Behaviour of the "[path]" placeholder have been changed in a refactor commit

[fa93hws]

• Operating System:
• Node Version:
• NPM Version:
• webpack Version:
• css-loader Version: 3.0+

Expected Behavior

[path] placeholder gives path-to-file

Actual Behavior

[path] placeholder gives path/to/file

How Do We Reproduce?

Should be very clear in this commit. #920
normalizeIdentifier is removed.

[alexander-akait]

Please create minimum reproducible test repo

[fa93hws]

Please create minimum reproducible test repo

It should be clear enough by seeing here https://github.com/webpack-contrib/css-loader/pull/920/files#diff-2b4ca49d4bb0a774c4d4c1672d7aa781L108.

If you are looking for the config

    modules: {
      mode: 'local',
      context: path.resolve(__dirname, '../src'),
      localIdentName: '[path][name]__[local]',
    },
    localsConvention: 'camelCaseOnly',
    importLoaders: 0,
sourceMap: true,

If you are looking for the repo
https://github.com/fa93hws/simple-typescript-react-starter

[alexander-akait]

Still can't understand, what is output was previously and what is output actually, no need link on source code, just clarify your problem

[fa93hws]

Still can't understand, what is output was previously and what is output actually, no need link on source code, just clarify your problem

The [path] in classname was mapped to path-to-file and it is mapped to path/to/file now.

[alexander-akait]

@fa93hws looks like regression, feel free to send a PR

[fa93hws]

@fa93hws looks like regression, feel free to send a PR

No problem. OK to have this function back? https://github.com/webpack-contrib/css-loader/pull/920/files#diff-2b4ca49d4bb0a774c4d4c1672d7aa781L108

[alexander-akait]

No, css classes can have non standard characters, so we can't back this function

[alexander-akait]

stop, why you think it is bug? Looks like it is expected behavior, and looks it is breaking change, i think you should use own localIdentName implementation if you want to have -

[fa93hws]

stop, why you think it is bug? Looks like it is expected behavior, and looks it is breaking change, i think you should use own localIdentName implementation if you want to have -

Because that commit says it's a "refactor" and a "refactor" normally won't change the behavior of the code.
That make me think it's probably a mistake?

[fa93hws]

stop, why you think it is bug? Looks like it is expected behavior, and looks it is breaking change, i think you should use own localIdentName implementation if you want to have -

Anyway, if it is expected behavior, I am happy to implement the - in my configuration.
Thanks

[alexander-akait]

Sometimes refactor has breaking change. Css allow to use / character so we need allow to use them in code also, but [path][name]__[local] to /path/to/name__hash looks very bad and it is very uncomfortably, maybe we should change / to - only for placeholders for better DX

[alexander-akait]

Let's think about this, because maybe it is really not good DX

[alexander-akait]

In version@2 it was -, right (i am look on code)?

[fa93hws]

In version@2 it was -, right (i am look on code)?

2.some.version is with - but 3.0 is not.
I'm not sure at which version it starts not to replace the / but it's after this PR: #920

I will replace the '/' to '-' only then.

[alexander-akait]

Let's keep open i think about this in near future, as workaround you can use custom localIdentName function to fix problem, feel free to feedback. Here interesting problem

[fa93hws]

Let's keep open i think about this in near future, as workaround you can use custom localIdentName function to fix problem, feel free to feedback. Here interesting problem

No problem, Thanks

[alexander-akait]

Yes, it is big regression 😞 thanks for catching this