Examples with StructField org.apache.spark.sql.types.StructField used on opensource projects

Example 46 with StructField

use of org.apache.spark.sql.types.StructField in project incubator-systemml by apache.

the class MLContextTest method testDataFrameGoodMetadataPYDML.

@Test
public void testDataFrameGoodMetadataPYDML() {
    System.out.println("MLContextTest - DataFrame good metadata PYDML");
    List<String> list = new ArrayList<String>();
    list.add("10,20,30");
    list.add("40,50,60");
    list.add("70,80,90");
    JavaRDD<String> javaRddString = sc.parallelize(list);
    JavaRDD<Row> javaRddRow = javaRddString.map(new CommaSeparatedValueStringToDoubleArrayRow());
    List<StructField> fields = new ArrayList<StructField>();
    fields.add(DataTypes.createStructField("C1", DataTypes.DoubleType, true));
    fields.add(DataTypes.createStructField("C2", DataTypes.DoubleType, true));
    fields.add(DataTypes.createStructField("C3", DataTypes.DoubleType, true));
    StructType schema = DataTypes.createStructType(fields);
    Dataset<Row> dataFrame = spark.createDataFrame(javaRddRow, schema);
    MatrixMetadata mm = new MatrixMetadata(3, 3, 9);
    Script script = pydml("print('sum: ' + sum(M))").in("M", dataFrame, mm);
    setExpectedStdOut("sum: 450.0");
    ml.execute(script);
}

Example 47 with StructField

use of org.apache.spark.sql.types.StructField in project incubator-systemml by apache.

the class MLContextFrameTest method testFrame.

public void testFrame(FrameFormat format, SCRIPT_TYPE script_type, IO_TYPE inputType, IO_TYPE outputType) {
    System.out.println("MLContextTest - Frame JavaRDD<String> for format: " + format + " Script: " + script_type);
    List<String> listA = new ArrayList<String>();
    List<String> listB = new ArrayList<String>();
    FrameMetadata fmA = null, fmB = null;
    Script script = null;
    ValueType[] schemaA = { ValueType.INT, ValueType.STRING, ValueType.DOUBLE, ValueType.BOOLEAN };
    List<ValueType> lschemaA = Arrays.asList(schemaA);
    FrameSchema fschemaA = new FrameSchema(lschemaA);
    ValueType[] schemaB = { ValueType.STRING, ValueType.DOUBLE, ValueType.BOOLEAN };
    List<ValueType> lschemaB = Arrays.asList(schemaB);
    FrameSchema fschemaB = new FrameSchema(lschemaB);
    if (inputType != IO_TYPE.FILE) {
        if (format == FrameFormat.CSV) {
            listA.add("1,Str2,3.0,true");
            listA.add("4,Str5,6.0,false");
            listA.add("7,Str8,9.0,true");
            listB.add("Str12,13.0,true");
            listB.add("Str25,26.0,false");
            fmA = new FrameMetadata(FrameFormat.CSV, fschemaA, 3, 4);
            fmB = new FrameMetadata(FrameFormat.CSV, fschemaB, 2, 3);
        } else if (format == FrameFormat.IJV) {
            listA.add("1 1 1");
            listA.add("1 2 Str2");
            listA.add("1 3 3.0");
            listA.add("1 4 true");
            listA.add("2 1 4");
            listA.add("2 2 Str5");
            listA.add("2 3 6.0");
            listA.add("2 4 false");
            listA.add("3 1 7");
            listA.add("3 2 Str8");
            listA.add("3 3 9.0");
            listA.add("3 4 true");
            listB.add("1 1 Str12");
            listB.add("1 2 13.0");
            listB.add("1 3 true");
            listB.add("2 1 Str25");
            listB.add("2 2 26.0");
            listB.add("2 3 false");
            fmA = new FrameMetadata(FrameFormat.IJV, fschemaA, 3, 4);
            fmB = new FrameMetadata(FrameFormat.IJV, fschemaB, 2, 3);
        }
        JavaRDD<String> javaRDDA = sc.parallelize(listA);
        JavaRDD<String> javaRDDB = sc.parallelize(listB);
        if (inputType == IO_TYPE.DATAFRAME) {
            JavaRDD<Row> javaRddRowA = FrameRDDConverterUtils.csvToRowRDD(sc, javaRDDA, CSV_DELIM, schemaA);
            JavaRDD<Row> javaRddRowB = FrameRDDConverterUtils.csvToRowRDD(sc, javaRDDB, CSV_DELIM, schemaB);
            // Create DataFrame
            StructType dfSchemaA = FrameRDDConverterUtils.convertFrameSchemaToDFSchema(schemaA, false);
            Dataset<Row> dataFrameA = spark.createDataFrame(javaRddRowA, dfSchemaA);
            StructType dfSchemaB = FrameRDDConverterUtils.convertFrameSchemaToDFSchema(schemaB, false);
            Dataset<Row> dataFrameB = spark.createDataFrame(javaRddRowB, dfSchemaB);
            if (script_type == SCRIPT_TYPE.DML)
                script = dml("A[2:3,2:4]=B;C=A[2:3,2:3]").in("A", dataFrameA, fmA).in("B", dataFrameB, fmB).out("A").out("C");
            else if (script_type == SCRIPT_TYPE.PYDML)
                // DO NOT USE ; at the end of any statment, it throws NPE
                script = pydml("A[$X:$Y,$X:$Z]=B\nC=A[$X:$Y,$X:$Y]").in("A", dataFrameA, fmA).in("B", dataFrameB, fmB).in("$X", 1).in("$Y", 3).in("$Z", 4).out("A").out("C");
        } else {
            if (inputType == IO_TYPE.JAVA_RDD_STR_CSV || inputType == IO_TYPE.JAVA_RDD_STR_IJV) {
                if (script_type == SCRIPT_TYPE.DML)
                    script = dml("A[2:3,2:4]=B;C=A[2:3,2:3]").in("A", javaRDDA, fmA).in("B", javaRDDB, fmB).out("A").out("C");
                else if (script_type == SCRIPT_TYPE.PYDML)
                    // DO NOT USE ; at the end of any statment, it throws
                    // NPE
                    script = pydml("A[$X:$Y,$X:$Z]=B\nC=A[$X:$Y,$X:$Y]").in("A", javaRDDA, fmA).in("B", javaRDDB, fmB).in("$X", 1).in("$Y", 3).in("$Z", 4).out("A").out("C");
            } else if (inputType == IO_TYPE.RDD_STR_CSV || inputType == IO_TYPE.RDD_STR_IJV) {
                RDD<String> rddA = JavaRDD.toRDD(javaRDDA);
                RDD<String> rddB = JavaRDD.toRDD(javaRDDB);
                if (script_type == SCRIPT_TYPE.DML)
                    script = dml("A[2:3,2:4]=B;C=A[2:3,2:3]").in("A", rddA, fmA).in("B", rddB, fmB).out("A").out("C");
                else if (script_type == SCRIPT_TYPE.PYDML)
                    // DO NOT USE ; at the end of any statment, it throws
                    // NPE
                    script = pydml("A[$X:$Y,$X:$Z]=B\nC=A[$X:$Y,$X:$Y]").in("A", rddA, fmA).in("B", rddB, fmB).in("$X", 1).in("$Y", 3).in("$Z", 4).out("A").out("C");
            }
        }
    } else {
        // Input type is file
        String fileA = null, fileB = null;
        if (format == FrameFormat.CSV) {
            fileA = baseDirectory + File.separator + "FrameA.csv";
            fileB = baseDirectory + File.separator + "FrameB.csv";
        } else if (format == FrameFormat.IJV) {
            fileA = baseDirectory + File.separator + "FrameA.ijv";
            fileB = baseDirectory + File.separator + "FrameB.ijv";
        }
        if (script_type == SCRIPT_TYPE.DML)
            script = dml("A=read($A); B=read($B);A[2:3,2:4]=B;C=A[2:3,2:3];A[1,1]=234").in("$A", fileA, fmA).in("$B", fileB, fmB).out("A").out("C");
        else if (script_type == SCRIPT_TYPE.PYDML)
            // DO NOT USE ; at the end of any statment, it throws NPE
            script = pydml("A=load($A)\nB=load($B)\nA[$X:$Y,$X:$Z]=B\nC=A[$X:$Y,$X:$Y]").in("$A", fileA).in("$B", fileB).in("$X", 1).in("$Y", 3).in("$Z", 4).out("A").out("C");
    }
    MLResults mlResults = ml.execute(script);
    // Validate output schema
    List<ValueType> lschemaOutA = Arrays.asList(mlResults.getFrameObject("A").getSchema());
    List<ValueType> lschemaOutC = Arrays.asList(mlResults.getFrameObject("C").getSchema());
    Assert.assertEquals(ValueType.INT, lschemaOutA.get(0));
    Assert.assertEquals(ValueType.STRING, lschemaOutA.get(1));
    Assert.assertEquals(ValueType.DOUBLE, lschemaOutA.get(2));
    Assert.assertEquals(ValueType.BOOLEAN, lschemaOutA.get(3));
    Assert.assertEquals(ValueType.STRING, lschemaOutC.get(0));
    Assert.assertEquals(ValueType.DOUBLE, lschemaOutC.get(1));
    if (outputType == IO_TYPE.JAVA_RDD_STR_CSV) {
        JavaRDD<String> javaRDDStringCSVA = mlResults.getJavaRDDStringCSV("A");
        List<String> linesA = javaRDDStringCSVA.collect();
        Assert.assertEquals("1,Str2,3.0,true", linesA.get(0));
        Assert.assertEquals("4,Str12,13.0,true", linesA.get(1));
        Assert.assertEquals("7,Str25,26.0,false", linesA.get(2));
        JavaRDD<String> javaRDDStringCSVC = mlResults.getJavaRDDStringCSV("C");
        List<String> linesC = javaRDDStringCSVC.collect();
        Assert.assertEquals("Str12,13.0", linesC.get(0));
        Assert.assertEquals("Str25,26.0", linesC.get(1));
    } else if (outputType == IO_TYPE.JAVA_RDD_STR_IJV) {
        JavaRDD<String> javaRDDStringIJVA = mlResults.getJavaRDDStringIJV("A");
        List<String> linesA = javaRDDStringIJVA.collect();
        Assert.assertEquals("1 1 1", linesA.get(0));
        Assert.assertEquals("1 2 Str2", linesA.get(1));
        Assert.assertEquals("1 3 3.0", linesA.get(2));
        Assert.assertEquals("1 4 true", linesA.get(3));
        Assert.assertEquals("2 1 4", linesA.get(4));
        Assert.assertEquals("2 2 Str12", linesA.get(5));
        Assert.assertEquals("2 3 13.0", linesA.get(6));
        Assert.assertEquals("2 4 true", linesA.get(7));
        JavaRDD<String> javaRDDStringIJVC = mlResults.getJavaRDDStringIJV("C");
        List<String> linesC = javaRDDStringIJVC.collect();
        Assert.assertEquals("1 1 Str12", linesC.get(0));
        Assert.assertEquals("1 2 13.0", linesC.get(1));
        Assert.assertEquals("2 1 Str25", linesC.get(2));
        Assert.assertEquals("2 2 26.0", linesC.get(3));
    } else if (outputType == IO_TYPE.RDD_STR_CSV) {
        RDD<String> rddStringCSVA = mlResults.getRDDStringCSV("A");
        Iterator<String> iteratorA = rddStringCSVA.toLocalIterator();
        Assert.assertEquals("1,Str2,3.0,true", iteratorA.next());
        Assert.assertEquals("4,Str12,13.0,true", iteratorA.next());
        Assert.assertEquals("7,Str25,26.0,false", iteratorA.next());
        RDD<String> rddStringCSVC = mlResults.getRDDStringCSV("C");
        Iterator<String> iteratorC = rddStringCSVC.toLocalIterator();
        Assert.assertEquals("Str12,13.0", iteratorC.next());
        Assert.assertEquals("Str25,26.0", iteratorC.next());
    } else if (outputType == IO_TYPE.RDD_STR_IJV) {
        RDD<String> rddStringIJVA = mlResults.getRDDStringIJV("A");
        Iterator<String> iteratorA = rddStringIJVA.toLocalIterator();
        Assert.assertEquals("1 1 1", iteratorA.next());
        Assert.assertEquals("1 2 Str2", iteratorA.next());
        Assert.assertEquals("1 3 3.0", iteratorA.next());
        Assert.assertEquals("1 4 true", iteratorA.next());
        Assert.assertEquals("2 1 4", iteratorA.next());
        Assert.assertEquals("2 2 Str12", iteratorA.next());
        Assert.assertEquals("2 3 13.0", iteratorA.next());
        Assert.assertEquals("2 4 true", iteratorA.next());
        Assert.assertEquals("3 1 7", iteratorA.next());
        Assert.assertEquals("3 2 Str25", iteratorA.next());
        Assert.assertEquals("3 3 26.0", iteratorA.next());
        Assert.assertEquals("3 4 false", iteratorA.next());
        RDD<String> rddStringIJVC = mlResults.getRDDStringIJV("C");
        Iterator<String> iteratorC = rddStringIJVC.toLocalIterator();
        Assert.assertEquals("1 1 Str12", iteratorC.next());
        Assert.assertEquals("1 2 13.0", iteratorC.next());
        Assert.assertEquals("2 1 Str25", iteratorC.next());
        Assert.assertEquals("2 2 26.0", iteratorC.next());
    } else if (outputType == IO_TYPE.DATAFRAME) {
        Dataset<Row> dataFrameA = mlResults.getDataFrame("A").drop(RDDConverterUtils.DF_ID_COLUMN);
        StructType dfschemaA = dataFrameA.schema();
        StructField structTypeA = dfschemaA.apply(0);
        Assert.assertEquals(DataTypes.LongType, structTypeA.dataType());
        structTypeA = dfschemaA.apply(1);
        Assert.assertEquals(DataTypes.StringType, structTypeA.dataType());
        structTypeA = dfschemaA.apply(2);
        Assert.assertEquals(DataTypes.DoubleType, structTypeA.dataType());
        structTypeA = dfschemaA.apply(3);
        Assert.assertEquals(DataTypes.BooleanType, structTypeA.dataType());
        List<Row> listAOut = dataFrameA.collectAsList();
        Row row1 = listAOut.get(0);
        Assert.assertEquals("Mismatch with expected value", Long.valueOf(1), row1.get(0));
        Assert.assertEquals("Mismatch with expected value", "Str2", row1.get(1));
        Assert.assertEquals("Mismatch with expected value", 3.0, row1.get(2));
        Assert.assertEquals("Mismatch with expected value", true, row1.get(3));
        Row row2 = listAOut.get(1);
        Assert.assertEquals("Mismatch with expected value", Long.valueOf(4), row2.get(0));
        Assert.assertEquals("Mismatch with expected value", "Str12", row2.get(1));
        Assert.assertEquals("Mismatch with expected value", 13.0, row2.get(2));
        Assert.assertEquals("Mismatch with expected value", true, row2.get(3));
        Dataset<Row> dataFrameC = mlResults.getDataFrame("C").drop(RDDConverterUtils.DF_ID_COLUMN);
        StructType dfschemaC = dataFrameC.schema();
        StructField structTypeC = dfschemaC.apply(0);
        Assert.assertEquals(DataTypes.StringType, structTypeC.dataType());
        structTypeC = dfschemaC.apply(1);
        Assert.assertEquals(DataTypes.DoubleType, structTypeC.dataType());
        List<Row> listCOut = dataFrameC.collectAsList();
        Row row3 = listCOut.get(0);
        Assert.assertEquals("Mismatch with expected value", "Str12", row3.get(0));
        Assert.assertEquals("Mismatch with expected value", 13.0, row3.get(1));
        Row row4 = listCOut.get(1);
        Assert.assertEquals("Mismatch with expected value", "Str25", row4.get(0));
        Assert.assertEquals("Mismatch with expected value", 26.0, row4.get(1));
    } else {
        String[][] frameA = mlResults.getFrameAs2DStringArray("A");
        Assert.assertEquals("Str2", frameA[0][1]);
        Assert.assertEquals("3.0", frameA[0][2]);
        Assert.assertEquals("13.0", frameA[1][2]);
        Assert.assertEquals("true", frameA[1][3]);
        Assert.assertEquals("Str25", frameA[2][1]);
        String[][] frameC = mlResults.getFrameAs2DStringArray("C");
        Assert.assertEquals("Str12", frameC[0][0]);
        Assert.assertEquals("Str25", frameC[1][0]);
        Assert.assertEquals("13.0", frameC[0][1]);
        Assert.assertEquals("26.0", frameC[1][1]);
    }
}

Example 48 with StructField

use of org.apache.spark.sql.types.StructField in project incubator-systemml by apache.

the class FrameRDDConverterUtils method convertFrameSchemaToDFSchema.

/**
 * This function will convert Frame schema into DataFrame schema
 *
 * @param fschema frame schema
 * @param containsID true if contains ID column
 * @return Spark StructType of StructFields representing schema
 */
public static StructType convertFrameSchemaToDFSchema(ValueType[] fschema, boolean containsID) {
    // generate the schema based on the string of schema
    List<StructField> fields = new ArrayList<>();
    // add id column type
    if (containsID)
        fields.add(DataTypes.createStructField(RDDConverterUtils.DF_ID_COLUMN, DataTypes.DoubleType, true));
    // add remaining types
    int col = 1;
    for (ValueType schema : fschema) {
        DataType dt = null;
        switch(schema) {
            case STRING:
                dt = DataTypes.StringType;
                break;
            case DOUBLE:
                dt = DataTypes.DoubleType;
                break;
            case INT:
                dt = DataTypes.LongType;
                break;
            case BOOLEAN:
                dt = DataTypes.BooleanType;
                break;
            default:
                dt = DataTypes.StringType;
                LOG.warn("Using default type String for " + schema.toString());
        }
        fields.add(DataTypes.createStructField("C" + col++, dt, true));
    }
    return DataTypes.createStructType(fields);
}

Example 49 with StructField

use of org.apache.spark.sql.types.StructField in project incubator-systemml by apache.

the class FrameRDDConverterUtils method convertDFSchemaToFrameSchema.

/**
 * NOTE: regarding the support of vector columns, we make the following
 * schema restriction: single vector column, which allows inference of
 * the vector length without data access and covers the common case.
 *
 * @param dfschema schema as StructType
 * @param colnames column names
 * @param fschema array of SystemML ValueTypes
 * @param containsID if true, contains ID column
 * @return 0-based column index of vector column, -1 if no vector.
 */
public static int convertDFSchemaToFrameSchema(StructType dfschema, String[] colnames, ValueType[] fschema, boolean containsID) {
    // basic meta data
    int off = containsID ? 1 : 0;
    boolean containsVect = false;
    int lenVect = fschema.length - (dfschema.fields().length - off) + 1;
    int colVect = -1;
    // process individual columns
    for (int i = off, pos = 0; i < dfschema.fields().length; i++) {
        StructField structType = dfschema.apply(i);
        colnames[pos] = structType.name();
        if (structType.dataType() == DataTypes.DoubleType || structType.dataType() == DataTypes.FloatType)
            fschema[pos++] = ValueType.DOUBLE;
        else if (structType.dataType() == DataTypes.LongType || structType.dataType() == DataTypes.IntegerType)
            fschema[pos++] = ValueType.INT;
        else if (structType.dataType() == DataTypes.BooleanType)
            fschema[pos++] = ValueType.BOOLEAN;
        else if (structType.dataType() instanceof VectorUDT) {
            if (containsVect)
                throw new RuntimeException("Found invalid second vector column.");
            String name = colnames[pos];
            colVect = pos;
            for (int j = 0; j < lenVect; j++) {
                colnames[pos] = name + "v" + j;
                fschema[pos++] = ValueType.DOUBLE;
            }
            containsVect = true;
        } else
            fschema[pos++] = ValueType.STRING;
    }
    return colVect;
}

Example 50 with StructField

use of org.apache.spark.sql.types.StructField in project incubator-systemml by apache.

the class RDDConverterUtils method binaryBlockToDataFrame.

public static Dataset<Row> binaryBlockToDataFrame(SparkSession sparkSession, JavaPairRDD<MatrixIndexes, MatrixBlock> in, MatrixCharacteristics mc, boolean toVector) {
    if (!mc.colsKnown())
        throw new RuntimeException("Number of columns needed to convert binary block to data frame.");
    // slice blocks into rows, align and convert into data frame rows
    JavaRDD<Row> rowsRDD = in.flatMapToPair(new SliceBinaryBlockToRowsFunction(mc.getRowsPerBlock())).groupByKey().map(new ConvertRowBlocksToRows((int) mc.getCols(), mc.getColsPerBlock(), toVector));
    // create data frame schema
    List<StructField> fields = new ArrayList<>();
    fields.add(DataTypes.createStructField(DF_ID_COLUMN, DataTypes.DoubleType, false));
    if (toVector)
        fields.add(DataTypes.createStructField("C1", new VectorUDT(), false));
    else {
        // row
        for (int i = 1; i <= mc.getCols(); i++) fields.add(DataTypes.createStructField("C" + i, DataTypes.DoubleType, false));
    }
    // rdd to data frame conversion
    return sparkSession.createDataFrame(rowsRDD.rdd(), DataTypes.createStructType(fields));
}