[AstroPy] [astropy-dev] Coordinates subpackage - request for help

Demitri Muna demitri.muna@gmail....
Thu Nov 15 10:05:48 CST 2012


Chris, thanks for the script! A few modifications need to be made though (see comments below).

On Nov 15, 2012, at 3:29 AM, David Berry <d.berry@jach.hawaii.edu> wrote:

> 1) Different conversion algorithms are always (or nearly always) going
> to produce different answers. So do you have a target accuracy for the
> astropy coordinates package? Do you want arcseconds accuracy or
> milli-arcsecond accuracy? The more accuracy you want, the more
> sophisticated the algorithms need to be.

To first order I'd like to see where our accuracy is and go from there. I think for a first release arcsec accuracy is probably good enough (to get this thing out the door), with the aim of better accuracy as an improvement. But let's see where we are.

> 2) Why does it need to be IDL? It's proprietary software, so many
> people will not use it. There are many tools in many freely available
> languages for creating tables of corresponding positions in different
> coordinate systems.   For instance, I could create such a table
> quickly using PyAST, which since we're talking Python seems an obvious
> choice.

This is purely for testing purposes and not meant to be released for use by anyone. I specifically chose IDL since the libraries are well-trusted and well-tested by the community. I think a selling feature would be to show that our code compares favorably with what people trust.

> 3) You include ecliptic in the list. Does the coordinates package
> support ecliptic? I could not see it in the code.

Erik implemented the actual systems; I'll let him address that.

> 4) Your list of coordinate systems starts at " ra/dec J2000" and ends
> at "equatorial J2000" - are you making a distinction between these?
> And I presume you mean FK5 J2000 (there is a common usage of "J2000
> RA/Dec" which uses the mean dynamical equator and equinox of the J2000
> epoch as reference, which is different to FK5)?

Admittedly, the list was hastily written. Basically any routines for any systems available, we want to test.

> 5) Was the omission of ICRS intentional?

See above.

> 6) No mention of the epoch of observation. As I mentioned in another
> message you need to specify the epoch of observation, to be able to
> convert accurately to and from an FK4 RA/Dec system (which I presume
> is what you mean by "B1950"). Strictly, you also need the epoch of
> observation when converting between FK5 and ICRS, albeit the
> dependency on the epoch is much weaker.

> 7) You need to specify an equinox for the ecliptic coordinates.

You are correct (see "hastily written" :). The request should have included a note to randomly generate any additional information needed.

As for the output, it would be ideal to generate the random points and save that as a file. Another script could then convert those points into various systems and either write out each result as a separate file or convert from one system to another, and write the results out to a new file.

Thanks again for the help!


Demitri Muna

Center for Cosmology and Particle Physics
New York University


More information about the AstroPy mailing list