Imports that are resolvable should not be reported as builtins

[rixth]

constants is a builtin (apparently) in node. however, if we can resolve the module name, we should assume that it is not a builtin.

Perhaps we could create a second rule that enforces not clobbering builtin names, but that behaviour shouldn't be part of this rule which is to enforce ordering.

[coveralls]

Coverage increased (+0.007%) to 96.232% when pulling 27a5fad on rixth:improve_type_detection into b08bd3e on benmosher:master.

[coveralls]

Coverage increased (+0.007%) to 96.232% when pulling 27a5fad on rixth:improve_type_detection into b08bd3e on benmosher:master.

[coveralls]

Coverage increased (+0.007%) to 96.232% when pulling 27a5fad on rixth:improve_type_detection into b08bd3e on benmosher:master.

[rixth]

Not sure what the difference is between the travis build & the appveyor build (nor why the latter is failing)

[ljharb]

however, if we can resolve the module name, we should assume that it is not a builtin.

Can you elaborate on this? In node, builtin names (like constants) take precedence no matter what's on the filesystem.

[j-f1]

@ljharb is correct:

Core modules are always preferentially loaded if their identifier is passed to require(). For instance, require('http') will always return the built in HTTP module, even if there is a file by that name.
^source

[rixth]

@ljharb that may be so for just constants, but deeper paths eg constants/foo will be required as expected

[ljharb]

Hmm, that's true - also constants/.

I wonder if it'd be simpler to change it so that if there's any path at all - ie, anything with a slash - that it's not a builtin?

[rixth]

@ljharb Happy to make this change -- would that be acceptable to you, or is there another contributor we should run this past?

[ljharb]

I'd be happy to hear from @jfmengels or @benmosher on it, but both of them have been taking open source breaks for awhile now.

If there aren't any node builtins that have a / in the name, then this seems like the right way to address the bug.

[benmosher]

I wonder if it'd be simpler to change it so that if there's any path at all - ie, anything with a slash - that it's not a builtin?

this makes sense to me if it matches the spec 👍

[ljharb]

@rixth are you no longer interested in updating this PR?

[rixth]

@ljharb Correct - it is over a year old, conflicting, and I no longer work at the company/codebase where this was problematic. It was always an edge case at best.