Convergence charts using matplotlib

[iboyd-ansys]

This work started as some experiments with using matplotlib to display solution convergence and diagnostic charts but seems to have stabilised enough that some aspects can be exposed.

It uses System Coupling's support for output of chart data in CSV format. In the longer term, we would like to expose a gRPC stream (SyC already has the basics for this in place for other clients).

These are the main aspects of the work:

• Identification of CSV columns and mapping to interfaces and data transfers in the model. The model elements are embedded in longer strings and only their display names are given.
• Specification of what is wanted to be plotted.
• Mapping to a "subplot" definition .
• Instantiation of matplotlib subplots from the definitions.
• Tying matplotlib "axes" data to data from the CSV file, so that x-y series data, axes ranges, etc. can be set.
• Keeping this general enough to support a different (streamed) data source in future.
• Supporting "static" post-solve charts.
• Work towards supporting "live" convergence plots that are updated while the solve takes place.

The visible API is a new command show_plot exposed on the solution API root.

Still todo:

• Some of the internals are tentative and subject to change. For example:
  • An attempt has been made to make the code adaptable to an alternative, streaming, source of data but we are yet to try out that approach.
  • How to handle multiple interfaces is still an open question (for examples, should these be in separate figures?). It is supported in some parts of the code but ultimately only one interface may be plotted in current version.
• Test coverage could be improved.
• Live charts only partially work in that they display and update during solve but the wrong data is being plotted. This is not exposed in any accessible API yet.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 45.71843% with 374 lines in your changes are missing coverage. Please review.

Project coverage is 74.27%. Comparing base (270563d) to head (a69e4d0). Report is 26 commits behind head on main.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #312 +/- ## =========================================== - Coverage 85.05% 74.27% -10.78% =========================================== Files 34 41 +7 Lines 1893 2581 +688 =========================================== + Hits 1610 1917 +307 - Misses 283 664 +381 ```

[ochernuk]

@iboyd-ansys a few notes:

• for steady analysis the x-axis shows "Time" - it should be "Iteration" instead
• It would be good to have the ability to select only one chart to show and what to show (e.g. only convergence, or only transfer values chart), perhaps via syc.solution.show_plot(interface_name = ..., transfer_name = ..., show_convergence = False, show_transfer_values = True, show_source = False, show_target = True)
• Regarding multiple interfaces, I think separate plots are ok by default for now (though in the future it'd be good to add flexibility for that, but I see that as fairly low priority)

[iboyd-ansys]

@ochernuk, regarding your notes:

@iboyd-ansys a few notes:

• for steady analysis the x-axis shows "Time" - it should be "Iteration" instead
• It would be good to have the ability to select only one chart to show and what to show (e.g. only convergence, or only transfer values chart), perhaps via syc.solution.show_plot(interface_name = ..., transfer_name = ..., show_convergence = False, show_transfer_values = True, show_source = False, show_target = True)
• Regarding multiple interfaces, I think separate plots are ok by default for now (though in the future it'd be good to add flexibility for that, but I see that as fairly low priority)

The first two points here should be addressed now. As far as the third one is concerned, with mutiple interfaces on the same plot, we are likely to end up with a lot of "subplots" and I am not sure how well it will behave once we get beyond a few subplots. I believe it works to call show_plot multiple times with different arguments so that might end up being the recommended approach.