Analysis: added matplotlib figure generation with several desirable performance metrics.

[nikwl]

Thanks so much for this awesome tool!

I've been wanting to see a more graphical representation of how my plots are doing. This pr adds the option to create a matplotlib figure that summarizes your plotting performance, including:

• Dumbbell plot visualizing start, end, and duration of your plots over time.
• Average plot rate over time (computed over a sliding window).
• Average plot time (computed over a sliding window).
• Cumulative number of plots over time.

It adds two new arguments to plotman analyze:
--logdir: a directory that contains a number of log files (this can be used instead --logfile).
--figfile: path to save the figure to.

It adds one new dependency: matplotlib

The figure creation is done by high level functions which are easily removable or interchangeable. If the maintainers don't like my method of integration, figure creation should be easy to reintegrate in a better way. Or if using matplotlib is undesirable then hopefully the body of the functions can be reused in some way to still be helpful for doing analysis.

Thanks!

[rafaelsteil]

Here's a little script I created that generated a PDF with charts using data from plotman export rafaelsteil@e4a0632

[altendky]

Thanks for your contribution here. I apologize for leaving it hanging for so long.

What do you think of separating it from analyze and making it its own plotman graph subcommand or such? I think it could get its own plotman.graph module as well.

[altendky]

There should probably be an extra down below for plotting graphing.

graph =
    matplotlib ~= 3.4
    numpy ~= 1.20

[altendky]

It looks like there may have been some confusion with the catch up merge. Take a look down through the diff and make sure all the changes are intended to be part of this PR. Sorry for my role in this what with taking forever to catch up on the community contributions.

[altendky]

Do we need dedicated log parsing here? If the existing parsing doesn't catch everything, maybe we could add the extra bits there? Though I have to admit we have two parsers right now. I was hoping to make some time to get everything using the newer one. I'll try to get that done first.

[altendky]

Is it particularly needed for this to be directly runnable rather than just using the plotman CLI access?

[nikwl]

Hey @altendky, thanks for checking this out and for all of your feedback. I really only got a chance to bring the fork up to speed yesterday - I do mean to fix all the above issues but I'm a little pressed for time right now and I didn't get a chance to double check anything or actually test the code. When I do finally get everything working I'll "un-draft" the pr.

Thanks!

[altendky]

Awesome, thanks. Don't rush on my behalf.

[nikwl]

Hey @altendky, sorry this has taken so long but I finally got around to updating this pr. If you could look it over when you get a chance that would be awesome. Thanks!

[altendky]

I think we're close. Still sorry for my slowness... again.

[altendky]

Oh, and add a changelog entry.

[altendky]

Maybe try nikwl#1 to catch up and resolve the conflicts.

[nikwl]

@altendky Thanks so much for the pr, that made things much, much simpler. I've updated the log parsing and tested it on my plots, seems to be working.

requested changes made

[altendky]

Welp, a bit closer with the hints. If you aren't sure how to handle these I can take a look later.