Ignoring paths is not supported in source_files

[carusooo]

I'm having issues with configuring spellcheck to ignore file paths when passed in via the source_files argument

From the documentation here and in PySpelling, it looks like this should be all that's required.

  sources:
  - '**/*.md*|!docs/api/**'`

This is how the action is invoked

jobs:
  spellcheck:
    name: Spellcheck
    runs-on: ubuntu-latest

    steps:
    - uses: actions/checkout@v4.1.0
      name: Checkout
    - id: changed_files
      name: Changed Files
      uses: tj-actions/changed-files@v45
      with:
        files: |
           **.mdx
    - name: List all changed files
      env:
        ALL_CHANGED_FILES: ${{ steps.changed_files.outputs.all_changed_files }}
      run: |
        for file in ${ALL_CHANGED_FILES}; do
        echo "$file was changed"
        done
    - uses: rojopolis/spellcheck-github-actions@0.41.0
      name: Spellcheck
      
      if: ${{ steps.changed_files.outputs.all_changed_files }}
      with:
        config_path: spellcheck.yaml
        source_files: ${{ steps.changed_files.outputs.all_changed_files }}
        task_name: Markdown

I suspect this is working as intended, wanted to confirm. I think I'll need to filter out the changed files in the github action

[jonasbn]

hi @carusooo Is this a public repository? if so then please provide a link

[carusooo]

It is a public repo, here's the spellcheck.yaml file. I ended up adding the filter to the changed file detection workflow step, since I think this action is basically connecting to the PySpelling CLI and passing in the files there.

Overall I think I need to revise my approach to spellcheck the built HTML files since mdx files aren't well handled by PySpelling.

[jonasbn]

Hi @carusooo

Thanks for sharing I will see what I can learn from it, perhaps something can be used in the action.

[facelessuser]

I'm not sure what is meant by not supporting mdx files. Pyspelling doesn't have a filter that has control to filter directly in Markdown, but has one that converts Markdown to HTML which is easier to selectively filter within.

I'm certain an interested party could possibly use write their own filter using a Markdown parser that allows more control than what we currently use that might allow such control, but I would generally argue HTML is much easier to filter with available libraries. Often I build my documentation from Markdown and then use the HTML filter.

As for file exclusions, PySpelling probably could make this more clear. I think we reference wcmatch (the library that controls this) but expect the user to take a look. But you can exclude directories. It uses the NEGATE flag in the wcmatch library by default that allows you to specify patterns that start with ! to exclude them from search.

So you could do something like what is shown below. You can provide multiple rules that will be evaluated simultaneously if you separate them with |. So you can add an exclusion rule and an inclusion rule (order doesn't matter).

  sources:
  - '!docs/api/**|**/*.mdx'

Assuming this action passes the source normally to Pyspelling, I imagine you could do something similar.

[facelessuser]

Since Pyspelling allows a list of source patterns, I wonder if I should just update it so all patterns our evaluated simultaneously so that this would work.

sources
- '**/*.mdx'
- '!docs/api/**'

There's no limitation in the underlying library to prevent this, I think we just loop through the sources...just because.

[facelessuser]

I realized looking at the OP again, I did not read things very well as the opening post did specify how exclusions can be used 😅 . I also see that PySpelling documentation does indeed mention how exclusions are done (I didn't realize I had documented that).

Regardless, I'm probably going to allow the following moving forward as I think it will be more intuitive for the user. I'm not exactly sure how this action interfaces with Pyspelling in this regard, but this should work via the --source command line option as well. I also will not invalidate the old approach (separating patterns with |) so those can continue to be used as well.

sources
- '**/*.mdx'
- '!docs/api/**'

[jonasbn]

Hi @facelessuser

I believe the core problem is that when you specify the use of source_files in the workflow specification. It does not apply the patterns from the configuration file.

Meaning that all ignored files get processed even if specified to be ignored by the configuration file.

So there is a conflict between the use from the Action perspective and the configuration file, since the action uses: --source.

From the documentation:

--source SOURCE, -S SOURCE
Specify override file pattern. Only applicable when specifying exactly one --name.

REF: https://facelessuser.github.io/pyspelling/

So I need to find a way to get the pattern from the configuration applied after the list of source_files is accumulated.

[facelessuser]

So you want to be able to append new sources to sources already in the pyspelling config? Is there a reason this approach is desired? The way PySpelling was designed was just to setup rules and then call them. If you want to override something, then you can from the command line. It's possible I don't understand the use case, but from my perspective, if you were going to constantly append certain files to the rules, my question would be why not permanently include those patterns in the config?

[jonasbn]

@facelessuser I completely understand PySpelling and it's approach, but in the context of the GitHub Action, I am interested in only capturing the changed files, meaning a more dynamic approach. Here --source seemed to be a good approach.

But as pointed out by @carusooo this does not work as expected since the --source overrides the configuration, as you indicate is a more static approach, so the two usage patterns collide.

[jonasbn]

@facelessuser I have a large repository consisting of Markdown files, it takes significant time to process, to only checking the changes files makes sense.

Example output:

Using pyspelling on configuration outlined in >.spellcheck.yaml<
Running task >Markdown<
Checking file >awk/learn_awk.md<
Checking file >node/use_nvm.md<
Checking file >git/use_tig.md<
Checking file >node/use_nodeenv.md<
Checking file >rm/are_you_sure.md<
Checking file >shell/nifty_aliases.md<
----------------------------------------------------------------
Using aspell to spellcheck Markdown
Running Task: Markdown...
Compiling Dictionary...
> Processing: awk/learn_awk.md
> Processing: node/use_nvm.md
> Processing: git/use_tig.md
> Processing: node/use_nodeenv.md
> Processing: rm/are_you_sure.md
> Processing: shell/nifty_aliases.md

Spelling check passed :)

The repository contains 559 Markdown files if I count correctly.

[facelessuser]

Okay, in that case, you would need to pass in exclude patterns with !exclude. Currently, that requires you to specify them via --source "!exlude|include" as each --source is evaluated independently. If it makes it easier, the change I mentioned earlier would allow you to pass them in separately as all the patterns would be evaluated simultaneously. So --source include1 --source include2 --source !exclude would work.

Let me know if that helps, if not you can join exclude and include with |.

[jonasbn]

@facelessuser I will see if I can come up with a solution