script plots

[astrofle]

Enable halting a script to bring up a plotter

[teuben]

Definitely working better. I have a good test script that emulates the EDGE HI pipeline graphics, but it has a few failing options. Will report back after I confirm if it's a pilot error.

One thing that it still reports:

Not running on IPython and trying to use the ShellGUI may result in unexpected behavior.

[astrofle]

Definitely working better. I have a good test script that emulates the EDGE HI pipeline graphics, but it has a few failing options. Will report back after I confirm if it's a pilot error.

One thing that it still reports:

Not running on IPython and trying to use the ShellGUI may result in unexpected behavior.

That is a reminder to you that we don't support Spyder ;)

[teuben]

Definitely working better. I have a good test script that emulates the EDGE HI pipeline graphics, but it has a few failing options. Will report back after I confirm if it's a pilot error.
One thing that it still reports:
Not running on IPython and trying to use the ShellGUI may result in unexpected behavior.

That is a reminder to you that we don't support Spyder ;)

I know, and my complaint about that IPython message was from a script, not via ipython.

On a more "serious" note. We don't have an example script that makes an example of scripting. I do have the edge pipeline script, and I've created a "10 liner" out of this, which will run through this. The features I would recommend we have in such an example pipeline script are:

• should do dysh and optional matplotlib plotting
• should run in batch mode (presumably where one creates png's) as well as interactive mode
• keeps all plots on the screen. so far I had some case where if I killed a window, it would kill others. This was not the case in the plotting infrastructure we had before and I wonder if that's a new feature of the new blocking script.

Should we add such a script to notebooks/developer, or is there another better location?

[teuben]

As i'm using the p.show(block=True) i realized for plt.show() the default is block=True. Is there a reason why you wanted block=False to be the default?

[etsmit]

I'm not able to use the callbacks in this mode, for example get_selected_regions(). For example, in the use case of

1. cycle through 10 spectra
2. for each, plot the spectrum, interactively select good regions, then
3. auto-take out a 1st order baseline and maybe add it to a buffer for averaging or write it back to disk

This fails because the script gets paused on the plot, where I can select the regions, but in order to get to the next step in the script (ps.baseline(..., include=ps.get_selected_regions())), there is no more ps._plotter to reference.

p.s. I'm putting this down because it's a likely use case for whenever I get around to making a receiver checkout script.

[etsmit]

See PR 1034 for a potential fix idea.

[astrofle]

As i'm using the p.show(block=True) i realized for plt.show() the default is block=True. Is there a reason why you wanted block=False to be the default?

I was thinking that maybe matplotlib users will be used to calling show, even in the dysh shell, and wanted to avoid blocking unless explicitly requested. We can change it if you think it makes more sense to be 1-to-1 with matplotlib.

[teuben]

I wrote a simple script, mimicking what the edge pipeline does: bring up two dysh plots, and bring up 2 mpl plots. They all 4 stay up, for simple interactions (pan & zoom, measure etc.) and can be killed one by one. It has the --batch option where png's are created and the script should be able run headless.

It works with the current blocking-plots branch.

Since this was a non-trivial thing, I think we should also "advertise" this approach.

You can my final drafted script in https://github.com/teuben/dyshp/blob/main/scripts/fake_plot.py but I have a few simpler versions for mpl and dysh only in that same directory.

[teuben]

I think it would still be nice to have block=True the default, as it is in matplotlib.

There are some strange blocking behaviors with mpl i the same script, as I've shown in my fake_plot.py script, but that may also be the peculiar way how I've written the interaction. See comments in that script, it's now a show stopper, more a feature?