WD-TAP-1.1 : tap_schema and VOSI-tables

[Mark Taylor]

On Tue, 4 Oct 2016, Markus Demleitner wrote:

> Hi DAL,
> 
> On Mon, Oct 03, 2016 at 03:28:49PM -0700, Patrick Dowler wrote:
> > VODataService would be unneeded/obsolete. The whole datatype mapping
> > and handling would be much simpler and direct. I can't see any
> > downside to this aside from it being different from then ad-hoc things
> > we did in 1.0 :-)
> > 
> > * proposal *
> > So, although it would delay implementations somewhat, I propose that
> > we add an xtype column to tap_schema.columns and use the data_type
> > column for the underlying primitive type in TAP-1.1
> 
> From a server-side perspective: I'm all for it; it clearly should
> have done like this from the start.
> 
> But then I suppose by the letter of the law, it would be an
> incompatible change, because clients that so far have been looking
> for adql:point (or whatever) will be finding char (or something like
> that).  I wonder if clients so far actually do that.  I suppose
> TOPCAT looks at types to try and produce meaningful examples (does
> it?).  Is anyone else out there using TAP_SCHEMA datatypes in a way
> that'd break with this change?

Not me.  Except in taplint it looks like I just treat the values
from VODataService vs:TableDataType and from TAP_SCHEMA.columns
datatype as opaque strings.  Such a change would no doubt present
a few headaches in taplint, and I don't guarantee to have spotted
all possible flaws in such a plan, but I don't object.

--
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b.taylor at bris.ac.uk +44-117-9288776  http://www.star.bris.ac.uk/~mbt/