Keep M coordinates when writing to WKB and reading from WKB

[kaiwinter]

The changes I made (adding ordinateFlags) are mainly taken over from the WKTReader/WKTWriter.

[kaiwinter]

Are there still open issues or may this PR be merged?

[oers]

Any status on this?

[oers]

I had to fix this in WKBWriter, to make it work with XYZM

    private void writeCoordinate(CoordinateSequence seq, int index, OutStream os)
            throws IOException
    {
        ByteOrderValues.putDouble(seq.getX(index), buf, byteOrder);
        os.write(buf, 8);
        ByteOrderValues.putDouble(seq.getY(index), buf, byteOrder);
        os.write(buf, 8);

        // only write 3rd dim if caller has requested it for this writer
        if (outputDimension >= 3) {
            // if 3rd dim is requested, only write it if the CoordinateSequence provides it
            double ordVal = Coordinate.NULL_ORDINATE;
            if (seq.getDimension() == 3)
                ordVal = seq.getOrdinate(index, 2);
            ByteOrderValues.putDouble(ordVal, buf, byteOrder);
            os.write(buf, 8);
        }
        // only write 4th dim if caller has requested it for this writer
        if (outputDimension == 4) {
            // if 4th dim is requested, only write it if the CoordinateSequence provides it
            double ordVal = Coordinate.NULL_ORDINATE;
            if (seq.getDimension() == 4)
                ordVal = seq.getOrdinate(index, 3);
            ByteOrderValues.putDouble(ordVal, buf, byteOrder);
            os.write(buf, 8);
        }
    }

Test for XYZM is missing here

[kaiwinter]

@oers Does the WKTWriter handle XYZM coordination correctly? If not this should better be fixed together in a new pull request maybe. If it does I could add your fix to this pull request.

[oers]

@kaiwinter yes, the textwriter works correctly. I tested both with XYZM Coordinates (reading and writing)

[kaiwinter]

@oers I don't understand the implementation tbh. I tried to adapt an existing unit test but couldn't even get the WKTReader to parse a XYZM coordinate. So I couldn't verify what your change does exactly. Maybe this should better be an own issue and pull request? What does @dr-jts think?

[oers]

The wrong method in WKB Writer never wrote the 4th dimension. My Method fixes this (the new one below. It still had a bug)

    private void writeCoordinate(CoordinateSequence seq, int index, OutStream os)
            throws IOException
    {
        ByteOrderValues.putDouble(seq.getX(index), buf, byteOrder);
        os.write(buf, 8);
        ByteOrderValues.putDouble(seq.getY(index), buf, byteOrder);
        os.write(buf, 8);

        // only write 3rd dim if caller has requested it for this writer
        if (outputDimension >= 3) {
            // if 3rd dim is requested, only write it if the CoordinateSequence provides it
            double ordVal = seq.getOrdinate(index, 2);
            ByteOrderValues.putDouble(ordVal, buf, byteOrder);
            os.write(buf, 8);
        }
        // only write 4th dim if caller has requested it for this writer
        if (outputDimension == 4) {
            // if 4th dim is requested, only write it if the CoordinateSequence provides it
            double ordVal = seq.getOrdinate(index, 3);
            ByteOrderValues.putDouble(ordVal, buf, byteOrder);
            os.write(buf, 8);
        }
    }

My Test is like this:

    @Test
    void wkbLineStringZM() throws ParseException {
        LineString lineZM = fac.createLineString(new Coordinate[]{new CoordinateXYZM(1,2,3,4), new CoordinateXYZM(5,6,7,8)});
        byte[] write = new WKBWriter(4).write(lineZM);


        LineString deserialisiert = (LineString) new PostgisWKBReader().read(write);

        assertEquals(lineZM, deserialisiert);

        for (int i = 0; i < lineZM.getNumPoints() ; i++) {
            Point geometryN = lineZM.getPointN(1);
            Point geometryN1 = deserialisiert.getPointN(1);

            assertFalse(Double.isNaN(geometryN.getCoordinate().getM()));
            assertFalse(Double.isNaN(geometryN.getCoordinate().getZ()));
            assertFalse(Double.isNaN(geometryN1.getCoordinate().getM()), "M is persisted");
            assertFalse(Double.isNaN(geometryN1.getCoordinate().getZ()), "Z is persisted");

            assertEquals(geometryN, geometryN1);
        }
    }

    @Test
    void wktLineStringZM() throws ParseException {
        LineString lineZM = fac.createLineString(new Coordinate[]{new CoordinateXYZM(1,2,3,4), new CoordinateXYZM(5,6,7,8)});
        String write = new WKTWriter(4).write(lineZM);

        LineString deserialisiert = (LineString) new PostgisWKTReader().read(write);

        assertEquals(lineZM, deserialisiert);

        for (int i = 0; i < lineZM.getNumPoints() ; i++) {
            Point geometryN = lineZM.getPointN(1);
            Point geometryN1 = deserialisiert.getPointN(1);

            assertFalse(Double.isNaN(geometryN.getCoordinate().getM()));
            assertFalse(Double.isNaN(geometryN.getCoordinate().getZ()));
            assertFalse(Double.isNaN(geometryN1.getCoordinate().getM()), "M is persisted");
            assertFalse(Double.isNaN(geometryN1.getCoordinate().getZ()), "Z is persisted");

            assertEquals(geometryN, geometryN1);
        }
    }

[kaiwinter]

@oers I added your changes to this pull request.

[oers]

I integrated the changes made here in a project that uses postgis 14. Itworks really well.

[jiayuasu]

@kaiwinter @dr-jts Thanks for the great work. Is there a reason why this PR hasn't been merged after a year? Is there anything others could help?

WKB format is often used for geometry serialization. Silently missing the M value will lead to incorrect results. We at Sedona uses WKB to serialize XYZM data for our Snowflake support.