TAP Implementation Issues (cont'd)
Tom McGlynn
Thomas.A.McGlynn at nasa.gov
Tue Oct 27 11:19:03 PDT 2009
I've been playing around with PgSphere and such getting the ADQL
geometry constructs to work. That's OK and now it's time to see how to
put it into TAP. A few encoding issues have cropped up...
The table in section 2.5 defines the XTYPEs to be used for POINT and
REGION objects, but does not mention CIRCLE, BOX or POLYGON objects. As
far as I can tell those types are as well defined as the first two. Is
this just an oversight, or is there some difference in the level of
support between these two lists of geometric types?
ADQL also defines a function coordsys which extracts the coordinate
system from a geometric object. Is this simply a string or is it a
REGION, i.e., should we specify xtype='adql:REGION' for columns where it
is output or not. Is there a suggested UCD or UTYPE that should be
given if I generate a column using this function?
The ADQL coord[12] functions extract a field from a point object. If I
create a column using one of these is there an appropriate UCD or UTYPE
that I should use. Is there a mapping from the coordinate system of the
point to the UCDs that I should use?
The xtypes for regions are of the form
xtype='adql:spec'
which suggests that there is an ADQL name space that will need to be
defined (I think that if I use that syntax I need the namespace defined
somewhere). What is the definition of that namespace?
Where is timestamp required to be supported (i.e., the ISO input format
requirements)? It looks like there are a lot of time-related datatypes
in the various databases. Is it only if the internal SQL datatype is
'timestamp' that we need worry about the requirement for ISO format
input? A quick gander suggested that MS SQLServer doesn't have such a
type (well it has one called timestamp but it has nothing to do with
dates so far as I can tell). Since the HEASARC uses MJD's I'm going to
ignore this requirement in detail. In practice for PQL we support MJD,
JD and ISO format data entry for all of our fields where our unit's
metadata indicates that the field is a date. That will be harder for
ADQL. The actual datatypes are either double or int.
Tom McGlynn
More information about the dal
mailing list