Backwards support of OBS_MODE for GADF v0.2

Gammapy version 1.0rc1

Bug description obs.fixed_pointing_info.mode crashes with ValueError if the pointing mode stored in the FITS data file is not equal to “WOBBLE”, “POINTING” or “DRIFT”. This is coded in PointingMode.from_gadf_string. This in turn blocks the data reduction.

This is not a good behavior and not even compatible with GADF, because GADF does not impose any list of acceptable OBS_MODE meta data (see here).

Example (using H.E.S.S. data):

[registerrier]

Currently, we support only 3 cases: POINTING, DRIFT and WOBBLE. The latter case is there to stay compliant with existing data releases (in particular from the joint Crab dataset) but is not listed among the possible keywords in gadf v0.3.

All in all, if we are to stay consistent with the gadf formats, we could:

• Check the HDUVERS and
  • if v<=0.2, set the FixedPointingInfo.mode to the exact OBS_MODE
  • else use the current expression (without the WOBBLE option)
• in the MapDatasetMaker.make_meta_table remove the explicit test to PointingMode.POINTING as imposing one or the other is inconsistent with what is stated in gadf v0.2 that OBS_MODE is irrelevant for data reduction

[adonath]

Yes, I see. I checked and it seems indeed in v0.2 basically arbitrary values are allowed (which is very weird BTW…). This means from the Gammapy side we have to check the GADF version?