CamelCase support?

[felixonmars]

Is it possible to treat camel case class names etc. as separate words, so that spelling errors could be found?

[luzpaz]

This would be amazing but I'm worried it would report a ton of false positives.

[felixonmars]

Maybe a switch would help avoid that?

[thdot]

I have implemented this feature locally a few weeks ago and had no problems with false positives.
The reason why I haven't pushed this is because the in-file replace functionality doesn't work with this quick hack. It's the same situation as for PR #174 which doesn't support the '-w' flag currently as well.

I think at some point we have to discuss how important the write-changes option is and how it influences new features like support for CamelCase, c-escapes or customized regular expressions to split the words. Since I use codespell only for reporting misspellings (it runs on every compile of my projects) I do not care much about this option.

Thoughts?

[larsoner]

It would be fine with me to say that only a subset of corrections can be automatically applied, so long as the others were still reported. For example adding a feature for within-CamelCase spelling errors, it seems reasonable to specify in the docs that these can be reported but not (at the moment) automatically corrected.

[lucasdemarchi]

I agree with @larsoner

[peternewman]

See also discussion in #314 .

[peternewman]

@thdot where is your CamelCase read-only checking code? I'd love to get it merged in!

[TysonAndre]

An example regex to extract individual words from camelCase and mixedACRONYMSpelling that may be useful (PCRE, though) /[a-z]+|[A-Z](?:[a-z]+|[A-Z]+(?![a-z]) (not aware of hyphens or accented characters, could skip that check if either were found)

This was from a project to check spelling and be aware of PHP's syntax (e.g. single-quoted strings)

https://github.com/TysonAndre/PhanTypoCheck/blob/0.0.3/src/TypoCheckUtils.php#L100-L102

[TysonAndre]

An example regex to extract individual words from camelCase and mixedACRONYMSpelling that may be useful (PCRE, though) /[a-z]+|[A-Z](?:[a-z]+|[A-Z]+(?![a-z]) (not aware of hyphens or accented characters, could skip that check if either were found)

I'd just like to add that my initial solution had quadratic performance for long strings of lowercase letters, and should not be used as-is. (it didn't specify a start boundary)

'/(?:[a-z][^a-zA-Z]*[A-Z]|_)/' can be used as a sanity check of whether a word is camelCase or snake_case (only needed if splitting a string is slow).

'/[a-z]+|[A-Z](?:[a-z]+|[A-Z]+(?![a-z]))/' can be used to extract individual parts

[BlaineEXE]

Thought I might note that we use codespell in our CI for github.com/rook/rook, and we have a function AtLeast() which codespell suggests be at least, and I'd love to have the CamelCase support so we can still find accidental atleast strings in documents while still allowing AtLeast() functions.

[matkoniecz]

I just want to mention that I would be perfectly happy with detection only quadratic camel case support (I found defaultNameOccurances only by accident as occurance was existing also as standalone typo)

[TysonAndre]

I mean the one that I posted fixed the quadratic runtime I had with my first attempt, so maintainers should avoid introducing similar bugs in however they solve it.

The buggy one was /(?:[a-z].*[A-Z]|_)/ and was slow for long lowercase strings due to causing backtracking and needing to start from every possible start position

The fixed one was /[a-z]+|[A-Z](?:[a-z]+|[A-Z]+(?![a-z])/