fix #329: fmt.diff checks same paths as ci._check_files

[rorour]

The problem I mentioned in today's meeting was in fact a bug that was fixed in version 22.12.0, and we only required >=22.0 (I was running 22.6.0.)

This update is enforcing new requirements, hence the change to pykern/pkcli/github.py and pykern/xlsx.py.

rshellweg passes ci format checks on my machine with these changes, but running pykern fmt run . will change files that pykern ci run does not diff. (Running pykern fmt run . in rshellweg will format those example files that were failing pykern ci run). Do we want pykern fmt run . to only run on <pkg>, tests, setup.py when run from the repo root?

[robnagler]

Do we want pykern fmt run . to only run on , tests, setup.py when run from the repo root?

Correct. We don't want to enforce standards on non-standard directories.

[robnagler]

Thanks for checking for bugs in black.

[rorour]

Do we want pykern fmt run . to only run on , tests, setup.py when run from the repo root?

Correct. We don't want to enforce standards on non-standard directories.

How should fmt run handle this? It seems like if fmt run is called from the command line we need to check each path that it receives and run it through ci._paths()?

[robnagler]

It seems like if fmt run is called from the command line we need to check each path that it receives and run it through ci._paths()?

fmt is a utility that abstracts our formatter. It might eventually know how to run JS formatter, for example. It does not need to know about repos except that it does need to know about exclusions since it delegates tree walking to the formatter, which is typical of formatters.

ci abstracts continuous integration, which includes formatting and other standards (check prints). ci is what enforces standards on a repo so it sets standards by calling fmt only where we set standards (tests, setup.py, etc.). Therefore, fmt doesn't need to be changed except to accept a list of files instead of a single directory.

[rorour]

Then should we add a new command to ci to replace our widespread use of pykern fmt run .?

[robnagler]

Then should we add a new command to ci to replace our widespread use of pykern fmt run .?

Not sure what you mean. Developers can still make sure the whole repo is formatted.

I'm not sure it is very widespread. For example, I never use pykern fmt run ., because I run pykern fmt run <file-saved>. I'm pretty sure Evan and Paul use the same command. I suspect VS code and PyCharm could do the same. This way there's no need to run fmt manually, which means there won't be ci failures for people forgetting to run pykern fmt run . manually.

Perhaps I'm misunderstanding.

[rorour]

Do we want pykern fmt run . to only run on , tests, setup.py when run from the repo root?

Correct. We don't want to enforce standards on non-standard directories.

This comment made me think you wanted some kind of change to the way fmt run currently works

[robnagler]

This comment made me think you wanted some kind of change to the way fmt run currently works

Ah, no. fmt enforces standards, of course.

[robnagler]

Great job sticking with the updates! Thanks.