lean-interact 0.5.0__tar.gz → 0.5.2__tar.gz

{lean_interact-0.5.0 → lean_interact-0.5.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lean-interact
-Version: 0.5.0
+Version: 0.5.2
 Summary: LeanInteract is a Python package that allows you to interact with the Lean theorem prover.
 Author-email: Auguste Poiroux <auguste.poiroux@epfl.ch>
 License: MIT License
@@ -49,7 +49,7 @@ Description-Content-Type: text/markdown
 - **🔗 Interactivity**: Execute Lean code and files directly from Python.
 - **🚀 Ease of Use**: LeanInteract abstracts the complexities of Lean setup and interaction.
 - **💻 Cross-platform**: Works on Windows, macOS, and Linux operating systems.
-- **🔧 Compatibility**: Supports all Lean versions between `v4.7.0-rc1` and `v4.19.0-rc2`.
+- **🔧 Compatibility**: Supports all Lean versions between `v4.7.0-rc1` and `v4.19.0`.
   - We backport the latest features of Lean REPL to older versions of Lean.
 - **📦 Temporary Projects**: Easily instantiate temporary Lean environments.
   - Useful for experimenting with benchmarks depending on [Mathlib](https://github.com/leanprover-community/mathlib4) like [ProofNet#](https://huggingface.co/datasets/PAug/ProofNetSharp) and [MiniF2F](https://github.com/yangky11/miniF2F-lean4).

{lean_interact-0.5.0 → lean_interact-0.5.2}/README.md

@@ -12,7 +12,7 @@
 - **🔗 Interactivity**: Execute Lean code and files directly from Python.
 - **🚀 Ease of Use**: LeanInteract abstracts the complexities of Lean setup and interaction.
 - **💻 Cross-platform**: Works on Windows, macOS, and Linux operating systems.
-- **🔧 Compatibility**: Supports all Lean versions between `v4.7.0-rc1` and `v4.19.0-rc2`.
+- **🔧 Compatibility**: Supports all Lean versions between `v4.7.0-rc1` and `v4.19.0`.
   - We backport the latest features of Lean REPL to older versions of Lean.
 - **📦 Temporary Projects**: Easily instantiate temporary Lean environments.
   - Useful for experimenting with benchmarks depending on [Mathlib](https://github.com/leanprover-community/mathlib4) like [ProofNet#](https://huggingface.co/datasets/PAug/ProofNetSharp) and [MiniF2F](https://github.com/yangky11/miniF2F-lean4).

{lean_interact-0.5.0 → lean_interact-0.5.2}/examples/beq_plus.py

@@ -156,14 +156,14 @@ def beql(
         except ValueError:
             if verbose:
                 console.print("Invalid theorems encountered, skipping this pair.")
-            continue
+            break
 
         formal_code = formal_1_code + formal_2_code
         # Preliminary check to ensure the formalization is well-typed.
         if check_proof_sub(server, formal_code, context_env, formal_2_start_line, "sorry", timeout_per_proof) is None:
             if verbose:
                 console.print("Ill-typed formalization encountered, skipping this pair.")
-            continue
+            break
 
         proof_exact = check_proof_sub(
             server, formal_code, context_env, formal_2_start_line, "exact?", timeout_per_proof
@@ -173,6 +173,8 @@ def beql(
             if verbose:
                 console.print("Proof exact")
                 console.print(Syntax(proof_exact, "lean4"))
+        else:
+            break
 
     return res[0] and res[1]
 
@@ -229,13 +231,13 @@ def beq_plus(
         except ValueError:
             if verbose:
                 console.print("Invalid theorem encountered, skipping this pair.")
-            continue
+            break
 
         formal_code = formal_1_code + formal_2_code
         if check_proof_sub(server, formal_code, context_env, formal_2_start_line, "sorry", timeout_per_proof) is None:
             if verbose:
                 console.print("Ill-typed formalization encountered, skipping this pair.")
-            continue
+            break
 
         # 1. Use BEqL
         proof_exact = check_proof_sub(
@@ -322,6 +324,9 @@ def beq_plus(
                     console.print(Syntax(proof_convert, "lean4"))
                 break
 
+        if not res[i]:
+            break
+
     return res[0] and res[1]
 
 
@@ -382,22 +387,35 @@ open scoped BigOperators"""
     ]
 
     console.print(f"{metric.__name__} metric on examples:")
+    counter_equivalent = 0
     for formalization_1, formalization_2 in formalization_pairs:
         console.print()
         console.rule()
         console.print("Comparing formalizations:")
         console.print(Syntax(formalization_1, "lean4"))
         console.print(Syntax(formalization_2, "lean4"))
-        console.print(
-            f"Proved equivalent: {metric(formalization_1, formalization_2, src_header, repl_config, timeout_per_proof=DEFAULT_TIMEOUT, verbose=True)}"
+        equivalent = metric(
+            formalization_1,
+            formalization_2,
+            src_header,
+            repl_config,
+            timeout_per_proof=DEFAULT_TIMEOUT,
+            verbose=True,
         )
+        counter_equivalent += equivalent
+        if equivalent:
+            console.print("[green]Proved equivalent[/]")
+        else:
+            console.print("[red]Equivalence not proven[/]")
+    console.print()
+    console.print(f"Total proved equivalent: {counter_equivalent}/{len(formalization_pairs)}")
 
 
-def proofnetverif(metric):
+def proofnetverif(metric, n_samples=100):
     repl_config = LeanREPLConfig(lean_version="v4.8.0", project=TempRequireProject("mathlib"), verbose=True)
 
     dataset = load_dataset("PAug/ProofNetVerif", split="valid")
-    dataset = dataset.shuffle(seed=42).select(range(100))
+    dataset = dataset.shuffle(seed=42).select(range(n_samples))
 
     metric_results = []
     for example in tqdm(dataset, desc=f"`{metric.__name__}` metric on ProofNetVerif dataset"):

{lean_interact-0.5.0 → lean_interact-0.5.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "lean-interact"
-version = "0.5.0"
+version = "0.5.2"
 description = "LeanInteract is a Python package that allows you to interact with the Lean theorem prover."
 keywords = ["Lean", "theorem proving", "autoformalization", "REPL"]
 license = { file = "LICENSE" }

{lean_interact-0.5.0 → lean_interact-0.5.2}/src/lean_interact/config.py

@@ -59,6 +59,7 @@ class LocalProject(BaseProject):
 
         with FileLock(f"{self.directory}.lock"):
             try:
+                subprocess.run(["lake", "exe", "cache", "get"], cwd=self.directory, check=False, stdout=stdout, stderr=stderr)
                 subprocess.run(["lake", "build"], cwd=self.directory, check=True, stdout=stdout, stderr=stderr)
             except subprocess.CalledProcessError as e:
                 logger.error("Failed to build local project: %s", e)
@@ -102,6 +103,7 @@ class GitProject(BaseProject):
             repo.submodule_update(init=True, recursive=True)
 
             try:
+                subprocess.run(["lake", "exe", "cache", "get"], cwd=project_dir, check=False, stdout=stdout, stderr=stderr)
                 subprocess.run(["lake", "build"], cwd=project_dir, check=True, stdout=stdout, stderr=stderr)
             except subprocess.CalledProcessError as e:
                 logger.error("Failed to build the git project: %s", e)

{lean_interact-0.5.0 → lean_interact-0.5.2}/src/lean_interact/server.py

@@ -71,6 +71,7 @@ class LeanServer:
             encoding="utf-8",
             text=True,
             bufsize=1,
+            start_new_session=True,
             preexec_fn=None
             if platform.system() != "Linux"
             else lambda: _limit_memory(self.config.memory_hard_limit_mb),
@@ -86,15 +87,24 @@ class LeanServer:
 
     def kill(self) -> None:
         if self._proc:
-            # Close file handles before killing the process
-            if self._proc.stdin:
-                self._proc.stdin.close()
-            if self._proc.stdout:
-                self._proc.stdout.close()
-            if self._proc.stderr:
-                self._proc.stderr.close()
-            self._proc.kill()
-            self._proc.wait()
+            try:
+                proc = psutil.Process(self._proc.pid)
+                # Terminate the process tree
+                children = proc.children(recursive=True)
+                for child in children:
+                    try:
+                        child.terminate()
+                    except Exception:
+                        pass
+                proc.terminate()
+                _, alive = psutil.wait_procs([proc] + children, timeout=1)
+                for p in alive:
+                    try:
+                        p.kill()
+                    except Exception:
+                        pass
+            except Exception:
+                pass
             self._proc = None
         gc.collect()
 

{lean_interact-0.5.0 → lean_interact-0.5.2}/src/lean_interact/utils.py

@@ -20,7 +20,7 @@ logger.addHandler(handler)
 ROOT_DIR = os.path.dirname(os.path.abspath(__file__))
 DEFAULT_CACHE_DIR = os.path.join(ROOT_DIR, "cache")
 DEFAULT_REPL_GIT_URL = "https://github.com/augustepoiroux/repl"
-DEFAULT_REPL_VERSION = "v1.0.6"
+DEFAULT_REPL_VERSION = "v1.0.8"
 
 os.makedirs(DEFAULT_CACHE_DIR, exist_ok=True)
 
@@ -334,7 +334,7 @@ def clean_theorem_string(theorem_string: str, new_theorem_name: str = "dummy", a
             clean_formal = re.sub(r"^[^\s:({\[]+", "", clean_formal).strip()
         clean_formal = f"theorem {new_theorem_name} " + clean_formal
         if add_sorry:
-            clean_formal += " := sorry"
+            clean_formal += " := by sorry"
         return clean_formal
     except Exception:
         return None

{lean_interact-0.5.0 → lean_interact-0.5.2}/tests/test_server.py

@@ -295,7 +295,7 @@ lean_exe "dummy" where
                 ],
             ),
         )
-        result = server.run(ProofStep(tactic="apply irrational_add_rat_iff.mpr", proof_state=0), verbose=True)
+        result = server.run(ProofStep(tactic="apply irrational_add_ratCast_iff.mpr", proof_state=0), verbose=True)
         self.assertEqual(result, ProofStepResponse(proof_state=1, goals=[], proof_status="Completed"))
 
     def test_restart_with_env(self):
@@ -416,6 +416,32 @@ lean_exe "dummy" where
         with self.assertRaises(ChildProcessError):
             server.run(Command(cmd="def z := 3"), verbose=True)
 
+    def test_timeout_respected(self):
+        if platform.system() == "Windows":
+            self.skipTest("(Temporary) Skipping test on Windows due to long path issues in the CI")
+
+        config = LeanREPLConfig(project=TempRequireProject("mathlib"))
+        server = AutoLeanServer(config)
+
+        response = server.run(
+            Command(cmd="import Mathlib\nset_option maxHeartbeats 0\nset_option maxRecDepth 100000"),
+            add_to_session_cache=True,
+        )
+        assert isinstance(response, CommandResponse)
+        root_env = response.env
+
+        # check that the next command takes less than 3 seconds
+        start = time.time()
+        with self.assertRaises(TimeoutError):
+            server.run(
+                Command(
+                    cmd="theorem amc12a_2003_p1 (u v : ℕ → ℕ) (h₀ : ∀ n, u n = 2 * n + 2) (h₁ : ∀ n, v n = 2 * n + 1) :\n    ((∑ k in Finset.range 2003, u k) - ∑ k in Finset.range 2003, v k) = 2003 := by simp only [h₀, h₁, Finset.sum_range_succ, Finset.sum_range_zero]; rfl",
+                    env=root_env,
+                ),
+                timeout=2,
+            )
+        self.assertLess(time.time() - start, 3)
+
     # def test_run_proof(self):
     #     server = AutoLeanServer(config=LeanREPLConfig(verbose=True))
     #     result = server.run(