Revert adding lien to the rare dictionary

[nikolaik]

Since lien is a frequently used legal term and also happens to be used when working with the Terraform Google
Terraform resource google_resource_manager_lien, it feels wrong to treat it as a typo.

Humbly suggesting to revert that change

This reverts commit 536ccb5.

Fixes #3630

[DimitriPapadopoulos]

Note that this typo is in the rare dictionary. Perhaps the rare dictionary should not be selected by default.

[peternewman]

Can you move these to the code dictionary please, which seems like the right place to me given they upset programming related text, but are still potentially valid/useful outside of that domain.

[nikolaik]

Can you move these to the code dictionary please, which seems like the right place to me given they upset programming related text, but are still potentially valid/useful outside of that domain.

Moved now

[DimitriPapadopoulos]

@peternewman @nikolaik It seems to me that lien is not related to programming, but to a specific domain.

[nikolaik]

@peternewman @nikolaik It seems to me that lien is not related to programming, but to a specific domain.

I agree with you @DimitriPapadopoulos but no strong opinion, so open to go either way

[peternewman]

@peternewman @nikolaik It seems to me that lien is not related to programming, but to a specific domain.

When you say specific domain do you mean Google Terraform? Isn't that the same as say:

codespell/codespell_lib/data/dictionary_code.txt

Line 52 in 4644f58

movei->movie

Which is only relevant to Assembler (AFAIK), to pick probably one of the more obscure ones?

[rjmoseley]

From my perspective it's not particularly related to programming, it's a legal term that just so happens to be used by Google Cloud Platform. Whilst it's that usage that has resulted in this debate, fundamentally it's just a word, and one I (native British English speaker) was familiar with long before I came across it's usage in GCP.

[nikolaik]

From what I'm hearing in this discussion and #3630 is that adding a well defined word with current use as a misspelling would not make sense and do more harm than good. Originally added in #3460 as a misspelling with the justification "Found in Emacs."

I changed this PR back back to a plain revert, since that now seems like the best outcome, to me.

Fixed.

[DimitriPapadopoulos]

In the long term, I would rather keep rare words in the rare dictionary, that's what it's for. Perhaps the rare dictionary shouldn't be selected by default.

I'll settle for removing this entry for now. Thank you @nikolaik.

Just noticed another possible alternative. From the README:

3. Entry with one word, but with automatic fix disabled:

   clas->class, disabled because of name clash in c++
   

Note that there isn't a comma at the end of the line. The last argument is treated as the reason why a suggestion cannot be automatically applied.

Perhaps we should disable automatic fixes by adding a comment to all or part of the words in the rare dictionary. Example:

$ cat DICTIONARY.txt 
lien->line, actually a word in the OED and SCOWL (And Friends)
$ 
$ cat test.txt 
lien
$ 
$ codespell -w -D DICTIONARY.txt test.txt 
test.txt:1: lien ==> line  | actually a word in the oed and scowl (and friends)
$ 
$ codespell -D DICTIONARY.txt test.txt 
test.txt:1: lien ==> line  | actually a word in the oed and scowl (and friends)
$ 

[nikolaik]

Thanks Dimitri!

[nikolaik]

Autofixing cases which rely on a person using their discretion, where there is ambiguity (dictionary definition exists) sounds wrong to me.

[DimitriPapadopoulos]

Autofixing cases which rely on a person using their discretion, where there is ambiguity (dictionary definition exists) sounds wrong to me.

So would you agree to reintroducing lien in the rare dictionary, but with autofixing disabled?

[nikolaik]

So would you agree to reintroducing lien in the rare dictionary, but with autofixing disabled?

I'm not in favour of reintroducing lien, for the reasons mentioned above. Though for existing fixes in the rare dictionary that have dictionary definitions, I would consider it, since that would be less of a breaking change for users.

I would like to add that one of the reasons this project appeals to me is that it provides an good out of the box experience, with defaults that fixes common mistakes, does not give false positives, and requires little intervention from users to give value. I use codespell via the pre-commit hook and one of the sources of friction are when a mistake has multiple possible fixes (ambiguous) which I don't mind. Though disabling autofix for fixes in the rare dictionary, that are part of the default experience, will add more friction, maybe too much.

[DimitriPapadopoulos]

So removing the rare dictionary from the default dictionaries would be the best option, wouldn't it?

[nikolaik]

I can't say, it depends a lot on the contents of the rare dictionary, but it might

[DimitriPapadopoulos]

'rare' for rare (but valid) words that are likely to be errors