[Bug] Codespell doesn't offer to capitalize word result if it doesn't start with the same first letter (?)

[luzpaz]

./src/effects/Reverb.cpp:455: Rever ==> Revert, fever
In the analysis example above fever is not capitalized

[peternewman]

I can't see this in the dictionary @luzpaz . Is it personal custom dictionary?

[luzpaz]

No... this is not a custom dictionary. And I'm not referring to how the words are displayed in the custom dictionary. This is specifically related to the logic that is used by codespell to capitalize/capital-case the suggested correctly-spelled words because the typo itself was capitalized/capital-cased. Make sense?

[peternewman]

Ah, which version are you running? I tried various searches via GitHub and failed to find it, but by inspecting the dictionary:

codespell/codespell_lib/data/dictionary.txt

Line 24590 in f22aa4a

rever->revert, refer, fever,

Please can you also share the whole line it's matching on and any command line flags/config you're using.

Also hopefully there's a PR to add reverb as a suggestion?

[peternewman]

Edit, agreed, it's broken:

text.txt:3: rever ==> revert, refer, fever
text.txt:4: Rever ==> Revert, refer, fever
text.txt:5: REVER ==> REVERT, REFER, FEVER

fix_case needs erm fixing. Well some tests adding first. It needs a for each around each suggestion individually, but also a quick check the reason isn't getting passed to it if present:

codespell/codespell_lib/_codespell.py

Lines 471 to 478 in 079b649

	def fix_case(word, fixword):
	if word == word.capitalize():
	return fixword.capitalize()
	elif word == word.upper():
	return fixword.upper()
	# they are both lower case
	# or we don't have any idea
	return fixword

[bl-ue]

Ah, I see @peternewman: the problem is that fix_case is being called essentially like this:

fix_case("Rever", "revert, refer, fever")

So it's capitalizing fixword, which is why only the first word is capitalized.

[peternewman]

Exactly. But the fix is not to just do a naive split, as the last one shouldn't be capitalised because it's a potential note.

[bl-ue]

Oh, that's a good point. So maybe we should change build_dict() and Misspelling to have a list of suggestions and an optional comment? That sounds like the best idea to me.

codespell/codespell_lib/_codespell.py

Lines 454 to 479 in 7db90e0

	def build_dict(filename, misspellings, ignore_words):
	with codecs.open(filename, mode='r', encoding='utf-8') as f:
	for line in f:
	[key, data] = line.split('->')
	# TODO for now, convert both to lower. Someday we can maybe add
	# support for fixing caps.
	key = key.lower()
	data = data.lower()
	if key in ignore_words:
	continue
	data = data.strip()
	fix = data.rfind(',')
	if fix < 0:
	fix = True
	reason = ''
	elif fix == (len(data) - 1):
	data = data[:fix]
	reason = ''
	fix = False
	else:
	reason = data[fix + 1:].strip()
	data = data[:fix]
	fix = False
	misspellings[key] = Misspelling(data, fix, reason)

I could easily undertake it.

[peternewman]

I think that handles it already, and reason is the reason. Basically just make sure the tests cover both cases, but yeah, feel free to get stuck in.

[peternewman]

@DimitriPapadopoulos fixed this in #2223

So maybe we should change build_dict() and Misspelling to have a list of suggestions and an optional comment? That sounds like the best idea to me.

codespell/codespell_lib/_codespell.py

Lines 454 to 479 in 7db90e0

	def build_dict(filename, misspellings, ignore_words):
	with codecs.open(filename, mode='r', encoding='utf-8') as f:
	for line in f:
	[key, data] = line.split('->')
	# TODO for now, convert both to lower. Someday we can maybe add
	# support for fixing caps.
	key = key.lower()
	data = data.lower()
	if key in ignore_words:
	continue
	data = data.strip()
	fix = data.rfind(',')
	if fix < 0:
	fix = True
	reason = ''
	elif fix == (len(data) - 1):
	data = data[:fix]
	reason = ''
	fix = False
	else:
	reason = data[fix + 1:].strip()
	data = data[:fix]
	fix = False
	misspellings[key] = Misspelling(data, fix, reason)

I could easily undertake it.

But something like this would be good still @bl-ue !