URLs and scope names about import-maps HOT 9 CLOSED

I am on vacation through April 1, but would be happy to review PRs clarifying this or other aspects in the meantime.

from import-maps.

Since a scope defines a path and is a prefix of the refererURL, it makes sense to me that paths within a scope are relative to the path of the scope.

As you point out, you can refer outside of the scope path by using ../:

{
  "path_prefix": "https://foo.bar",
  "packages": {
    "A": { "path": "A@1", "main": "index.js" },
    "B": { "path": "B@1", "main": "index.js" }
  },
  "scopes": {
    "A": {
      "packages": {
        "C": { "path": "../C@1", "main": "index.js" }
      }
    },
    "B": {
      "packages": {
        "C": { "path": "../C@2", "main": "index.js" }
      }
    }
  }
}

from import-maps.

I find it really confusing, especially considering that this rule behavior is ambiguous in itself. Module names are not strictly correlated to the filesystem paths, so why make the scopes package paths relative to the scope names rather than the scopes paths? ie has a user, I could expect the following:

https://foo.bar/A@1/C@2/index.js
https://foo.bar/B@1/C@2/index.js

But not really this, since I never specified that A and B were valid URLs:

https://foo.bar/A/C@2/index.js
https://foo.bar/B/C@2/index.js

I can guess that the main argument for this behavior is that it's much simpler to just use the scope name rather than the package paths, but my point is that automatically composing path is a confusing behavior in itself, which counterbalances the "less repetition" argument. I'd prefer things to be explicit rather than implicit and subject to varying expectations.

from import-maps.

Scopes are independent from packages and don't have names - they are very explicitly and only paths. The reason is the role of a scope: it's used to look up a map based on the refererURL.

from import-maps.

I'm not sure I understand. Does that make my example above wrong, and should be this instead: ?

{
  ...,
  "scopes": {
    "A@1": {
       ...,
    },
    "B@1": {
       ...,
    }
  }
}

from import-maps.

Ah, yes. Though we should be careful to be clear with statements like "Resolving C from A". The "from" part really should be a path to be most clear.

from import-maps.

I see - if the scopes keys are URLs/paths, then I think the relative behavior semantically makes much more sense, thanks!

from import-maps.

cc @domenic for the discussion. Emphasizing that scopes are path-keyed and never referring to "scope name" would probably be helpful.

from import-maps.

This tripped me up too and seems like a common catch from the way the examples are described.

from import-maps.