search

npm / rfcs

Public change requests/proposals & ideation
Other
730 stars 240 forks source link

[RRFC] Obey the specific dist-tag #606

Closed gemwuu closed 2 years ago

gemwuu commented 2 years ago

Motivation ("The Why")

Since @ljharb has already proposed an RRFC of obey user specifier, which will be implemented at the next major version of npm@9.

In my cases, when installing a new dependency using npm install antd@conch, I want my antd to keep the exact conch dist-tag in my package.json, because this is exactly the version I need in my project. By saying version, I mean the version conch points to, not the moment I run the install command.

And in the obey user specifier RRFC, @wesleytodd suggested a new flag to support that behavior is more reasonable.

Example

Assuming the dist-tag conch of antd currently binding to the exact version 4.20.7.

How

Current Behaviour

  1. [x] Running npm install antd@conch should save the semantic versioning into package.json, hence ^4.20.7 shall be saved;
  2. [x] Running npm install antd@conch --save-exact should save the exact version into package.json, hence 4.20.7 shall be saved.

Desired Behaviour

  1. [ ] Running npm install antd@conch --force should save the dist-tag into package.json, hence conch shall be saved;

References

wesleytodd commented 2 years ago

npm install antd@conch --force

I think that another flag would be my preference. I think --force in this case is ambiguous and since it is a common flag might already have some associated behavior (although I cannot think of one off the top of my head).

Something like --no-resolve-dist-tag? IDK, that seems weird to, but just thinking something which is more explicit to the expected behavior for this feature would be good.

EDIT: Sorry I just noticed you opened a PR, and I think I also see that you did update the flag. Should we close this issue in favor of the RFC itself?

gemwuu commented 2 years ago

npm install antd@conch --force

I think that another flag would be my preference. I think --force in this case is ambiguous and since it is a common flag might already have some associated behavior (although I cannot think of one off the top of my head).

Something like --no-resolve-dist-tag? IDK, that seems weird to, but just thinking something which is more explicit to the expected behavior for this feature would be good.

EDIT: Sorry I just noticed you opened a PR, and I think I also see that you did update the flag. Should we close this issue in favor of the RFC itself?

Of course