A strawman for aggregate literals #21993
Conversation
odersky
force-pushed
the
try-seqliterals
branch
from
a728906 to
6e904fb
Compare
|
From the doc I understand that implicit conversions are not applicable. Is weak conformance applicable? |
|
The successor of weak conformance (int literal widening) is usually applicable. That follows from the fact that SeqLiterals expand to varargs and varargs do perform that widening. |
docs/_docs/internals/syntax.md
Outdated
| @@ -278,6 +278,7 @@ SimpleExpr ::= SimpleRef | |||
| | ‘new’ ConstrApp {‘with’ ConstrApp} [TemplateBody] New(constr | templ) | |||
| | ‘new’ TemplateBody | |||
| | ‘(’ ExprsInParens ‘)’ Parens(exprs) | |||
| | ‘[’ ExprInBrackets ‘)’ SeqLiteral(exprs, TypeTree()) | |||
There was a problem hiding this comment.
| | ‘[’ ExprInBrackets ‘)’ SeqLiteral(exprs, TypeTree()) | |
| | ‘[’ ExprInBrackets ‘]’ SeqLiteral(exprs, TypeTree()) |
just noticed while having a look
odersky
force-pushed
the
try-seqliterals
branch
from
3830475 to
599d51c
Compare
|
I got a crash in the repl by just having |
|
Also, maybe this is expected as its just for testing, but some of the instances for scala> val m: Set[Int] = [1,2,3]
1 warning found
-- Warning: --------------------------------------------------------------------
1 |val m: Set[Int] = [1,2,3]
| ^
|Given search preference for scala.compiletime.ExpressibleAsCollectionLiteral[Set[Int]] between alternatives
| scala.compiletime.ExpressibleAsCollectionLiteral.bitsetFromLiteral.type
|and
| (scala.compiletime.ExpressibleAsCollectionLiteral.setFromLiteral :
| [T]: scala.compiletime.ExpressibleAsCollectionLiteral.setFromLiteral[T])
|will change.
|Current choice : the first alternative
|New choice from Scala 3.7: none - it's ambiguous
val m: Set[Int] = BitSet(1, 2, 3) |
|
@bishabosha In fact, the given instances are set up so that it works properly only from 3.7 onwards. |
Does that mean the warning is a false positive? |
|
No, the warning is real. The given instances are set up so that they work properly only from 3.7. |
A test case that shows that we can have an "inline type class" that allows to use a typeclass-based scheme for sequence literals where instances can be created with macros.
Also: Move test to run
We now don't try to instantiate selected type variables. Instead, we use a default as fallback if the expected type is underspecified according to the definition in Implicits. This is simpler and more expressive.
Also, test infix operations taking [...] as a right operand.
odersky
force-pushed
the
try-seqliterals
branch
from
28700de to
4b8a832
Compare
odersky
added
the
needs-sip
|
See https://contributors.scala-lang.org/t/pre-sip-a-syntax-for-collection-literals/6990/186 for the Pre SIP discussion. |
|
Closing to relieve the PR queue. We can come back to this later. |
A test case that shows that we can have an "inline type class" that allows to use a typeclass-based scheme for sequence literals where instances can be created with macros.