Misleading documentation of Range.map - it returns an IndexedSeq not a Range

[bjornregnell]

Problem

Copied from the doc of 3.2.0: https://www.scala-lang.org/api/3.2.0/scala/collection/immutable/Range.html#

The doc says map on Range returns a new Range but it returns an IndexedSeq (which defaults to Vector).

[bjornregnell]

The implementation is here:
https://github.com/scala/scala/blob/c7aef6aac6487e4f59233662c0d9da3c7852b7c4/src/library/scala/collection/immutable/Range.scala#L150
But I cannot find the actual doc comment...

[SethTisue]

The doc comment is inherited from IterableOnce

[SethTisue]

Any fix should presumably involve also fixing any other methods, e.g. filter. Perhaps they don't need to be fixed one at a time, but can be adjusted using the inheritance mechanism? @scala/collections

[som-snytt]

There is a distinction between the C and the CC. This came up recently in the "infer the override" feature, where the C was overridden by CC (which was Set) even though C is more specific. At the scaladoc level, there is just the $coll.

Also, per the original question about presenting types to students in REPL, it could do something nice about what to display (as the type). That is, what the REPL reports is not tied inexorably to the API. It could have a special mode or plugin for specialized reporting.

Also it's IterableOnceOps, as visible in "definition classes".

[bjornregnell]

Could a fix just bolt a concrete doc comment onto the actual map in the actual Range.scala file? There are other doc comments in the same file...

[bjornregnell]

Also, per the original question about presenting types to students in REPL, it could do something nice about what to display (as the type). That is, what the REPL reports is not tied inexorably to the API. It could have a special mode or plugin for specialized reporting.

Would not the REPL lie then? I mean, when coding in eg. vs code with Metals the user still sees the discrepancy between IndexedSeq and the actual Vector and scratches its head. I can only see that being solved by nailing the return type of Range.map to Vector but then loose the flexibility to be able to upgrade some time later to the super-fast-not-yet-invented-collection-type-always-better-than-good-old-Vector.

[som-snytt]

REPL already lies by filtering output.

scala> class C
class C

scala> new C
val res0: C = C@115ca7de

scala> new C().getClass
val res1: Class[_ <: C] = class C

scala> new C().getClass.toString.toUpperCase
val res2: String = CLASS $LINE3.$READ$$IW$C

scala>

[som-snytt]

This is a great first issue, I've learned so much, I feel like the newbie that I am.

[som-snytt]

I introduced @define ccoll $coll for this purpose, but I just noticed that $Coll is not a typo.

Here are the backticks leaking in a code block. Probably sufficient to strip them in a code block. Conversely, never interpolate in a backticked phrase, which currently warns for bad var. Interpolated code should always be in the triple braces.

I was momentarily confused about where Range sets its Coll, because it doesn't:

I will retain coll for natural language term for C, ccoll for CC, and Coll for the class for CC.

[som-snytt]

I was worried about breaking the doc, so it's a relief to see that it's already a bit frayed at the edges.