Limit impact of R & Python dependency detection to those content types for which it applies.

[sagerb]

Our behavior in a multi-type project directory (in which all of the publisher's project files exist in a single subdirectory hiearchy) can lead to some incorrect dependency detection.

The example here is a combination of a python and static-html project. The inspection done for the static-html file, will detect an inaccurate dependency on python, if the developer has created a requirements.txt file in that subdirectory. It would likewise also incorrectly detect an R dependency if a file named renv.lock was present.

Currently, the inspect functionality does content specific dependency detection, and then falls through a generic check for these two dependency files, adding an R or a Python dependency if they are found. This occurs for all content types, which is incorrect.

To fix this, we need to limit this generic check to only apply to a subset of content types, where R or Python dependencies make sense.

---

Analysis of issue reported from Tareef:

I had a chance to talk with Tareef and have him show me how he got into the state he did. I now understand it to be a bit different than what was discussed before:

• He was using a single subdirectory with multiple types of projects within it, similar (but different) to our mixed content test case. Within this subdirectory, he had both a Python project and a Quarto project, which he used to output a single HTML page.
• He selected to deploy from his generated HTML file.
• Our inspection correctly created a deployment w/ content type of html, but incorrectly identified an implicit dependency on Python because his multi-type project subdirectory contained a requirements.txt from his python api app.
  Deploying the project ran into issues during the deployment process w/ Connect. I’m not sure what happened there, but I’m pretty confident it can be reproduced.

I believe the workaround for him is to remove the [python] section from the config file, which will remove the Python dependency and redeploy the project to Connect. (He didn’t get a chance to try this because he wants to actually publish the project as a Quarto project, which would be re-run on a schedule. I’ll update the issue as I try to reproduce it.)