This PR turns on the new `do` elaborator in Init, Lean, Std, Lake and the testsuite. --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
142 lines
3.3 KiB
Text
142 lines
3.3 KiB
Text
module
|
||
|
||
import Lean.Elab.Command
|
||
import Lean.Elab.Syntax
|
||
import all Lean.Elab.Syntax
|
||
|
||
abbrev isValidAtom := Lean.Elab.Term.toParserDescr.isValidAtom
|
||
|
||
open Lean Elab Command
|
||
|
||
/-!
|
||
Test various classes of atoms that should be allowed or not.
|
||
-/
|
||
|
||
def test (expected : Bool) (strings : List String) : CommandElabM Unit := do
|
||
let mut wrong : List String := []
|
||
for s in strings do
|
||
if isValidAtom s != expected then
|
||
wrong := s :: wrong
|
||
if isValidAtom (" " ++ s) != expected then
|
||
wrong := s!"{s} (with leading whitespace)" :: wrong
|
||
if isValidAtom (s ++ " ") != expected then
|
||
wrong := s!"{s} (with trailing whitespace)" :: wrong
|
||
if isValidAtom (" " ++ s ++ " ") != expected then
|
||
wrong := s!"{s} (with leading and trailing whitespace)" :: wrong
|
||
|
||
if wrong.isEmpty then
|
||
logInfo <| "All " ++ if expected then "accepted" else "rejected"
|
||
else
|
||
logError <|
|
||
s!"The following atoms should be {if expected then "" else "in"}valid but are not:\n" ++
|
||
String.join (wrong.reverse.map (s! " * {·}\n"))
|
||
|
||
|
||
syntax "#test_valid" term : command
|
||
syntax "#test_invalid" term : command
|
||
|
||
macro_rules
|
||
| `(#test_valid%$tok $t) => `(#eval%$tok test true $t)
|
||
| `(#test_invalid%$tok $t) => `(#eval%$tok test false $t)
|
||
|
||
|
||
/-!
|
||
# No Empty Atoms
|
||
-/
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid [""]
|
||
|
||
|
||
/-!
|
||
# Preventing Character Literal Confusion
|
||
|
||
Atoms are required to be suitably unlike character literals. This means that if they start with a
|
||
single quote, the next character must also be a single quote.
|
||
|
||
Two single quotes (and variations on it) has long-term usage as an infix operator in Mathlib.
|
||
-/
|
||
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["if", "''", "''ᵁ", "if'", "x'y'z", "x''y"]
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["'x'", "'ᵁ", "'c", "'no'", "'"]
|
||
|
||
|
||
/-!
|
||
# No Internal Whitespace
|
||
-/
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["open mixed", "open mixed"]
|
||
|
||
|
||
/-!
|
||
# No Confusion with String Literals
|
||
|
||
Internal double quotes are allowed.
|
||
-/
|
||
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["what\"string\"is_this?"]
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["\"","\"\"", "\"f\""]
|
||
|
||
/-!
|
||
# No Confusion with Escaped Identifiers
|
||
|
||
This doesn't confuse the parser, but it may confuse a user if they can define an atom that can never
|
||
be parsed.
|
||
-/
|
||
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["prefix«", "prefix«test", "prefix«test»", "prefix«test»foo"]
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["«", "«test", "«test»", "«test»foo"]
|
||
|
||
|
||
/-!
|
||
# No Confusion with Name Literals
|
||
|
||
The first two characters of an atom may not be a valid start of a name literal
|
||
-/
|
||
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["``", "`!", "x`"]
|
||
|
||
/-!
|
||
The next set all begin with U0x2035, REVERSED PRIME, rather than back-tick, and are thus accepted.
|
||
-/
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["‵", "‵x", "‵«", "‵xyz", "‵«x.y", "‵«x.y.z»"]
|
||
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["`", "`x", "`«", "`xyz", "`«x.y", "`«x.y.z»"]
|
||
|
||
|
||
/-!
|
||
# No Leading Digits
|
||
-/
|
||
|
||
/-- info: All accepted -/
|
||
#guard_msgs in
|
||
#test_valid ["prefix5", "prefix22test", "prefix1test0", "prefix8test8foo"]
|
||
|
||
/-- info: All rejected -/
|
||
#guard_msgs in
|
||
#test_invalid ["0", "1test", "0test3"]
|