https://github.com/wardbrian/ffi-examples

https://github.com/wardbrian/ffi-examples

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/wardbrian/ffi-examples
• Owner: WardBrian
• License: mit
• Created: 2023-09-22T15:06:48.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-10-19T19:02:10.000Z (over 1 year ago)
• Last Synced: 2025-01-28T15:23:33.252Z (4 months ago)
• Language: Python
• Size: 44.9 KB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# FFI Examples

This repo contains the slides and examples for my talk at the 2023 Flatiron Wide Annual Meeting (FWAM).

## Requirements

A C and C++ compiler, `make`, python with the pybind11 and numpy packages.

The [slides](./slides.md) are best viewed with [slides](https://github.com/maaslalani/slides).

## Outline

I will be focusing on the case of calling compiled code from a

higher-level language. This is because it is the more common (and useful) case,

and because it is the case which is the most similar between pairs of languages.

- Basic concepts

  - Linking and namespaces, `nm`

  - Name mangling and `extern "C"`

    - I am not familiar with Fortran but my understanding is

      you can use `iso_c_binding` to 'pretend' you're a C library.

- Basic FFI

  - Shared libraries, manual lookup

    - Calling conventions

    - I will be using Python, but this style is most

      useful if you are providing bindings for multiple

      languages.

      This style works in Python, Julia, R[^1], MATLAB, and likely more.

    - `ctypes`

[^1]: R has some extra weird limitations, but they're all things that you can work around if needed.

- Language-specific options (`Python.h`, `pybind11`, etc.)

  - Can be more powerful or convenient, but really only work for one language at a time.

- Common concerns[^2]

  - Error handling

  - Cross-language memory management

  - Converting between idioms (e.g. out parameters)

[^2]: Note that these are also concerns for API design generally, not just FFI.

- General tools that do it all for you (SWIG, etc.)

  - Can be powerful!

  - Can also be a pain and generate a _lot_ of code.

  - Personally, happy medium is things like clang2py which automate boilerplate of the "manual" approach