Question about PEX binaries - is there a way to run a compound/multi-step command on startup (ex:

cmd1 && cmd2

)? Currently I have a binary definition that looks like this:

pex_binary(
    name="scrapy",
    script="scrapy",
    dependencies=[
        ":config",
        ":poetry#Scrapy",
        "//:poetry#spacy",
        "scraper/src/py/scraper/scripts:lib",
        "scraper/src/py/scraper/scripts:shell_scripts",
    ],
    environment="linux_docker",
    # Optimal settings for Docker builds
    layout="packed",
    execution_mode="venv",
)

But I want to run another command prior to running

scrapy

for setup like so:

python -m spacy download en_core_web_sm && scrapy <args>

My current plan is to create a shell script that downloads the spaCy model and then runs scrapy, and use that as the

entry_point

for the PEX - is that the right way to tackle this problem? Or is there a better solution?

Why not create a Python file and make that the entry point? The Python file can go nuts with compound actions.

This was my second alternative, which does make sense. Thanks for the pointer

I'm not sure why that wasn't your go-to idea, but scies offer that. Pants has no scie plugin yet though.

no "official" plugin

If it's not in Pants it effectively doesn't exist since we offer no way to discover plugins besides what you just did.

Yep, agreed.

@better-van-82973 a scie though, pretty much no matter how its build is packaged up in a plugin will take more lines of configuration and be more obtuse than a Python file for this purpose.

@wide-midnight-78598 I found your code here for running uvicorn: https://github.com/sureshjoshi/pants-plugins/commit/31e238c638af02c388aaf17081da83afe7725c82 And that’s actually pretty close to what I want! I think the part where I got confused is that packages like uvicorn and scrapy expose external scripts to run and I was trying to run those. But since they’re just Python packages I can invoke them from a script as well

Well, I read quickly, but that example requires 0 custom code in Pants 2.16.+ because you can embed default env and args in a PEX: + https://www.pantsbuild.org/v2.16/docs/reference-pex_binary#codeargscode + https://www.pantsbuild.org/v2.16/docs/reference-pex_binary#codeenvcode

Thanks for the help! I think I found another way around the problem for now but this is really good to know for other packages that expose executable scripts