Basic Automated Test Harness

[PrimaryCanary]

Here is a very simple framework for ensuring all the formatters properly configured. Currently, it has a small issue in that it doesn't actually work.

Each test consists of two files of the format formatters/mode-it-should-be-active-in/formatter.ext and formatters/mode-it-should-be-active-in/formatter.ext.formatted. The former is unformatted code while the latter is the expected output after running the formatter. test-apheleia.el should (but doesn't) return the result of calling the given formatter on the unformatted code to stdout.

All output from test-apheleia.el should be coming from the apheleia--run-formatters function. I think any stdout is being smothered because it's called from the async inferior process. How would I make test-apheleia.el return the formatted buffer to stdout?

TODO:

• Make test-apheleia.el work.
• Ensure all tests contain the proper text.
• Add supported modes and formatters table to README.
• Add test to ensure README table matches apheleia-modes-alist and apheleia-formatters.
• Run tests from Docker container.
• Add a test target in the Makefile.
• Add package-lint target in the Makefile.
• Run the tests in CI.

[raxod502]

I love how simple it is to add a test, just adding the two files and no extra busywork required!

[raxod502]

-type f,l

Do we want to find symlinks, or just files?

[PrimaryCanary]

I've used symlinks for modes that are likely to be working on the same code with the same formatter e.g. rust-mode and rustic-mode both work on Rust and likely use rustfmt. I can strengthen the requirement if you like.

[raxod502]

Fair enough, that makes sense.

[PrimaryCanary]

The good:

• The Docker container successfully builds and runs all formatters.
• The tests will run in CI.

The bad:

• There are currently two formatters building from source. I couldn't find a binary for ocamlformat and the Brittany binary Didn't Work On My Machine^TM.
• The container builds very slowly as a result. We might have to use your expertise to speed up CI and builds.
• No point tests which would make testing and fixing Point is sometimes moved small distances #2 easier.

The ugly:

• I haven't tested the CI settings because I don't use CircleCI. It looks like it will work.

If none of those issues are too big, I'd say this is ready to merge.

[raxod502]

I fixed things up so CI will run in only 2-3 minutes, and made various other improvements. The main thing remaining is that there seems to be something wrong with json-mode in Emacs 25, do you think you could take a look at that?

[raxod502]

This thread is being closed automatically by Tidier because it is labeled with "waiting on response" and has not seen any activity for 90 days. But don't worry—if you have any information that might advance the discussion, leave a comment and I will be happy to reopen the thread :)

[raxod502]

I wrote my own version of this in #72, partly drawing from the code here. The only things missing are package-lint integration, and the auto-generated table of formatters in the README.