Support running local scripts with pipenv run

[multimeric]

Currently pipenv run will run either:

• An executable in the $PATH
• A script listed in Pipfile's [scripts]

However, it is quite a common occurrence that users will want to run a local script, e.g. ./app.py. Currently there are a few ways to do this, but each has their downsides

• Add it to the [scripts] section - this works but assumes the user knows beforehand which scripts they may want to execute. Not good for ad-hoc running of scripts
• pipenv run python app.py - this forces the user to specify the interpreter. This means the shebang line is ignored, and requires the user to know which interpreter they want (if it's a shell script it shouldn't be run with python). Unix philosophy dictates that we leave the choice of interpreter up to the script itself (hence the existance of the shebang)
• Start a pipenv shell, and then run ./app.py. This is (imo) too much of a heavyweight solution. Starting a new shell shouldn't be required to run a single script.

Thus, could we implement the ability for pipenv run to execute arbitrary local scripts? Or is there some philosophical reason why we aren't allowing this?

[uranusjr]

I was thinking about this for a while (in a more general sense). Maybe we can add a fallback that when the first argument looks like a path, or starts with - (a dash), send the command line to python, so you can do

pipenv run -m foo   # Runs the foo module, like python -m foo
pipenv run ./foo.py  # Like python ./foo.py

(The path case has python preprended because it makes generalisation easier, and also works better for Windows.)

This would, however, requires a PEEP, and I just don’t currently have the time and capacity to draft it. I would be interested in assisting if you are interested in submitting a PEEP request.

https://github.com/pypa/pipenv/blob/master/peeps/PEEP-000.md

[multimeric]

Maybe we can add a fallback that when the first argument looks like a path, or starts with - (a dash), send the command line to python

I don't think we should assume that the script is a python script. For example, what if someone make a bash script that uses the pipenv python, but itself runs using the bash interpreter? If someone wants to run a local python script I think adding a shebang line #!/usr/bin/env python (admittedly this is Unix/Linux only, I'm not sure how Windows does this) or simply running pipenv run python foo.py would be sufficient.

This would, however, requires a PEEP, and I just don’t currently have the time and capacity to draft it.

I'd be happy to do this. Do I just PR a markdown document into the peep folder? What else should I add on top of what I have already written?

[uranusjr]

Personally I feel it is more productive to assume python, since pipenv run bash ./foo is always available, and the goal of pipenv run is never to duplicate shell parsing. This kind of discussion is exactly why a PEEP is needed though. Another possibility is to implement a shebang parser (it’s not that hard).

Yes, just send a PR to add a Markdown document to the peeps directory. The PEEP would need to summarise discussions around the particular topic, why certain ideas are rejected, and why a certain one is chosen. You don’t need to write all that immediately though. Start with copy-pasting your original post into it to kick off discussion, and gradually add things as they are discussed/resolved/rejected.

See #2819 for an example of discussions for PEEP-002. Also you might be interested in reasing some of Python’s PEP to get a feeling how this kind of documents can be structured. PEP 517 is a good example IMO.