gitattributes seems not to be used

[pawamoy]

Preliminary Steps

Please confirm you have...

• reviewed How Linguist Works,
• reviewed the Troubleshooting docs,
• considered implementing an override,
• verified an issue has not already been logged for your issue (linguist issues).

Problem Description

So it says Linguist takes into account any override. I tried locally to override *.1 files to be detected as "vendored" or "documentation", using .gitattributes (in order to get rid of them in the stats), but it does not work.

I tried many different combinations:
*.1 linguist-vendored, *.1 linguist-vendored=true, *.1 linguist-documentation, licenses/* linguist-vendored, but my code is always reported as being 86.12% Roff.

$ cat .gitattributes 
licenses/* linguist-vendored

$ linguist . --breakdown
86.12%  Roff
13.88%  Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

Roff:
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.1
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.2
{{cookiecutter.repository_name}}/licenses/GNU_Lesser_General_Public_License_v2.1
{{cookiecutter.repository_name}}/licenses/SIL_Open_Font_License_1.1

URL of the affected repository:

https://github.com/Pawamoy/cookiecutter-cookiecutter

Last modified on:

Not relevant. It's not even working locally. I have installed linguist version 6.0.1 with gem install github-linguist through rbenv, ruby version 2.4.0.

Expected language:

It would be better if the licenses text were detected as text, but this is not the issue here, since I am responsible for having named the files with a .1 extension (as in European_Union_Public_License_1.1). I don't mind that.

Detected language:

Roff, which is fine.

[Alhadis]

Try pushing a minor change to one of those license files. GitHub caches each repository's language breakdown, which means changes to .gitattributes files won't become visible until after the next push.

[pawamoy]

Yes I understand that, but I'm trying to do this locally, not to see the changes on GitHub.

[Alhadis]

Does *.1 linguist-language=Text work?

[pawamoy]

No 😂 Still detected as Roff!

[lildude]

Two things:

1. your override won't match as licenses is not in the root of your repo
2. you need to commit your .gitattributes for it to take effect

$ linguist --breakdown
86.12%  Roff
13.88%  Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

Roff:
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.1
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.2
{{cookiecutter.repository_name}}/licenses/GNU_Lesser_General_Public_License_v2.1
{{cookiecutter.repository_name}}/licenses/SIL_Open_Font_License_1.1

[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ echo "licenses/* linguist-vendored" > .gitattributes
[lildude-MBP : ...ookiecutter-cookiecutter (master●)]$ linguist --breakdown
86.12%  Roff
13.88%  Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

Roff:
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.1
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.2
{{cookiecutter.repository_name}}/licenses/GNU_Lesser_General_Public_License_v2.1
{{cookiecutter.repository_name}}/licenses/SIL_Open_Font_License_1.1

[lildude-MBP : ...ookiecutter-cookiecutter (master●)]$ git add .gitattributes
[lildude-MBP : ...ookiecutter-cookiecutter (master●)]$ git commit -m 'Add attribs'
[master 0062a89] Add attribs
 1 file changed, 1 insertion(+)
 create mode 100644 .gitattributes
[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ linguist --breakdown 
86.12%  Roff
13.88%  Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

Roff:
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.1
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.2
{{cookiecutter.repository_name}}/licenses/GNU_Lesser_General_Public_License_v2.1
{{cookiecutter.repository_name}}/licenses/SIL_Open_Font_License_1.1

[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ echo "*/licenses/* linguist-vendored" >! .gitattributes
[lildude-MBP : ...ookiecutter-cookiecutter (master●)]$ git commit -am 'Update attribs'
[master 20791da] Update attribs
 1 file changed, 1 insertion(+), 1 deletion(-)
[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ linguist --breakdown
100.00% Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ echo "*.1 linguist-vendored" >! .gitattributes
[lildude-MBP : ...ookiecutter-cookiecutter (master●)]$ git commit -am 'Update attribs'
[master 8d27937] Update attribs
 1 file changed, 1 insertion(+), 1 deletion(-)
[lildude-MBP : ...cookiecutter-cookiecutter (master)]$ linguist --breakdown
59.96%  Roff
40.04%  Python

Python:
hooks/post_gen_project.py
setup.py
specs/support/runner.py
specs/support/settings.py
specs/template_spec.py
{{cookiecutter.repository_name}}/hooks/post_gen_project.py

Roff:
{{cookiecutter.repository_name}}/licenses/European_Union_Public_License_1.2

[lildude-MBP : ...cookiecutter-cookiecutter (master)]$

[pchaigno]

A single asterisk before /licenses/ works if you're sure licenses is in a subdirectory. If it may be deeper in the hierarchy of directories, you'll need two wildcards:

**/licenses/* linguist-vendored

[pawamoy]

So, the important points:

• .gitattributes must be committed again after each modification for linguist to take it into account,
• patterns in .gitattributes are not magically matched against whole path of a file, we need to use wildcards when necessary.

Would a PR adding these two points in the related README section be accepted?

Thank you @lildude, @pchaigno and @Alhadis!

[Alhadis]

Also note that a pattern like licenses/ won't work, it needs a trailing asterisk: licenses/*.

This is unlike a .gitignore file, where entire directories can be marked by specifying the directory name, rather than a globbing pattern:

# .gitignore
path/to/folder      # Works
path/to/folder/*    # Works

# .gitattributes
path/to/folder      # Doesn't work
path/to/folder/*    # Works

[pchaigno]

Would a PR adding these two points in the related README section be accepted?

👍 for me. The second point is not specific to Linguist tough. Maybe point to the appropriate documentation and change one of the examples to show a more complex case (e.g., **/dir/*)?

[pawamoy]

Indeed it's not specific to linguist, a simple reminder + doc link + example would be good enough. Let's see that in a PR :)