Remind me, @polite-garden-50641, when we run Django's makemigrations command internally, we don't do it via

./pants run

, right?

https://github.com/pantsbuild/pants/issues/12129 ?

yup. we don't. However, I think it will work, we do use

./pants run

since when the pex file runs it is like running a python script (the only process that runs in isolation is the pex build) once the code runs it is not isolated and it has access to the local file system whatever the cwd is. I don't use it since it is faster for me to just use the underlying VE which I already have rather than wait for pants to build the pex file and run it.

and it has access to the local file system whatever the cwd is.

How does Django know to ouput relative to CWD, then and not the caller's

__file__

which will be in the tempdir? This seems like shouldn't work, so thats surprising

the way I see it, and I might be wrong, doing

./pants run

is just like doing

./pants package

and then just running the pex file (

./dist/my_script.pex

)

Spiritually, that is true. However the point stands that in your Python code doing

__file__

will not give you the sandbox path, which isn't relative to CWD (at least not in a meaningful way). So I'm not sure how Django can remedy that fact in a meaningful way. I guess, what's the full command/code look like? Maybe I'm missing something?

Screen Shot 2022-05-26 at 10.22.51 AM.png

Ah, yeah specifically the issue here is if it did need to make changes. It wouldn't write them to the repo. Try making a DB change and running the command

naa..it works...

....doens't that screenshot show it doesnt work? The migrations file is not in repo. It's in the tempdir

look where it wrote the migration script... ok. I stand corrected.

So the good news is I'm not taking crazy pills (I mean, I suppose I could be, we just can't prove it here). The bad news is we still can't codegenerate files easily in Pants 😛

yes. so as I mentioned, I used our virtualenv since it is faster... I am very impatient and won't wait for computers... I also find the fastest way to do things for my workflow...

Generally speaking. I think if users are reaching outside of Pants (especially into a virtualenv) it's a signal we're doing something wrong.

I agree.. pants can do many things better and faster.

Good 😈

case in point (pants run not using a cached pex) https://github.com/pantsbuild/pants/issues/10513

Yeah, OK, so no crazy pills, and no workaround, so we probably want to do that ticket

Your latest comment is 100% valid though, and I expect to see that as a stumbling block. Not sure if there's a way to "win" here 😕