5bb7f37645
This PR adds a new `first_par` tactic combinator that runs multiple tactics in parallel and returns the first successful result (cancelling the others). The `try?` tactic's `atomicSuggestions` step now uses `first_par` to try three grind variants in parallel: - `grind? +suggestions` - uses library suggestion engine - `grind? +locals` - unfolds local definitions from current file - `grind? +locals +suggestions` - combines both This leverages `TacticM.parFirst` which already provides the "first success wins" parallel execution with cancellation. ### Depends on - [x] depends on: #11946 🤖 Prepared with Claude Code --------- Co-authored-by: Claude <noreply@anthropic.com>
43 lines
1.3 KiB
Text
43 lines
1.3 KiB
Text
import Lean.LibrarySuggestions.Default
|
||
|
||
/-!
|
||
# Test for `first_par` combinator in try?
|
||
|
||
This tests that `try?` uses `first_par` internally to try multiple grind variants
|
||
(`grind? +suggestions`, `grind? +locals`, `grind? +locals +suggestions`) in parallel
|
||
and return the first successful result.
|
||
|
||
The test creates a scenario where:
|
||
1. Basic tactics (rfl, assumption, simp) fail
|
||
2. Basic grind fails
|
||
3. But `grind +locals +suggestions` succeeds by finding a local theorem via library search
|
||
-/
|
||
|
||
structure MyPair (α : Type) where
|
||
fst : α
|
||
snd : α
|
||
|
||
def foo (x : MyPair Nat) := x
|
||
|
||
-- A theorem about MyPair that grind +suggestions can find via library search
|
||
theorem myPair_eq (p : MyPair Nat) (h1 : p.fst = 1) (h2 : p.snd = 1) :
|
||
p = foo ⟨1, 1⟩ := by
|
||
cases p; simp_all [foo]
|
||
|
||
-- A goal where:
|
||
-- - atomic block fails (no simple solution, basic grind can't prove it)
|
||
-- - first_par succeeds: grind +locals +suggestions finds myPair_eq
|
||
-- The +locals +suggestions flags are filtered out in the suggestion output
|
||
/--
|
||
info: Try these:
|
||
[apply] grind only [myPair_eq, foo]
|
||
[apply] grind =>
|
||
instantiate only [myPair_eq]
|
||
instantiate only [foo]
|
||
-/
|
||
#guard_msgs in
|
||
example (p : MyPair Nat) (h1 : p.fst = 1) (h2 : p.snd = 1) : p = ⟨1, 1⟩ := by
|
||
fail_if_success grind
|
||
fail_if_success grind +locals
|
||
fail_if_success grind +suggestions
|
||
try?
|