1129160d80
Previously the formatter was using the builtin token table rather that the one in the current environment. This could lead to round-tripping failures for user-defined notations. For an illustrative example, given the following notation ```lean infixl:65 "+'" => Int.add notation:65 a:65 "+'-" b:66 => Int.add a (id b) ``` then `5 +' -1` would parse as `Int.add 5 (-1)` and incorrectly pretty print as `5+'-1`, which in turn would parse as `Int.add 5 (id 1)`. Now it pretty prints as `5+' -1`.
16 lines
456 B
Text
16 lines
456 B
Text
/-!
|
|
# Verify that the formatter uses the current token table
|
|
|
|
Formerly, the formatter was only looking at the builtin token table
|
|
when deciding where to insert additional whitespace between tokens.
|
|
This lead to `5+' -1` in the following pretty printing as `5+'-1`.
|
|
-/
|
|
|
|
infixl:65 "+'" => Int.add
|
|
notation:65 a:65 "+'-" b:66 => Int.add a (id b)
|
|
|
|
/-- info: 5+' -1 : Int -/
|
|
#guard_msgs in #check 5 +' -1
|
|
|
|
/-- info: 5+'-1 : Int -/
|
|
#guard_msgs in #check 5 +'- 1
|