RFC change semantics for parens-less 0 argument templates

[timotheecour]

current implementation implicitly resolves 0-arg templates as a call; IMO it's a design mistake that should be fixed, and it should require a pragma {.implicitcall.} eg:

template fun(): string = "foo"
doAssert fun() == "foo" # ok
doAssert fun == "foo" # should CT error
template fun2(): string {.implicitcall.} = "foo"
doAssert fun2() == "foo" # should CT error
doAssert fun2 == "foo" # ok

alternative proposal (just different syntax, same idea)

instead of {.implicitcall.}, use template fun2 without parens:

template fun(): string = "foo"
doAssert fun() == "foo" # ok
doAssert fun == "foo" # should CT error
template fun2: string = "foo"
doAssert fun2() == "foo" # should CT error
doAssert fun2 == "foo" # ok

links

• template treated as template(): untyped when using in template without bracket · Issue #14330 · nim-lang/Nim
• Scoping and overload resolution nim-lang/compilerdev#5 (comment)
• Calling nullary templates without () doesn't work inside calls inside other templates nim-lang/Nim#13515