Idea: spellcheck as part of the tests?

[carols10cents]

Maybe this is more work than it's worth, considering that we'd need to ignore code blocks and add words like "enum" to whatever dictionary that gets used, buuuut it's a thought I had that I wanted to see what other people thought.

aspell suggests "enema" for "enum", btw.

[steveklabnik]

I like this idea!

[carols10cents]

https://github.com/azerupi/mdBook/issues/141 is relevant to these interests!

[JIghtuse]

So, in #275 I found a few spellings thanks to aspell.

Now I discovered a needed change to tell aspell about encodings of "personal dictionaries" it uses. It is now possible to generate a dictionary file with valid words (should be reviewed manually, but can be created only once and updated incrementally when new specific words appear) and then check book contents with global "language" dictionary and a personal one.

Here is a script to generate dictionary and check files if dictionary exists: https://gist.github.com/JIghtuse/0a92074f56007d2010396063aa221bf5#file-spellcheck-sh

And here is a current dictionary file for the book: https://gist.github.com/JIghtuse/0a92074f56007d2010396063aa221bf5#file-dictionary-txt

Would be glad to help integrating it into CI! 😸

[carols10cents]

That would be awesome!!!

I think the requirements for getting this into CI are:

• If there are misspellings, the script should exit with a non-zero status, and hopefully print out what's misspelled so that we can fix them easily.
• If there are not misspellings, the script should exit with a 0 status.
• Add your script and dictionary to the repo, in the root is fine
• Add the invocation of your script to .travis.yml, probably at the beginning of the script section, let's see how that feels

Bonus (optional) tasks, I can take care of these if you want:

• Make a second PR that isn't meant to be merged that has a misspelling to test out the failure case
• Add documentation to the README on how to set up and run aspell locally, and how to add words to the dictionary

Please let me know if any of this is unclear or if you get stuck with anything!

[JIghtuse]

@carols10cents pretty clear, thanks! I'll try to make these changes asap.

[JIghtuse]

Hey! I finally implemented it.

aspell can be weird at times! I've made a few workarounds.

• I didn't found a way to show line number with typo. Solution: script uses grep to find it. Not a great solution due to possible false positives, but I don't know how to improve it for now.

• aspell thinks it is fine to modify personal dictionary even if no errors was found. So dictionary passed to aspell cannot live in git. Solution: copy dictionary to /tmp and pass this path to aspell.

[JIghtuse]

@carols10cents please look at the last three commits in https://github.com/JIghtuse/book/commits/master and if it is fine, I'll create a PR.

[JIghtuse]

As an alternative, here is a python script doing the same checks: https://gist.github.com/JIghtuse/a059d63bf31b56c0e71ac03e69584f48

Code looks a bit clearer for me and doesn't have aspell quirks, but it generates a bit more false positives, so it is needed to extend dictionary with words like:

السلام
عليكم
שָׁלוֹם
Здравствуйте
नमस्ते
こんにちは

Also, there is no interactive mode like aspell provides, but probably this script is easier to maintain than proposed bash solution. What do you think?

[carols10cents]

@JIghtuse the bash script solution looks fine to me!! I'm excited to see it in action-- open that PR! :)