nimpy

Native language integration with Python has never been easier!

Implementing a Python Module in Nim

# mymodule.nim - file name should match the module name you're going to import from python
import nimpy

proc greet(name: string): string {.exportpy.} =
  return "Hello, " & name & "!"

# Compile on Windows:
nim c --app:lib --out:mymodule.pyd --threads:on --tlsEmulation:off --passL:-static mymodule
# Compile on everything else:
nim c --app:lib --out:mymodule.so --threads:on mymodule

# test.py
import mymodule
assert mymodule.greet("world") == "Hello, world!"
assert mymodule.greet(name="world") == "Hello, world!"

Calling Python From Nim

import nimpy
let os = pyImport("os")
echo "Current dir is: ", os.getcwd().to(string)

# sum(range(1, 5))
let py = pyBuiltinsModule()
let s = py.sum(py.range(0, 5)).to(int)
assert s == 10

Note: here nimpy relies on your local python installation.

Importing Nim Extensions Directly

For a convenient way to import your Nim extension modules directly, you can use Nimporter.

Misc

The library is designed with ABI compatibility in mind. That is the compiled module doesn't depend on particular Python version, it should properly work with any. The C API symbols are loaded in runtime from whichever process has launched your module.

Troubleshooting, Q&A

Importing the compiled module from Python fails

If you're getting ImportError: dynamic module does not define module export function ... make sure that the module you're importing from Python has exactly the same name as the nim file which the module is implemented in.

Nimpy fails to find (proper) libpython

The most reliable way to find libpython is find_libpython python package:

pip3 install find_libpython
python3 -c 'import find_libpython; print(find_libpython.find_libpython())'

Then you can specify path to libpython using nimpy.py_lib.pyInitLibPath. Tracking issue: #171.

Nim strings are converted to Python bytes instead of string

nimpy converts Nim strings to Python strings usually, but since Nim strings are encoding agnostic and may contain invalid utf8 sequences, nimpy will fallback to Python bytes in such cases.

Is there any numpy compatibility?

nimpy allows manipulating numpy objects just how you would do it in Python, however it is not much more efficient. scinim offers API for performance critical numpy interop, and it is advised to consider it first.

Nimpy also exposes lower level Buffer protocol, see raw_buffers.nim. tpyfromnim.nim contains a very basic test for this.

Does nim default garbage collector (GC) and ARC/ORC work?

Yes. nimpy internally does everything needed to run the GC properly (keeps the stack bottom actual, and appropriate nim references alive), and doesn't introduce any special rules on top. So the GC question boils down to proper GC usage in nim shared libraries, you'd better lookup elsewhere. The following guidelines are by no means comprehensive, but should be enough for the quick start:

If it's known there will be only one nimpy module in the process, you should be fine.
If there is more than one nimpy module, it is recommended to move nim runtime out to a separate shared library. However it might not be needed if nim references are known to never travel between nim shared libraries.
If you hit any GC problems with nimpy, whether you followed these guidelines or not, please report them to nimpy tracker :)

Windows, threads and MinGW

When compiling with --threads:on Nim will imply --tlsEmulation:on (Windows only) which prevents Nim runtime from initing properly when being called from a foreign thread (which is always the case in case of Python module).

Adding --tlsEmulation:off when using MinGW toolchain (Nim's default on Windows) will introduce a dependency on libgcc_s_seh-*.dll, that newer python versions are often unable to find.

One way to overcome this is to link with libgcc statically, by passing -static to linker, or --passL:-static to Nim.

Exporting Nim types as Python classes

Warning! This is experimental.

An exported type should be a ref object and inherit PyNimObjectExperimental directly or indirectly.
The type will only be exported if at least one exported "method" is defined.
A proc will be exported as python type method only if it's first argument is of the corresponding type and is called self. If the first argument is not called self, the proc will exported as a global module function.

# mymodule.nim
type TestType = ref object of PyNimObjectExperimental
  myField: string

proc setMyField(self: TestType, value: string) {.exportpy.} =
  self.myField = value

proc getMyField(self: TestType): string {.exportpy.} =
  self.myField

# test.py
import mymodule
tt = mymodule.TestType()
tt.setMyField("Hello")
assert(tt.getMyField() == "Hello")

Future directions

High level buffer API

nomissbowling/nimpy