Support inline ignores

[whoopsmith]

I'd like to use codespell as part of an automated check workflow. For that I need to be able to ignore certain cases on a per instance bases.

Here's an example. In the code base there is a command called 'Trim All'. The comment contains a statement like:
// Track how many Trim Alls have been issued.

codespell finds Alls as a possible error.

I don't want to add 'Alls' to the ignore words list as it could be a real error in other cases. But I don't want this message generated on every run either.
The ignore word list only supports single words So I can't do 'Trim Alls' as an ignore.

Some sort of in line keystring that allows the next occurrence of a specific word error would be useful. Most code formatters, pretty-printers, and static analysis tools have some sort of inline // disable-whatever option that allows you to skip the trigger on a per instance basis.

[jolivier23]

Or something like pylint does where you say
/* codespell: ignore <expression|list of words> */ with a matching "find" or "unignore" comment

[eavanvalkenburg]

This is a must when using codespell with a large multi-dev project, because you don't want to then alter the whitelist for every individual thing and what is correct in some contexts is not correct in others, so similar to pylint, flake8, etc adding a inline comment that tells codespell to ignore that particular line would be very welcome!

[vadz]

What about extending the ignore file format to have a second column restricting the entry to being used for the given file or, better, giving line or range of lines in the given file? Sometimes you can't modify the sources themselves, but you can always add a file with codespell exceptions.

[eavanvalkenburg]

In my case I don't want to change some project level thing for my one file that needs an exclusion and where in some contexts I need the exception, while in other contexts it doesn's and should be flagged, so just the single line to be ignored would be far the easiest then, my case is with Home Assistant, where all the integrations are created independently of the main but there is a pre-commit hook with codespell, that shouldn't be updated by individual developers working on single things, but mine failed because I had a field coming from a external service that I had to check against which was flagged by codespell, I ended up having to move the checking list to a seperate json file (which were excluded by the project level codespell config) but it made my code more complex (reading a external file, instead of just a constant in a dict).

[peternewman]

The --exclude-file/-x option already has the ability to list lines to ignore (although in a separate file, rather than decorating the line).

For example:
https://github.com/OpenLightingProject/ola/blob/master/.codespellignorelines

[vadz]

Thanks for the -x option hint! I've totally missed it and it's perfect for my needs.

[odashi]

Any updates on this?

[peternewman]

Any updates on this?

If --exclude-file/-x doesn't suit your usage @odashi , then someone still needs to propose a format (either a comment at the end of a line), or a comment on the line before, either with one syntax that works on multiple programming languages, or more likely a variable prefix (e.g. # or // or whatever and then a static match phrase).

Then someone (not necessarily the same person) will need to implement it.

Suggestions and pull requests welcome.

[ssbarnea]

Maybe someone can tell me how to avoid codespell yelling about that python line KEGEX = r"^st(?:dout)?$" -- dout being the problem. Am I supposed to skip entire file because of this? Sounds like using a BFG to kill a moskito.

[abhinavsingh]

I have a test with following chunked encoded packet. It thinks nd is not spelled correctly. Should I ignore the entire test file?

b'HTTP/1.1 200 OK\r\nTransfer-Encoding: chunked\r\n\r\n5\r\nHello\r\n5\r\n Worl\r\n1\r\nd\r\n0\r\n\r\n',

Reference: https://github.com/abhinavsingh/proxy.py/pull/1064/files#diff-fc90f112bb0886415d939834b630f798e633c63c977a60a69b60b8fd7fcecd42R94

[abhinavsingh]

I have a test with following chunked encoded packet. It thinks nd is not spelled correctly. Should I ignore the entire test file?

b'HTTP/1.1 200 OK\r\nTransfer-Encoding: chunked\r\n\r\n5\r\nHello\r\n5\r\n Worl\r\n1\r\nd\r\n0\r\n\r\n',

Reference: https://github.com/abhinavsingh/proxy.py/pull/1064/files#diff-fc90f112bb0886415d939834b630f798e633c63c977a60a69b60b8fd7fcecd42R94

Quickest hack is to add following at top of the file:

    .. spelling::

       nd

[peternewman]

So @ssbarnea and @abhinavsingh (and others on this thread), you've currently got a few options, in rough order from simplest, but biggest impact on finding other typos through the most complicated but restricted.

1. Skip the entire file with --skip
2. Skip the relevant line(s) with --exclude-file
3. Skip the specific typo with --ignore-words-list (e.g. dout or nd)
4. Hide the typo with --ignore-regex (possibly better for @abhinavsingh case)

Quickest hack is to add following at top of the file:

    .. spelling::

       nd

I'm not aware of that syntax, is it specific to Jupyter notebooks?