UVFITS rdate question

[telegraphic]

Hi all, wondering whether this UVFITS code:

        if (self.rdate is None) or (self.rdate == ""):
            rdate_obj = Time(np.floor(self.time_array[0]), format="jd", scale="utc")

Should account for the 0.5 Julian Date fraction offset due to UT1? E.g.

        if (self.rdate is None) or (self.rdate == ""):
            rdate_obj = Time(np.floor(self.time_array[0]) + 0.5, format="jd", scale="utc")

[kartographer]

Hey @telegraphic! The astropy Time object handles (or at least, should handle) the JD "noon offset" under the hood when converting to Gregorian dates. But I did do a quick check to make sure that was the case:

In [1]: from astropy.time import Time

In [2]: Time(2456789.0, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[2]: '2014-05-11'

In [3]: Time(2456789.49, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[3]: '2014-05-11'

In [4]: Time(2456789.5, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[4]: '2014-05-12'

In [5]: Time(2456789.51, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[5]: '2014-05-12'

Looks like it behaves as expected to me! 😄

[telegraphic]

@kartographer agreed, but notice there's an errant np.floor in the code:

In [4]: Time(np.floor(2456789.5), format='jd',scale='utc').strftime("%Y-%m-%d")
Out[4]: '2014-05-11'

In [4]: Time(2456789.5, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[4]: '2014-05-12'

In [5]: Time(np.floor(2456789.51), format='jd',scale='utc').strftime("%Y-%m-%d")
Out[5]: '2014-05-11'

In [5]: Time(2456789.51, format='jd',scale='utc').strftime("%Y-%m-%d")
Out[5]: '2014-05-12'

So I think my suggestion is: remove np.floor?

I have now also noticed that DATE-OBS in the AIPS 117 memo is YYYY-MM-DD, and pyuvdata is currently writing the full ISOT:

        hdu.header["DATE-OBS"] = Time(self.time_array[0], scale="utc", format="jd").isot

which I think should be changed to:

        hdu.header["DATE-OBS"] = Time(self.time_array[0], scale="utc", format="jd").strftime("%Y-%m-%d")

I can confirm that MIRIAD is confused when importing UVFITS files, as it uses both DATE-OBS and RDATE

[kartographer]

Ahh, I totally missed the call to np.floor -- yes, that's definitely a bug that should be fixed.

[telegraphic]

Thanks @kartographer , I wasn't sure if there was a motivation for it, as it seemed rather deliberate! I've issued a PR from my branch. The PR #1436 did introduce a unit test failure with DUT1 now slightly different in one test -- not sure if the new value or golden value is off.