Skip bypassing unapply for scala 2 case classes to allow for single-element named tuple in unapply by aherlihy · Pull Request #23603 · scala/scala3

aherlihy · 2025-07-25T18:47:03Z

From my understanding, for Scala 2 case classes, the compiler will generate an unapply that returns an Option[(...)] which contains the fields of the class as a tuple and this first case:

scala3/compiler/src/dotty/tools/dotc/transform/PatternMatcher.scala

Lines 346 to 356 in 6146b90

    
           if (isSyntheticScala2Unapply(unapp.symbol) && caseAccessors.length == args.length) 
        
             def tupleSel(sym: Symbol) = 
        
               // If scrutinee is a named tuple, cast to underlying tuple, so that we can 
        
               // continue to select with _1, _2, ... 
        
               ref(scrutinee).ensureConforms(scrutinee.info.stripNamedTuple).select(sym) 
        
             val isGenericTuple = defn.isTupleClass(caseClass) && 
        
               !defn.isTupleNType(tree.tpe match { case tp: OrType => tp.join case tp => tp }) // widen even hard unions, to see if it's a union of tuples 
        
             val components = 
        
               if isGenericTuple then caseAccessors.indices.toList.map(tupleApp(_, ref(scrutinee))) 
        
               else caseAccessors.map(tupleSel) 
        
             matchArgsPlan(components, args, onSuccess)

is to bypass this unapply and directly select the case class parameters for performance. However that causes the fixes added by this PR: #22953 to be skipped since the check comes first. I think the simplest fix is to skip the Scala 2 special case for a single-element selector, but not 100% sure.

…lement named tuple in unapply

smarter

I think this is fine as a workaround to avoid the runtime crash now, but it really needs a FIXME comment explaining the context and calling for a refactor of the code.

aherlihy · 2025-07-29T14:32:14Z

+        // enable selecting the field. See i23131.scala for test cases.
+        val wasSingleNamedArgForNamedTuple =
+          args.length == 1 && args.head.removeAttachment(FirstTransform.WasNamedArg).isDefined &&
+            isGetMatch(unappType) && unapp.select(nme.get, _.info.isParameterless).tpe.widenDealias.isNamedTupleType


@smarter I added a more precise check but I'm not really sure this is the best way to do it

I think this is fine, since .select(nme.get, _.info.isParameterless) appears both here and below in val get = ..., it could be factored out in a method getOfGetMatch or something like that.

Sure, I can add a method for that.

…r single-element named tuple in unapply" to 3.7.3 (#23645) Backports #23603 to the 3.7.3-RC1. PR submitted by the release tooling. [skip ci]

Skip bypassing unapply for scala 2 case classes to allow for single-e…

2183e53

…lement named tuple in unapply

aherlihy requested a review from smarter July 25, 2025 18:48

more precise check

4e684dc

Gedochao requested a review from odersky July 29, 2025 09:43

Gedochao assigned odersky Jul 29, 2025

smarter approved these changes Jul 29, 2025

View reviewed changes

Make check more precise for single named tuple selector in pattern match

7f95284

aherlihy commented Jul 29, 2025

View reviewed changes

minimize diff

3c5e4a0

aherlihy force-pushed the i23131 branch from c6b09bc to 3c5e4a0 Compare July 29, 2025 14:52

WojciechMazur added the backport:nominated If we agree to backport this PR, replace this tag with "backport:accepted", otherwise delete it. label Jul 29, 2025

WojciechMazur added this to the 3.7.3 milestone Jul 29, 2025

Refactor out getOfGetMatch

73da315

aherlihy enabled auto-merge July 30, 2025 08:47

aherlihy merged commit e0aca2d into scala:main Jul 30, 2025
44 of 45 checks passed

aherlihy deleted the i23131 branch July 30, 2025 14:09

WojciechMazur mentioned this pull request Aug 1, 2025

Backport "Skip bypassing unapply for scala 2 case classes to allow for single-element named tuple in unapply" to 3.7.3 #23645

Merged

WojciechMazur added backport:done This PR was successfully backported. and removed backport:nominated If we agree to backport this PR, replace this tag with "backport:accepted", otherwise delete it. labels Aug 1, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Skip bypassing unapply for scala 2 case classes to allow for single-element named tuple in unapply#23603

Skip bypassing unapply for scala 2 case classes to allow for single-element named tuple in unapply#23603
aherlihy merged 5 commits intoscala:mainfrom
aherlihy:i23131

aherlihy commented Jul 25, 2025

Uh oh!

smarter left a comment

Uh oh!

aherlihy Jul 29, 2025

Uh oh!

smarter Jul 29, 2025

Uh oh!

aherlihy Jul 30, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

	if (isSyntheticScala2Unapply(unapp.symbol) && caseAccessors.length == args.length)
	def tupleSel(sym: Symbol) =
	// If scrutinee is a named tuple, cast to underlying tuple, so that we can
	// continue to select with _1, _2, ...
	ref(scrutinee).ensureConforms(scrutinee.info.stripNamedTuple).select(sym)
	val isGenericTuple = defn.isTupleClass(caseClass) &&
	!defn.isTupleNType(tree.tpe match { case tp: OrType => tp.join case tp => tp }) // widen even hard unions, to see if it's a union of tuples
	val components =
	if isGenericTuple then caseAccessors.indices.toList.map(tupleApp(_, ref(scrutinee)))
	else caseAccessors.map(tupleSel)
	matchArgsPlan(components, args, onSuccess)

Conversation

aherlihy commented Jul 25, 2025

Uh oh!

smarter left a comment

Choose a reason for hiding this comment

Uh oh!

aherlihy Jul 29, 2025

Choose a reason for hiding this comment

Uh oh!

smarter Jul 29, 2025

Choose a reason for hiding this comment

Uh oh!

aherlihy Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants