feat: Add /tesseract to PYTHONPATH, similar to native execution

[linusseelinger]

Description of changes

Add /tesseract to PYTHONPATH in Tesseract containers. This takes containerized execution of tesseract_api.py closer to native execution regarding finding Python modules in the same directory etc.

Testing done

Local testing

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.35%. Comparing base (7a92d45) to head (360b754).

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #372       +/-   ##
===========================================
+ Coverage   66.27%   76.35%   +10.07%     
===========================================
  Files          29       29               
  Lines        3345     3345               
  Branches      525      525               
===========================================
+ Hits         2217     2554      +337     
+ Misses        936      558      -378     
- Partials      192      233       +41     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[apaleyes]

looks good, thanks! do you want to take this all the way and add an example to showcase this change?

[xalelax]

I wanted to write the same; full example on just this is probably overkill, but maybe mention it:

• in a paragraph here https://docs.pasteurlabs.ai/projects/tesseract-core/latest/content/creating-tesseracts/advanced.html
• or somewhere in this example somewhere in https://docs.pasteurlabs.ai/projects/tesseract-core/latest/content/examples/building-blocks/localdependency.html

[apaleyes]

@xalelax why is a full example an overkill? in the simplest case it is literally 3 files, no more than 20 lines each

[xalelax]

I meant that it's such a small change that probably doing a whole example is a bit too much, not in the sense that it would take us a lot to write it, but that it would be eccessive detail for such a small feature. I'd just mention somewhere in the docs that the path of tesseract_api.py is in the PYTHONPATH (or something more simple and expressive than that), but I'd be fine with an example as well.

[dionhaefner]

I'm a bit confused by this PR, specifically this statement:

This takes containerized execution of tesseract_api.py closer to native execution regarding finding Python modules in the same directory etc.

From my understanding (and testing), native execution via tesseract-runtime does not put the script on PYTHONPATH. I.e., we have now diverged between containerized and non-containerized usage? @linusseelinger @xalelax

[xalelax]

@dionhaefner you are correct; this would be the case only if one happens to run tesseract-runtime when cwd is the folder where tesseract_api.py lives, which might be common when developing stuff, but not universal even when scoped to just dev work.

TBH I'm not a fan of the concept in general, and if not backed by some user story I'd just revert this, or achieve parity by adding to the PYTHONPATH the parent of TESSERACT_API_PATH (once it's done on the runtime, there should be no need to add it in the dockerfile btw)

@linusseelinger WDYT?

(also, sorry for the late reply, I had missed this 🙏 )

[dionhaefner]

achieve parity by adding to the PYTHONPATH the parent of TESSERACT_API_PATH

Yes either this or add a sharp edge to the docs (with a snippet that does sys.path.append(os.path.dirname(__file__))). I'm not sure if there are unintended side effects when adding the parent dir to the path. It's certainly unintuitive though that users also need to add Python modules to package_data in order to play well with containerization. So if they already need to make a conscious choice on where to put their auxiliary modules, they might as well add them to sys.path?