File placement improvements

[donald-e-boyce]

The intention of this pull request is to make the output file placement in find-orientations and fit-grains cleaner and simpler while cleaning up computational code that has hardwired file names and file handling that would be better done elsewhere, i. e. the config module.

Roughly, this is what has been done:

• put output files in the analysis_dir instead of in the working_dir with long names based on the analysis_name
• allow for slashes in analysis_name and automatically create subdirectories as needed for the analysis_dir
  • this is particularly useful when the config file has multiple (YAML) documents
• move path manipulations and hardwired file names from computational modules to the config module
• use pathlib for handling file and directory names
• update docstrings where useful (for affected modules/functions)

You activate the new file placement with the new_file_placement keyword in the root level of the config file, e.g.

analysis_name: ruby/scan-0  

# working_dir:
new_file_placement: on

If the new_file_placement is False (the default), then nothing is changed; the old file placement is used.

Originally, this was intended mainly for find-orientations, but fit-grains is slightly affected since it looks for a file written by find-orientations, i.e. the accepted orientations.

Running this on the multiruby and single_GE NIST_Ruby examples, I get the exact same results using the new or the old file placement.

[pep8speaks]

Hello @donald-e-boyce! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file hexrd/cli/fit_grains.py:

Line 127:59: E225 missing whitespace around operator

Comment last updated at 2024-12-11 16:24:01 UTC

[psavery]

Overall, I think this is a huge improvement! It is much better to contain the logic for files/directories within the config, rather than within the find-orientations and fit-grains functions. I left some comments - they are only minor things.

Can you make sure that the tests pass? It looks like some of the tests are failing, because apparently comparing Path objects with strings always fails (for example, cfg.working_dir == '/some/path' always fails now, because Path('/some/path') != '/some/path').

I think you just need to get the tests to use a Path object now in the comparison, or convert the Path object to str, whichever you prefer.

There are apparently some packaging issues too that I will look over.

[psavery]

Thanks Don! The changes look good so far! Now we just need to fix the tests (as mentioned here).

[psavery]

Looks good to me! I fixed the packaging issues in master (so they won't be an issue after merging).