Disable loading of all modules in astroid

Author: chvp

src/backend/workers/python/papyros/linting.py

@@ -7,32 +7,18 @@
 from pylint.lint import Run
 from pylint.reporters.text import TextReporter
 
-# Workaround for Pyodide + Python 3.12 + astroid 2.15.8: pylint hangs
-# indefinitely when analyzing code that imports `re` because astroid
-# recursively parses the stdlib `re` package. Short-circuit astroid's
-# module loading for `re` (and related stdlib modules) to return a
-# tiny synthetic module instead.
+# Workaround for Pyodide + astroid: pylint hangs indefinitely when astroid
+# tries to recursively parse imported modules' ASTs (e.g. re, pandas).
+# Since we run in a single-threaded WebAssembly environment, this causes an
+# infinite hang. We short-circuit ALL module resolution to return synthetic
+# empty modules. This preserves core linting (syntax errors, undefined
+# variables, unused imports, style checks, custom checkers) while only
+# losing type-inference-based checks on imported symbols.
 from astroid.manager import AstroidManager as _AstroidManager
 from astroid.builder import AstroidBuilder as _AstroidBuilder
 
-_BLOCKED_MODULES = {
-    "re",
-    "re._compiler",
-    "re._parser",
-    "re._constants",
-    "re._casefix",
-    "sre_compile",
-    "sre_parse",
-    "sre_constants",
-}
-
-_orig_ast_from_module_name = _AstroidManager.ast_from_module_name
-
 def _patched_ast_from_module_name(self, modname, context_file=None, use_cache=True):
-    if modname in _BLOCKED_MODULES:
-        # Return an empty synthetic module so astroid's inference short-circuits.
-        return _AstroidBuilder(self).string_build("", modname=modname)
-    return _orig_ast_from_module_name(self, modname, context_file=context_file, use_cache=use_cache)
+    return _AstroidBuilder(self).string_build("", modname=modname)
 
 _AstroidManager.ast_from_module_name = _patched_ast_from_module_name
 

src/backend/workers/python/papyros/pylint_config.rc

@@ -43,5 +43,5 @@ const-rgx=[_A-Za-z0-9]{1,30}$
 # I0011 Warning locally suppressed using disable-msg
 # I0012 Warning locally suppressed using disable-msg
 # old version: disable=I0011,I0012,W0704,W0142,W0212,W0232,W0702,R0201,W0614,R0914,R0912,R0915,R0913,R0904,R0801,C0303,C0111,C0304,R0903,W0141,W0621,C0301,W0631,R0911,C1001
-disable=W0311,W0621,W0622,R0902,R0903,C0111,C0301,C0303,C0304,C0413,I0011
+disable=W0311,W0621,W0622,R0902,R0903,C0111,C0301,C0303,C0304,C0413,I0011,E0401,E0611,E1101
 evaluation=max(10.0 - ((float(5 * error + warning + refactor + convention) / statement) * 10), 0)

test/__tests__/state/Runner.test.ts

@@ -71,6 +71,15 @@ const c = a + b;`;
         expect(Array.isArray(diagnostics)).toBe(true);
     }, 60000);
 
+    it("should lint code that uses pandas without hanging", async () => {
+        const papyros = new Papyros();
+        await papyros.launch();
+        papyros.runner.programmingLanguage = ProgrammingLanguage.Python;
+        papyros.runner.code = "import pandas as pd\ndf = pd.DataFrame({'a': [1, 2, 3]})\n";
+        const diagnostics = await papyros.runner.lintSource();
+        expect(Array.isArray(diagnostics)).toBe(true);
+    }, 60000);
+
     it("should be able to handle sleep", async () => {
         const papyros = new Papyros();
         await papyros.launch();