patch for geometry type

[croquiscom]

for issue #299

I made a patch and test for geometry type.

There is a strange point.
For insertion, I give a buffer of WKBPoint type (1 byte for byte-order, 4 byte for type, 8 byte for x, 8 byte for y)
But when I fetch it from the database, I get four more bytes of zero in front of the buffer.

I can not find any information about this in the MySQL packet documentation.
(is it right to return first-four-bytes-removed buffer? hmm..)

[croquiscom]

I thought about returning an object instead of a buffer. (like Date type)
But I can not determine which one is better, [x,y] or {x:x, y:y}, so I let it be.

By the way, if I do so and someone does not want converted object, then someone may use { typeCast: false } option.
But { typeCast: false } option give a string always.

How about check whether it is a binary or not when typeCast:false?

// in RowDataPacket.prototype.parse

var value = (typeCast)
  ? this._typeCast(fieldPacket, parser)
  : (field.charsetNr === Charsets.BINARY
    ? parser.parseLengthCodedBuffer()
    : parser.parseLengthCodedString());

[felixge]

A few thoughts:

• There is a type casting test that has all kinds of types in it, can you add your test case in there, rather than adding a new one?
• IMO casting into {x: x, y: y} seems like a great default - please do that
• Type casting into buffers when typeCast: false seems like a great idea for binary fields as well as this geometry stuff you have here.

Let me know what you think! Thanks for the help!

[croquiscom]

I worked as you wrote.

One problem of returning object is that the client cannot distinguish an object is a linestring or a multipoint,
though it may not be a problem for most cases because the client may know the data type of the column.

[felixge]

Ok, I'm not perfectly happy with your implementation as you're essentially doing stuff manually that is supposed to be handled by the parser class (incrementing the internal buffer offset, providing high level parse functions).

But that being said ... it doesn't interfere with anything I care about and the test is good - so I'm merging. If you want to extend the parser to include the float/other types you need and improve the code ... that would still be seriously kick ass : ).