CREATE TYPE doesn't recognize complex/complexd cell type

[Dimitar Misev]

rasql -q 'CREATE TYPE test_2d_complex_mdd AS complex MDARRAY [d0, d1]'
Executing update query... rasdaman error 300: Parsing error 300 in line 1, column 36: Unexpected name complex.

rasql -q 'CREATE TYPE test_2d_complex_mdd AS complexd MDARRAY [d0, d1]'
Executing update query... rasdaman error 971: Update error 971 in line 1, column 1, near token complexd: Invalid cell type.

Furthermore, complex/complexd isn't recognized as a cast cell type.

[Dimitar Misev]

complex is already a command token for the complex constructor, so it's hard to reuse as a type token as well. In rasdl it's no problem as it's a separate grammar.

We could use something similar to gdal for example, which makes things clearer as well (​https://www.gdal.org/gdal_8h.html#a22e22ce0a55036a96f652765793fb7a4)

CInt16
CInt32
CFloat32
CFloat64

Some data are also in CInt16 (one version of Sentinel 1 at least), related ticket: #1095

[Peter Baumann]

complex/complexd was an artificial token anyway, no such thing existed - hence, I see no problem in aligning it with relevant practice, such as GDAL.

[Dimitar Misev]

I think at the same time then it will be good to introduce aliases for the other types, e.g.

• octet = int8
• char = uint8
• short = int16
• …

These are much clearer and similarly aligned with the state of the art in ​C/C++, ​GDAL, ​numpy, etc.

[Peter Baumann]

there were times when commonly accepted practice said "avoid being explicit about #bits as it may change as haw advances". Probably sizings have become stable over the years, and alignment with current practice (while retaining the old names) definitely is an asset.