Fix parquet io

[kthyng]

Looks like the parquet exporter wasn't updated when output changes happened with OpenDrift. This is my attempt to fix it. As a disclaimer, I wrote these changes using ChatGPT because I was not familiar enough with parquet nor the details of the original code to get it right. I tested it with OpenDrift's tests and my own subsidiary package tests and it looks like it is working but I encourage an extra eye on this if possible.

[knutfrode]

The lines 2230-2239 seem a bit overly complicated. Could this not be simply:

if outfile is not None:
    ext = os.path.splitext(outfile)[1].lower()
    if ext == ".parquet":
         pass  # let the parquet exporter set self.result in its close()
    else:
         # reimport from file to get the complete simulation (lazily) in memory
         self.result = xr.open_dataset(outfile)  

[knutfrode]

Also I am note completely sure about replacing export_buffer_length is None with outfile is None,
as one could have an outfile but still set export_buffer_length to None to get just a single write at the end.
In this case it would be unnecessary to reimport from that file, as the complete simulation is already in memory.
On the other hand, it might perhaps be ok for consistency.

[knutfrode]

The deletion of self.environment and self.environment_profiles has here been removed. Tests are passing, so might be ok, but I seem to recall that this deletion was necessary for some reason. Was this removed deliberately, or was it just ChatGPT not seing the point?

[knutfrode]

Hi,
Apart from the small concerns regarding the updates to basemodel (see comments above), this seems fine to me.

However, I have never used the parquet writer, so perhaps @poplarShift or @mateuszmatu would be interested to have a look before merging this?
If no comments until tomorrow, I guess we can just merge it, and they can have a look later.
The writer is anyway not functional at the moment.

[mateuszmatu]

I have not looked more into this, as I also prefer netCDF. Maybe @poplarShift can update this?

[kthyng]

@knutfrode Thanks for your comments. I can integrate them later today or Monday in case we hear from @poplarShift first.

[poplarShift]

Hi all, sorry for the silence on this.
I've been a bit hesitant on whether this I/O module should simply be discontinued (even though I'd quite like the functionality) and just serve as inspiration/for future reference.

Apart from the maintenance thing there's also the mysterious issue #1321 which I still haven't figured out.
Maybe for now the most pragmatic thing is to take it out again?

[knutfrode]

But then I guess that if Kristen can make it work again, then we could keep it, and it will still serve as inspiration for others for future development (or generalisation).