Warn about old VLBA data that needs FFT artifact correction

[kettenis]

No description provided.

[kettenis]

Justin Linford from the VLBA told me is happy with the way this would warn (their) users about data that may lack certain corrections when imported into CASA. So I'm removing draft status from this PR.

[kettenis]

@dpetry Any comments on this one? I'll probably have to rebase this PR after #1318 gets merged, but that shouldn't be a big issue.

[dpetry]

Sorry, didn't see this before. Will look at it in detail tomorrow.

[dpetry]

@kettenis , hi, I had a first look. I don't think all warnings should be reduced to normal infos.
I think those messages which are about deprecated or incorrectly used FITS-IDI features and similar, should remain
warnings or at least have some string in them which people can grep for.
But I'll give a complete review tomorrow ..

[dpetry]

I agree that all the "treating as scalar" warnings can be turned to INFO.
Also the Double should be Float can be INFO.
But the "not yet implemented" and "will ignore it" messages should stay
warnings such that people realise that some of the input is not making it
into the MS.

[kettenis]

@dpetry thanks for the review. Yes maybe I went a little bit too far in demoting the warnings got standard FITS-IDI tables that we don't handle yet. For the BASELINE, BANDPASS and CALIBRATION tables that is not an issue, since those tables are not present in typical archival data. I'm not even sure if code exists to produce these. For the INTERFEROMETER_MODEL your concern is certainly justified. Even though there is no code in CASA to use the correlator model, there is general consensus that this data should be added to the MS. A new table was proposed as part of MSv3 and is still on the table for MSv4.

I've updated the pull request and also rebased it against the current master.

[dpetry]

@kettenis , this looks good except for the very last message about the unconventional extension.
IMHO that should remain a warning since people should be told if their input
contains something which cannot be interpreted and is ignored.
(Sorry if I'm delaying you.)

[dpetry]

@kettenis , hi Mark, did you see my comment above? Please just turn that last message back into a warning and I approve.

[kettenis]

@dpetry
Hi Dirk,

Unfortunately the VLBA data tends to have a few non-standard tables attached to it. Some of those are empty (no rows) or really are of no interest to the modern user (TAPE_STATISTICS). And these tables tend to come at the end so the warning you're talking about tends to come at the end of the output the user sees. The VLBA stakeholders are worried that this means users will miss the warnings.

How about printing that warning only if the number of rows in the table is greater than zero and exclude the TAPE_STATISTICS table?

      if (extname != "TAPE_STATISTICS" && nrows() > 0) {
        *itsLog << LogIO::WARN << "Extension " << extname
                << " not part of the FITS-IDI convention. Will ignore it." << LogIO::POST;
      }

[dpetry]

@kettenis , yes, that's a great idea. Please go ahead with this.

[kettenis]

@dpetry Thanks, done!

[dpetry]

Approved.