OIFITS version

[jsy1001]

We should clarify which version(s) of the OIFITS standard applies to the files used by OImaging. Does it depend on the version of the OIFITS data given as input?

[GillesDuvert]

@jsy1001 could you elaborate a bit, I have no clear example, except that, if ever, pray God, the correlation matrix was used, the IR softwares should use it?

[jsy1001]

I believe that OImaging will accept either version 1 or 2 OIFITS files. The same is true of BSMEM, MiRA, and WiSARD (I think). Do we care whether the image reconstruction backends write out OIFITS v2 when given OIFITS v1 as input (or perhaps the reverse) ? Or should OImaging convert v1 to v2?

At the moment I think v1 inputs are not converted to v2 anywhere, but we should confirm this.

[GillesDuvert]

Ah, I see... I guess WISARD writes more or less what it got, plus the extra model columns, so it should be v2 out if v2 in. never really tested it. Having to deal with those peculiarities would be better left outside the scope of the IR backends... until we face a problem.

[FerreolS]

I agree with @GillesDuvert , up to now all software seems to accept both version as input. I propose to postpone this issue until we face it (e.g. when a software impose v2 input). We will probably be able to make patch in the GUI without bothering software people. Probably @GillesDuvert will be retired for a long time if it happens ;-)

[emmt]

Since OIFITS-V2 is somewhat backward compatible, we could also read any version and write V2 when saving? I think this would be the simplest way to handle cases where mutilple files are provided that mix different versions of OI-FITS.

[jsy1001]

I think that would be a good solution.

There isn't a way to specify multiple files in OImaging yet, is there?

[emmt]

I don't think so.

[GillesDuvert]

The "merging" of various OIFITS files is better done within OIFitsExplorer, so I would let the job of merge and conversion to V2 to it. The OIFitsExplorer feature doing the merge can be plugged into OImaging if people do not want to use OIE beforehand.

[emmt]

I agree, what I proposed was just a sensitive behavior of the image reconstruction algorithm when directly used from the command line.

[FerreolS]

As @GillesDuvert I think that the merging / filtering of OIFITS should be done with OIFitsExplorer .

If software just add keywords and HDU to the OIFITS file without suppressing anything else as mentioned in #12 , then it will be blind to the OIFITS version. The output will have the same version as the input.

If V2 seems mandatory, then we will probably be able to perform conversion in OImaging by the mean of https://github.com/JMMC-OpenDev/OITools

[jsy1001]

BSMEM now always writes valid OIFITS v2 as output, regardless of whether the input is v1 or v2.