBinaryBlock(sc, rddIn, mcMatrix);
                rddOut.saveAsHadoopFile(fnameOut, LongWritable.class, FrameBlock.class, oinfo.outputFormatClass);
                break;
            }
        case BIN2MAT:
            {
                InputInfo iinfo = InputInfo.BinaryBlockInputInfo;
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                JavaPairRDD<Long, FrameBlock> rddIn = sc.hadoopFile(fnameIn, iinfo.inputFormatClass, LongWritable.class, FrameBlock.class).mapToPair(new LongWritableFrameToLongFrameFunction());
                JavaPairRDD<MatrixIndexes, MatrixBlock> rddOut = FrameRDDConverterUtils.binaryBlockToMatrixBlock(rddIn, mc, mcMatrix);
                rddOut.saveAsHadoopFile(fnameOut, MatrixIndexes.class, MatrixBlock.class, oinfo.outputFormatClass);
                break;
            }
        case DFRM2BIN:
            {
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                //Create DataFrame 
                SparkSession sparkSession = SparkSession.builder().sparkContext(sc.sc()).getOrCreate();
                StructType dfSchema = FrameRDDConverterUtils.convertFrameSchemaToDFSchema(lschema, false);
                JavaRDD<Row> rowRDD = FrameRDDConverterUtils.csvToRowRDD(sc, fnameIn, separator, lschema);
                Dataset<Row> df = sparkSession.createDataFrame(rowRDD, dfSchema);
                JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.dataFrameToBinaryBlock(sc, df, mc, false).mapToPair(new LongFrameToLongWritableFrameFunction());
                rddOut.saveAsHadoopFile(fnameOut, LongWritable.class, FrameBlock.class, oinfo.outputFormatClass);
                break;
            }
        case BIN2DFRM:
            {
                InputInfo iinfo = InputInfo.BinaryBlockInputInfo;
                OutputInfo oinfo = OutputInfo.BinaryBlockOutputInfo;
                JavaPairRDD<Long, FrameBlock> rddIn = sc.hadoopFile(fnameIn, iinfo.inputFormatClass, LongWritable.class, FrameBlock.class).mapToPair(new LongWritableFrameToLongFrameFunction());
                SparkSession sparkSession = SparkSession.builder().sparkContext(sc.sc()).getOrCreate();
                Dataset<Row> df = FrameRDDConverterUtils.binaryBlockToDataFrame(sparkSession, rddIn, mc, lschema);
                //Convert back DataFrame to binary block for comparison using original binary to converted DF and back to binary 
                JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.dataFrameToBinaryBlock(sc, df, mc, true).mapToPair(new LongFrameToLongWritableFrameFunction());
                rddOut.saveAsHadoopFile(fnameOut, LongWritable.class, FrameBlock.class, oinfo.outputFormatClass);
                break;
            }
        default:
            throw new RuntimeException("Unsuported converter type: " + type.toString());
    }
    sec.close();
}
Also used : MatrixBlock(org.apache.sysml.runtime.matrix.data.MatrixBlock) CSVFileFormatProperties(org.apache.sysml.runtime.matrix.data.CSVFileFormatProperties) SparkSession(org.apache.spark.sql.SparkSession) StructType(org.apache.spark.sql.types.StructType) ValueType(org.apache.sysml.parser.Expression.ValueType) MatrixIndexes(org.apache.sysml.runtime.matrix.data.MatrixIndexes) Dataset(org.apache.spark.sql.Dataset) Text(org.apache.hadoop.io.Text) JavaRDD(org.apache.spark.api.java.JavaRDD) OutputInfo(org.apache.sysml.runtime.matrix.data.OutputInfo) InputInfo(org.apache.sysml.runtime.matrix.data.InputInfo) DMLRuntimeException(org.apache.sysml.runtime.DMLRuntimeException) FrameBlock(org.apache.sysml.runtime.matrix.data.FrameBlock) JavaPairRDD(org.apache.spark.api.java.JavaPairRDD) LongWritableFrameToLongFrameFunction(org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongWritableFrameToLongFrameFunction) SparkExecutionContext(org.apache.sysml.runtime.controlprogram.context.SparkExecutionContext) JavaSparkContext(org.apache.spark.api.java.JavaSparkContext) LongWritable(org.apache.hadoop.io.LongWritable) LongFrameToLongWritableFrameFunction(org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction) CopyFrameBlockPairFunction(org.apache.sysml.runtime.instructions.spark.functions.CopyFrameBlockPairFunction)

Example 3 with LongFrameToLongWritableFrameFunction

use of org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction in project incubator-systemml by apache.

the class FrameTest method testFrameGeneral.

private void testFrameGeneral(InputInfo iinfo, OutputInfo oinfo, boolean bFromDataFrame, boolean bToDataFrame) throws IOException, DMLException, ParseException {
    boolean oldConfig = DMLScript.USE_LOCAL_SPARK_CONFIG;
    DMLScript.USE_LOCAL_SPARK_CONFIG = true;
    RUNTIME_PLATFORM oldRT = DMLScript.rtplatform;
    DMLScript.rtplatform = RUNTIME_PLATFORM.HYBRID_SPARK;
    int rowstart = 234, rowend = 1478, colstart = 125, colend = 568;
    int bRows = rowend - rowstart + 1, bCols = colend - colstart + 1;
    int rowstartC = 124, rowendC = 1178, colstartC = 143, colendC = 368;
    int cRows = rowendC - rowstartC + 1, cCols = colendC - colstartC + 1;
    HashMap<String, ValueType[]> outputSchema = new HashMap<String, ValueType[]>();
    HashMap<String, MatrixCharacteristics> outputMC = new HashMap<String, MatrixCharacteristics>();
    TestConfiguration config = getTestConfiguration(TEST_NAME);
    loadTestConfiguration(config);
    List<String> proArgs = new ArrayList<String>();
    proArgs.add(input("A"));
    proArgs.add(Integer.toString(rows));
    proArgs.add(Integer.toString(cols));
    proArgs.add(input("B"));
    proArgs.add(Integer.toString(bRows));
    proArgs.add(Integer.toString(bCols));
    proArgs.add(Integer.toString(rowstart));
    proArgs.add(Integer.toString(rowend));
    proArgs.add(Integer.toString(colstart));
    proArgs.add(Integer.toString(colend));
    proArgs.add(output("A"));
    proArgs.add(Integer.toString(rowstartC));
    proArgs.add(Integer.toString(rowendC));
    proArgs.add(Integer.toString(colstartC));
    proArgs.add(Integer.toString(colendC));
    proArgs.add(output("C"));
    fullDMLScriptName = SCRIPT_DIR + TEST_DIR + TEST_NAME + ".dml";
    ValueType[] schema = schemaMixedLarge;
    //initialize the frame data.
    List<ValueType> lschema = Arrays.asList(schema);
    fullRScriptName = SCRIPT_DIR + TEST_DIR + TEST_NAME + ".R";
    rCmd = "Rscript" + " " + fullRScriptName + " " + inputDir() + " " + rowstart + " " + rowend + " " + colstart + " " + colend + " " + expectedDir() + " " + rowstartC + " " + rowendC + " " + colstartC + " " + colendC;
    double sparsity = sparsity1;
    double[][] A = getRandomMatrix(rows, cols, min, max, sparsity, 1111);
    writeInputFrameWithMTD("A", A, true, schema, oinfo);
    sparsity = sparsity2;
    double[][] B = getRandomMatrix((int) (bRows), (int) (bCols), min, max, sparsity, 2345);
    ValueType[] schemaB = new ValueType[bCols];
    for (int i = 0; i < bCols; ++i) schemaB[i] = schema[colstart - 1 + i];
    List<ValueType> lschemaB = Arrays.asList(schemaB);
    writeInputFrameWithMTD("B", B, true, schemaB, oinfo);
    ValueType[] schemaC = new ValueType[colendC - colstartC + 1];
    for (int i = 0; i < cCols; ++i) schemaC[i] = schema[colstartC - 1 + i];
    Dataset<Row> dfA = null, dfB = null;
    if (bFromDataFrame) {
        //Create DataFrame for input A 
        StructType dfSchemaA = FrameRDDConverterUtils.convertFrameSchemaToDFSchema(schema, false);
        JavaRDD<Row> rowRDDA = FrameRDDConverterUtils.csvToRowRDD(sc, input("A"), DataExpression.DEFAULT_DELIM_DELIMITER, schema);
        dfA = spark.createDataFrame(rowRDDA, dfSchemaA);
        //Create DataFrame for input B 
        StructType dfSchemaB = FrameRDDConverterUtils.convertFrameSchemaToDFSchema(schemaB, false);
        JavaRDD<Row> rowRDDB = FrameRDDConverterUtils.csvToRowRDD(sc, input("B"), DataExpression.DEFAULT_DELIM_DELIMITER, schemaB);
        dfB = spark.createDataFrame(rowRDDB, dfSchemaB);
    }
    try {
        Script script = ScriptFactory.dmlFromFile(fullDMLScriptName);
        String format = "csv";
        if (oinfo == OutputInfo.TextCellOutputInfo)
            format = "text";
        if (bFromDataFrame) {
            script.in("A", dfA);
        } else {
            JavaRDD<String> aIn = sc.textFile(input("A"));
            FrameSchema fs = new FrameSchema(lschema);
            FrameFormat ff = (format.equals("text")) ? FrameFormat.IJV : FrameFormat.CSV;
            FrameMetadata fm = new FrameMetadata(ff, fs, rows, cols);
            script.in("A", aIn, fm);
        }
        if (bFromDataFrame) {
            script.in("B", dfB);
        } else {
            JavaRDD<String> bIn = sc.textFile(input("B"));
            FrameSchema fs = new FrameSchema(lschemaB);
            FrameFormat ff = (format.equals("text")) ? FrameFormat.IJV : FrameFormat.CSV;
            FrameMetadata fm = new FrameMetadata(ff, fs, bRows, bCols);
            script.in("B", bIn, fm);
        }
        // Output one frame to HDFS and get one as RDD //TODO HDFS input/output to do
        script.out("A", "C");
        // set positional argument values
        for (int argNum = 1; argNum <= proArgs.size(); argNum++) {
            script.in("$" + argNum, proArgs.get(argNum - 1));
        }
        MLResults results = ml.execute(script);
        format = "csv";
        if (iinfo == InputInfo.TextCellInputInfo)
            format = "text";
        String fName = output("AB");
        try {
            MapReduceTool.deleteFileIfExistOnHDFS(fName);
        } catch (IOException e) {
            throw new DMLRuntimeException("Error: While deleting file on HDFS");
        }
        if (!bToDataFrame) {
            if (format.equals("text")) {
                JavaRDD<String> javaRDDStringIJV = results.getJavaRDDStringIJV("A");
                javaRDDStringIJV.saveAsTextFile(fName);
            } else {
                JavaRDD<String> javaRDDStringCSV = results.getJavaRDDStringCSV("A");
                javaRDDStringCSV.saveAsTextFile(fName);
            }
        } else {
            Dataset<Row> df = results.getDataFrame("A");
            //Convert back DataFrame to binary block for comparison using original binary to converted DF and back to binary 
            MatrixCharacteristics mc = new MatrixCharacteristics(rows, cols, -1, -1, -1);
            JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.dataFrameToBinaryBlock(sc, df, mc, bFromDataFrame).mapToPair(new LongFrameToLongWritableFrameFunction());
            rddOut.saveAsHadoopFile(output("AB"), LongWritable.class, FrameBlock.class, OutputInfo.BinaryBlockOutputInfo.outputFormatClass);
        }
        fName = output("C");
        try {
            MapReduceTool.deleteFileIfExistOnHDFS(fName);
        } catch (IOException e) {
            throw new DMLRuntimeException("Error: While deleting file on HDFS");
        }
        if (!bToDataFrame) {
            if (format.equals("text")) {
                JavaRDD<String> javaRDDStringIJV = results.getJavaRDDStringIJV("C");
                javaRDDStringIJV.saveAsTextFile(fName);
            } else {
                JavaRDD<String> javaRDDStringCSV = results.getJavaRDDStringCSV("C");
                javaRDDStringCSV.saveAsTextFile(fName);
            }
        } else {
            Dataset<Row> df = results.getDataFrame("C");
            //Convert back DataFrame to binary block for comparison using original binary to converted DF and back to binary 
            MatrixCharacteristics mc = new MatrixCharacteristics(cRows, cCols, -1, -1, -1);
            JavaPairRDD<LongWritable, FrameBlock> rddOut = FrameRDDConverterUtils.dataFrameToBinaryBlock(sc, df, mc, bFromDataFrame).mapToPair(new LongFrameToLongWritableFrameFunction());
            rddOut.saveAsHadoopFile(fName, LongWritable.class, FrameBlock.class, OutputInfo.BinaryBlockOutputInfo.outputFormatClass);
        }
        runRScript(true);
        outputSchema.put("AB", schema);
        outputMC.put("AB", new MatrixCharacteristics(rows, cols, -1, -1));
        outputSchema.put("C", schemaC);
        outputMC.put("C", new MatrixCharacteristics(cRows, cCols, -1, -1));
        for (String file : config.getOutputFiles()) {
            MatrixCharacteristics md = outputMC.get(file);
            FrameBlock frameBlock = readDMLFrameFromHDFS(file, iinfo, md);
            FrameBlock frameRBlock = readRFrameFromHDFS(file + ".csv", InputInfo.CSVInputInfo, md);
            ValueType[] schemaOut = outputSchema.get(file);
            verifyFrameData(frameBlock, frameRBlock, schemaOut);
            System.out.println("File " + file + " processed successfully.");
        }
        System.out.println("Frame MLContext test completed successfully.");
    } finally {
        DMLScript.rtplatform = oldRT;
        DMLScript.USE_LOCAL_SPARK_CONFIG = oldConfig;
    }
}
Also used : FrameFormat(org.apache.sysml.api.mlcontext.FrameFormat) StructType(org.apache.spark.sql.types.StructType) HashMap(java.util.HashMap) MLResults(org.apache.sysml.api.mlcontext.MLResults) TestConfiguration(org.apache.sysml.test.integration.TestConfiguration) ArrayList(java.util.ArrayList) FrameBlock(org.apache.sysml.runtime.matrix.data.FrameBlock) LongWritable(org.apache.hadoop.io.LongWritable) LongFrameToLongWritableFrameFunction(org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction) Script(org.apache.sysml.api.mlcontext.Script) DMLScript(org.apache.sysml.api.DMLScript) ValueType(org.apache.sysml.parser.Expression.ValueType) FrameSchema(org.apache.sysml.api.mlcontext.FrameSchema) IOException(java.io.IOException) MatrixCharacteristics(org.apache.sysml.runtime.matrix.MatrixCharacteristics) DMLRuntimeException(org.apache.sysml.runtime.DMLRuntimeException) RUNTIME_PLATFORM(org.apache.sysml.api.DMLScript.RUNTIME_PLATFORM) Row(org.apache.spark.sql.Row) FrameMetadata(org.apache.sysml.api.mlcontext.FrameMetadata)

Aggregations

LongWritable (org.apache.hadoop.io.LongWritable)3 DMLRuntimeException (org.apache.sysml.runtime.DMLRuntimeException)3 LongFrameToLongWritableFrameFunction (org.apache.sysml.runtime.instructions.spark.utils.FrameRDDConverterUtils.LongFrameToLongWritableFrameFunction)3 FrameBlock (org.apache.sysml.runtime.matrix.data.FrameBlock)3 JavaPairRDD (org.apache.spark.api.java.JavaPairRDD)2 JavaRDD (org.apache.spark.api.java.JavaRDD)2 StructType (org.apache.spark.sql.types.StructType)2 ValueType (org.apache.sysml.parser.Expression.ValueType)2 MatrixCharacteristics (org.apache.sysml.runtime.matrix.MatrixCharacteristics)2 CSVFileFormatProperties (org.apache.sysml.runtime.matrix.data.CSVFileFormatProperties)2 IOException (java.io.IOException)1 ArrayList (java.util.ArrayList)1 HashMap (java.util.HashMap)1 Text (org.apache.hadoop.io.Text)1 JavaSparkContext (org.apache.spark.api.java.JavaSparkContext)1 Dataset (org.apache.spark.sql.Dataset)1 Row (org.apache.spark.sql.Row)1 SparkSession (org.apache.spark.sql.SparkSession)1 DMLScript (org.apache.sysml.api.DMLScript)1 RUNTIME_PLATFORM (org.apache.sysml.api.DMLScript.RUNTIME_PLATFORM)1