[mlir] Python: Parse ModuleOp from file path

For extremely large models, it may be inefficient to load the model into memory in Python prior to passing it to the MLIR C APIs for deserialization. This change adds an API to parse a ModuleOp directly from a file path.

Author: nikalra

mlir/include/mlir-c/IR.h

@@ -309,6 +309,10 @@ MLIR_CAPI_EXPORTED MlirModule mlirModuleCreateEmpty(MlirLocation location);
 MLIR_CAPI_EXPORTED MlirModule mlirModuleCreateParse(MlirContext context,
                                                     MlirStringRef module);
 
+/// Parses a module from file and transfers ownership to the caller.
+MLIR_CAPI_EXPORTED MlirModule
+mlirModuleCreateParseFromFile(MlirContext context, MlirStringRef fileName);
+
 /// Gets the context that a module was created with.
 MLIR_CAPI_EXPORTED MlirContext mlirModuleGetContext(MlirModule module);
 

mlir/include/mlir/Bindings/Python/Nanobind.h

@@ -23,6 +23,7 @@
 #endif
 #include <nanobind/nanobind.h>
 #include <nanobind/ndarray.h>
+#include <nanobind/stl/filesystem.h>
 #include <nanobind/stl/function.h>
 #include <nanobind/stl/optional.h>
 #include <nanobind/stl/pair.h>

mlir/lib/Bindings/Python/IRCore.cpp

@@ -6,6 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include <filesystem>
 #include <optional>
 #include <utility>
 
@@ -299,7 +300,7 @@ struct PyAttrBuilderMap {
     return *builder;
   }
   static void dunderSetItemNamed(const std::string &attributeKind,
-                                nb::callable func, bool replace) {
+                                 nb::callable func, bool replace) {
     PyGlobals::get().registerAttributeBuilder(attributeKind, std::move(func),
                                               replace);
   }
@@ -3047,6 +3048,19 @@ void mlir::python::populateIRCore(nb::module_ &m) {
           },
           nb::arg("asm"), nb::arg("context").none() = nb::none(),
           kModuleParseDocstring)
+      .def_static(
+          "parse",
+          [](const std::filesystem::path &path,
+             DefaultingPyMlirContext context) {
+            PyMlirContext::ErrorCapture errors(context->getRef());
+            MlirModule module = mlirModuleCreateParseFromFile(
+                context->get(), toMlirStringRef(path.string()));
+            if (mlirModuleIsNull(module))
+              throw MLIRError("Unable to parse module assembly", errors.take());
+            return PyModule::forModule(module).releaseObject();
+          },
+          nb::arg("asm"), nb::arg("context").none() = nb::none(),
+          kModuleParseDocstring)
       .def_static(
           "create",
           [](DefaultingPyLocation loc) {

mlir/lib/CAPI/IR/IR.cpp

@@ -22,6 +22,7 @@
 #include "mlir/IR/Location.h"
 #include "mlir/IR/Operation.h"
 #include "mlir/IR/OperationSupport.h"
+#include "mlir/IR/OwningOpRef.h"
 #include "mlir/IR/Types.h"
 #include "mlir/IR/Value.h"
 #include "mlir/IR/Verifier.h"
@@ -328,6 +329,15 @@ MlirModule mlirModuleCreateParse(MlirContext context, MlirStringRef module) {
   return MlirModule{owning.release().getOperation()};
 }
 
+MlirModule mlirModuleCreateParseFromFile(MlirContext context,
+                                         MlirStringRef fileName) {
+  OwningOpRef<ModuleOp> owning =
+      parseSourceFile<ModuleOp>(unwrap(fileName), unwrap(context));
+  if (!owning)
+    return MlirModule{nullptr};
+  return MlirModule{owning.release().getOperation()};
+}
+
 MlirContext mlirModuleGetContext(MlirModule module) {
   return wrap(unwrap(module).getContext());
 }

mlir/python/mlir/_mlir_libs/_mlir/ir.pyi

@@ -46,6 +46,7 @@ import abc
 import collections
 from collections.abc import Callable, Sequence
 import io
+from pathlib import Path
 from typing import Any, ClassVar, TypeVar, overload
 
 __all__ = [
@@ -2123,7 +2124,7 @@ class Module:
         Creates an empty module
         """
     @staticmethod
-    def parse(asm: str | bytes, context: Context | None = None) -> Module:
+    def parse(asm: str | bytes | Path, context: Context | None = None) -> Module:
         """
         Parses a module's assembly format from a string.