Updates for xija_gui_fit

[jzuhone]

Description

This PR fixes a number of annoying bugs in xija_gui_fit and refactors the design of a few features.

• The console access has been removed. It was an experimental feature that never worked very well anyway, and getting it to work to design scripting of the interface was proving intractable.
• Improved the handling of "bad times" which one wants to store in the JSON file and the "mask times", which includes the bad times but also includes times that are masked out temporarily in the GUI for fitting.
• Added a splitter between the plots window and the parameter panel so you can adjust the horizontal size of the two parts
• Fixed some issues with the scaling of the x-axis of plots, especially when adding/deleting. This was partially achieved by adding a dummy Figure, Axes pair that is never displayed but also never goes away and is used as the shared axes.
• The command entry box, used for freezing and thawing parameters as well as adding times to mask and removing the mask times, has been removed, and the functionalities have been separated out. Now, freeze and thaw entry boxes appear above the parameters panel. A new button for "filters" appears on the left panel (see image below), which handles adding mask times. To freeze or thaw a parameter, enter the regular expression into either box (e.g., solarheat*__P*) and hit Enter.
• General reorganization of the buttons on the left panel.
• The ability to add a "bad time" to the model JSON file from the GUI has been added and is part of the "filters" panel.

The new look of the window:

The new "Filters" subwindow, showing how to add a mask and ignore all masks, and how to add a "bad time":

Testing

• [ ] Passes unit tests on MacOS, linux, Windows (at least one required)
• [x] Functional testing

[jzuhone]

@taldcroft likely the only FSDS-relevant bits are in https://github.com/sot/xija/pull/116/files#diff-78ad09ad6fe546ab9d4b1d835063650634f425cff3b09df8417a13be438d6838

[taldcroft]

@jzuhone - awesome work glad to see this tool getting continued attention.

I think this will fall in the category of a package update that tags along in a normal Ska FSDS request. It wasn't obvious in 5 seconds what is going on with the diff you highlighted above, but from FSDS perspective the key is to highlight interface impacts at the top level (in the PR description) and make sure in advance that likely users are aware. That pretty much means ACIS ops, @matthewdahmer and occasionally me.

When we make the Ska update request then this can be highlighted if necessary as a notable change.

[jzuhone]

@taldcroft the change in the diff I linked should be a no-op, it's just that it's the only part that could conceivably affect model evaluation if I screwed something up.

[taldcroft]

@jzuhone - I gave this a spin in ska3-next and it works, but I do see a bunch of new warnings:

The process has forked and you cannot use this CoreFoundation functionality safely. You MUST exec().
Break on __THE_PROCESS_HAS_FORKED_AND_YOU_CANNOT_USE_THIS_COREFOUNDATION_FUNCTIONALITY___YOU_MUST_EXEC__() to debug.

@javierggt will probably have a new 2022.2rc4 release out shortly that includes https://github.com/sot/Ska.Matplotlib/pull/24, which changes the default plot_cxctime color (discussed previously on slack). Since this PR won't go in until after 2022.2 (https://github.com/sot/skare3/pull/774) you might as well start testing there.

[taldcroft]

One feature request (and maybe it's already there?) is to be able to get any flight model with the name, e.g.

xija_gui_fit aca --stop=2022:001 --days=200 --set-data='aca0=-10'

This would grab the latest flight model from chandra_models with xija_get_model_spec for any model_name that doesn't end in .json. Or something like that.

[jzuhone]

@taldcroft

when are you getting this warning?

The process has forked and you cannot use this CoreFoundation functionality safely. You MUST exec().
Break on __THE_PROCESS_HAS_FORKED_AND_YOU_CANNOT_USE_THIS_COREFOUNDATION_FUNCTIONALITY___YOU_MUST_EXEC__() to debug.

I don't see it.

[taldcroft]

Huh, I tried again and I don't see that now. Well, never mind for the moment.

[matthewdahmer]

@taldcroft @jzuhone I saw that same error, "__THE_PROCESS_HAS_FORKED...", once on my Apple Silicon Mac. During that instance, once I saw it I was not able to get xija_gui_fit to run until I rebooted, then it worked fine.

[jzuhone]

Weird. Anyway, I added the new feature @taldcroft requested. It also prompts if you have an unsaved model when you hit the Quit button.

[taldcroft]

I added the new feature @taldcroft requested. It also prompts if you have an unsaved model when you hit the Quit button. Nice! I also like the new Thaw/Freeze boxes, when I was using the previous version I had to think for a few seconds to remember the right words.

One more feature request - put the xija version into the App title? In testing just now I had been accidentally getting the wrong version so it would be nice to see the version.

[jzuhone]

Ok to merge this before things get too busy?

[taldcroft]

Fine by me to merge, but @matthewdahmer - are you good with this?

[matthewdahmer]

@jzuhone @taldcroft I am OK with this update. I tested it on my Mac with a PLINE refit and seems to work well!

[taldcroft]

BTW @jzuhone have you tested this with ska3-next on supported platforms?

I guess that isn't a lien on merging this since we can always come back with fixes as needed. My light testing was OK on that but it's worth noting that @javierggt experienced troubles with aca_view (another Qt app) and required export QT_MAC_WANTS_LAYER=1 to get it to work. (It is slightly perplexing that xija_gui_fit works without that).

[jzuhone]

@taldcroft I have not yet tested against ska3-next--is there a documented way yet to set up a ska3-next environment?

[taldcroft]

@jzuhone - the conda command under Testing in https://github.com/sot/skare3/pull/774 should work.

[jzuhone]

@taldcroft I tested under ska3-next and all appears to work fine.

[taldcroft]

👍

[jzuhone]

@taldcroft what's your position on who should merge this?