--preservePyml

[nevrome]

To conclude the --ordered saga I drafted an implementation for the --preservePyml argument as discussed in #300. I had to make some design decisions you may or may not like, @stschiff and @TCLamnidis, so please comment.

I went for a simple --preservePyml switch that only works in case there is a single relevant source package. If there are more, and the option is active, then forge stops with an error.
With --preservePyml forge runs through normally, but then replaces the following fields of the generated POSEIDON.yml file from the source POSEIDON.yml file:

• description
• contributor
• packageVersion
• lastModified
• readmeFile
• changelogFile

Copying the readmeFile and changelogFile fields requires of course also to copy the respective files.

You'll note that the package title is not copied. It can be easily reproduced by the user with the -o or -n options.

There is a bit of a weird interaction with the --minimal flag. With --preservePyml and --minimal the resulting package may have a README and a CHANGELOG file, but not a .janno or a .bib file. This sounds like one of these unexpected interactions that could bite us eventually...

I haven't written any tests yet.

[codecov]

Codecov Report

Attention: Patch coverage is 77.31959% with 22 lines in your changes missing coverage. Please review.

Project coverage is 68.50%. Comparing base (a141fc3) to head (14baad7).

Files	Patch %	Lines
src/Poseidon/CLI/Forge.hs	78.12%	4 Missing and 17 partials ⚠️
src/Poseidon/Utils.hs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #304      +/-   ##
==========================================
+ Coverage   68.21%   68.50%   +0.28%     
==========================================
  Files          26       26              
  Lines        3492     3540      +48     
  Branches      394      399       +5     
==========================================
+ Hits         2382     2425      +43     
  Misses        716      716              
- Partials      394      399       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[stschiff]

Thanks, will take a look.

[stschiff]

OK, I checked the code, without trying it out, and I can green light all of these changes. Looks good to me. Would be nice to turn at least one of the existing forge-golden-tests into a test for --preservePyml. Would be easy enough I think.

Regarding the --minimal flag, I must admit I never fully understood what use-case that actually supported. I understand why it's in init, but for forge I never got the point. In any case if we are serious about it, we should not allow --minimal and --preservePyml together, simple as that: With --minimal a user has expectations for empty fields in the Yaml file, while with --preservePyml those very fields are filled. These two expectations are fundamentally incompatible with each other. So I would suggest not to allow that.

[nevrome]

I used --minimal in the past for pipelines, where I wanted to keep the intermediate data consistently in packages. This was before --onlyGeno, though, so maybe it's less relevant now. Still convenient to have it, so I don't want to touch it at the moment.

Anyway: I refactored forge to replace the increasingly complex boolean logic by a clear separation of output modes with the new type ForgeOutMode. We now have four mutually exclusive output options: --onlyGeno, --minimal, --preservePyml and the default (all covered by golden tests). There are no more unpredictable interactions between them. I think it's much more clear like this.