View on GitHub

check-spelling-docs

Documentation for check-spelling

Cache dictionaries

There is some cost to dictionaries, so it seems to be worth caching them (especially with .dic+.aff).

Implementation

Dictionaries will go to a specific directory and actions/cache will track the directory.

The dictionary retrieval code can choose to skip retrieving any file(s) it already has.

Invalidation

Each dictionary's ETAG will be stored.

Use the ETAG to retrieve (or not) the content.

The code doesn't currently handle the case that there is no etag from the server, that should be added (in the interim, please use servers that generate etags...)

Clearing cache entries manually

Programmatically

You can get a list of caches using actions/cache: list-github-actions-caches-for-a-repository
You can delete a cache entry by id using actions/cache: delete-a-github-actions-cache-for-a-repository-using-a-cache-id

Interactively

Caches are accessible below workflows on the actions page of each repository and can be deleted directly.

Storage

There have been a couple of iterations so far... Layouts 1 and 2 (with etags) don't handle collisions in dictionary names particularly well...

Proposed layout

dictionary/
- {sha}/
  - {extension}
- ...

urls/
- {random}/
  - raw.url
  - resolved.url
  - etag
  - sha
- ...

Approach

Given a new unresolved url.
Resolve it.
Check urls/*/resolved.url to see if it is already present (=> 6).
Create urls/{random}.
Write raw.url and resolved.url to the directory.
Use the matching urls/{random} directory.
Use the etag (if available) to see if it's current (=> done).
Save the file to a temp.
Calculate the sha.
Write the new etag and sha to the directory (from 6).
Write dictionary/{sha}/{extension} ({sha} from 9, {extension} from 2).

Cleanup after all resolution is done

detect any directory/{sha} for which there is no urls/{*}/sha containing that value and delete it.
detect any urls/{*}/sha for which there's more than one urls/{*} with the same sha file value and warn about it.