Decide on filename convention / mimic filenames that correspond to the actual raw data.

[hugobuddel]

Currently the output only contains the CATG:

• DARK_LM_RAW.fits

The filenames of real raw data will probably look like this:

• METIS.2024-02-29T01:23:45.678.fits

Where the timestamp is the MJD-OBS. This has the benefit that each file is unique.

We should not follow the official scheme, because it has certain drawbacks:

• It is impossible to have a colon in a filename on windows. (But who uses that anyway? :innocent: )
• We can't see at a glance what the file is for, because the CATG isn't there.

I therefore propose this scheme:

• METIS.2024-02-29T01_23_45.678.DARK_LM_RAW.fits

The reason to put the CATG value last is that the recipes will put the filename of the input data in the headers of the produced data, but it will crop this to only a small number of characters. So it is necessary to put the MJD-OBS first to ensure that the headers of the processed data uniquely point to the used raw data.

Note that for the LMN mode and IFU modes we might have observations with technically the exact same time of observation, if the LM detector is read out at the exact same time as the N or IFU detector. In that case we can just add a millisecond to one of them.

[hugobuddel]

The above is implemented in #7.

This works, however, we perhaps use too many ambiguous separators with filenames like METIS.2024-01-02T03_45_00.123.DETLIN_LM_RAW-dit1.0.fits:

• . to separate the base filename and the extension
• . between the ESO parts of the filename and our part (METIS, mjdobs, and our part)
• T to separate date and time
• - to separate year/month/day
• _ to separate hour/minute/second
• . to separate second/millisecond
• - to separate CATG and parameters
• _ to separate components of CATG
• `` (nothing) to separate parameter name and value
• . to separate integer and fractional part of values

I propose to keep this messy and just proceed, until someone thinks of something nicer.