Making C and Python Talk to Each Other

https://leetarxiv.substack.com/p/making-c-and-python-talk-to-each

by muragekibicho • 1 year ago

158 101 1 year ago

leetarxiv.substack.com

muragekibicho• 1 year ago

Lots of articles focus on Cython and optimizing Python using C code.

This article is about embedding Python scripts inside a C codebase

SandmanDP• 1 year ago

I’ve been curious, what are the motivations for most projects to use Lua for enabling scripting in C over this? Is the concern around including an entire Python interpreter in a project and Lua is lighter?

90s_dev• 1 year ago

Network effect.

bandoti• 1 year ago

Lua is much lighter but the key is that it’s probably one of the easiest things to integrate (just copy the sources/includes and add them to build it’ll work)—like a “header only” kind of vibe.

But, you can strip down a minimal Python build and statically compile it without too much difficulty.

I tend to prefer Tcl because it has what I feel the perfect amount of functionality by default with a relatively small size. Tcl also has the better C APIs of the bunch if you’re working more in C.

Lua is very “pushy” and “poppy” due to its stack-based approach, but that can be fun too if you enjoy programming RPN calculators haha :)

crote2• 1 year ago

Lua is absolutely trivial to isolate. As the embedder, you have complete control over what the interpreter and VM are doing. Don't want your Lua scripts to have file access? Don't hook up those functions and you're done. Want to prevent against endless loops? Tell the VM to stop after 10.000 instructions. Want to limit the amount of memory a script can use? Absolutely trivial. This makes Lua very attractive for things like game development. You can run untrusted addon code without any worry that it'll be able to mess up the game - or the rest of the system.

Doing the same with Python is a lot harder. Python is designed first and foremost to run on its own. If you embed Python you are essentially running it besides your own code, with a bunch of hooks in both directions. Running hostile Python code? Probably not a good idea.

OskarS• 1 year ago

Another thing to mention is that until very recently (Python 3.12, I think?) every interpreter in the address space shared a lot of global state, including most importantly the GIL. For my area (audio plugins) that made Python a non-starter for embedding, while Lua works great.

I agree though: biggest reason is probably the C API. Lua's is so simple to embed and to integrate with your code-base compared to Python. The language is also optimized for "quick compiling", and it's also very lightweight.

These days, however, one might argue that you gain so much from embedding either Python or JavaScript, it might be worth the extra pain on the C/C++ side.

spacechild12• 1 year ago

People already mentioned that Lua is very lightweight and easy to integrate. It's also significantly faster than Python. (I'm not even talking about LuaJIT.)

Another big reason: the Lua interpreter does not have any global variables (and therefore also no GIL) so you can have multiple interpreters that are completely independent from each other.

Derbasti• 1 year ago

I've done both. Let me tell you, embedding Lua into a C program is magnitudes easier than embedding Python.

The main reason is the endless fragility of Python's garbage collector. To be clear, the API is trying to be as helpful as possible, but it's still a whole bunch of complexity that's easy to mess up. Incidentally, this discussion is left out in the linked article, which makes it less than useless. In my experience with many a third party C/Py interface, data leaks are incredibly common in such code.

Lua of course also has a garbage collector, but it essentially only knows two types of values: POD, and tables, with only the latter needing much consideration. The interaction model is based on a stack-based virtual machine, which is more complex than Python's full-function abstraction, but conveniently hides most of the garbage collector complexity. So long as you're just reshuffling things on the stack (i.e. most of the time), you don't need to worry about the garbage collector at all.

jebarker9• 1 year ago

Lots of people argue that AI R&D is currently done in Python because of the benefits of the rich library ecosystem. This makes me realize that's actually a poor reason for everything to be in Python since the actually useful libraries for things like visualization could easily be called from lower level languages if they're off the hot path.

giancarlostoro2• 1 year ago

I think its more than just because of the available libraries. I think that industry has just predominantly preferred Python. Python is a really rich modern language, it might be quirky, but so is every single language you can name. Nothing is quite as quirky as JavaScript though, maybe VB6 but that's mostly dead, though slightly lingering.

Mind you I've programmed in all the mentioned languages. ;)

whattheheckheck• 1 year ago

It's the ease of distribution of packages and big functionality being a pip install away

kristjansson• 1 year ago

One ... could? But it doesn't seem particularly ergonomic.

jebarker9• 1 year ago

Ergonomics isn't the point, performance is.

ashishb2• 1 year ago

I rewrote a simple RAG ingestion pipeline from Python to Go.

It reads from a database. Generates embeddings. Writes it to a vector database.

  - ~10X faster
  - ~10X lower memory usage

The only problem is that you have to spend a lot of time figuring out how to do it.

All instructions on the Internet and even on the vector database documentation are in Python.

chpatrick• 1 year ago

If speed and memory use aren't a bottleneck then "a lot of time figuring out how to do it" is probably the biggest cost for the company. Generally these things can be run offline and memory is fairly cheap. You can get a month of a machine with a ton of RAM for the equivalent of one hour of developer time of someone who knows how to do this. That's why Python is so popular.

kgeist• 1 year ago

I rewrote a simple RAG ingestion pipeline from Python to Go

I also wrote a RAG pipeline in Go, using OpenSearch for hybrid search (full-text + semantic) and the OpenAI API. I reused OpenSearch because our product was already using it for other purposes, and it supports vector search.

For me, the hardest part was figuring out all the additional settings and knobs in OpenSearch to achieve around 90% successful retrieval, as well as determining the right prompt and various settings for the LLM. I've found that these settings can be very sensitive to the type of data you're applying RAG to. I'm not sure if there's a Python library that solves this out of the box without requiring manual tuning too

crote2• 1 year ago

could easily be called from lower level languages

Could? Yes. Easily? No.

People write their business logic in Python because they don't want to code in those lower-level languages unless they absolutely have to. The article neatly shows the kind of additional coding overhead you'd have to deal with - and you're not getting anything back in return.

Python is successful because it's a high-level language which has the right tooling to create easy-to-use wrappers around low-level high-performance libraries. You get all the benefits of a rich high-level language for the cold path, and you only pay a small penalty over using a low-level language for the hot path.

jebarker9• 1 year ago

The problem I see (day to day working on ML framework optimization) is that it's not just a case of python calling lower level compiled code. Pytorch, for example, has a much closer integration of python and the low level functions than that and it does cause performance bottlenecks. So in theory I agree that using high level languages to script calls to low level is a good idea, but in practice that gets abused to put python in the hot path. Perhaps if the lower level language were the bulk of the framework and just called python for helper functions we'd see better performance-aware design from developers.

efavdb• 1 year ago

FWIW I would be up to write in c or something else, but use python for the packages / network effects.

mkoubaa• 1 year ago

Nobody has ever, in the history of Python, called the Python C API easy.

bigger_cheese• 1 year ago

I have been using Python recently and have found a lot of the data visualization tools seem to be wrappers around other languages (mostly JavaScript), things like, agGrid, Tabulator, Plotly etc.

Sometimes you end up embedding chunks of javascript directly inside your python

For example the docs for Streamlit implementation of AgGrid contain this: https://staggrid-examples.streamlit.app/Advanced_config_and_...

Grimblewald6• 1 year ago

It's not just about library availability. Python wins because it lets you offload the low-level performance work to people who really know what they’re doing. Libraries like NumPy or PyTorch / Keras wrap highly optimized C/C++ code—so you get near-C/++ performance without having to write or debug C yourself, and without needing a whole computer science degree to do so properly.

It's a mistake to assume C is always faster. If you don’t have a deep understanding of memory layout, compiler flags, vectorization, cache behavior, etc. your hand-written C code can easily be slower than high-level Python using well-optimized libraries. See [1] for a good example of that.

Sure, you could call those same libs from C, but then you're reinventing Python's ecosystem with more effort and more chances to shoot yourself in the foot. Python gives you access to powerful, low-level tools while letting you focus on higher-level problems—in a language that’s vastly easier to learn and use.

That tradeoff isn't just convenience—it's what makes modern AI R&D productive at scale.

[1] https://stackoverflow.com/questions/41365723/why-is-my-pytho...

jebarker9• 1 year ago

I feel like you're re-stating the same claim that crote made that there's a clean cut between python and lower level libraries meaning that the user doesn't need to know what is happening at the lower level to achieve good performance. This is not true in many cases if you are aiming to achieve peak performance - which we should be for training and serving AI systems since they are already so resource hungry.

a_t48• 1 year ago

Useful, I’m going to be doing something similar w/C++ soon.

kvemkon• 1 year ago

Once I needed to implement a simple python plugin engine in a C/C++ software, I've been successfully using the official guide [1].

[1] https://docs.python.org/3/extending/embedding.html

eth_hack77• 1 year ago

Thanks a lot for the article. Here's a QQ: did you measure the time of some basic operations python vs C? (e.g. if I do a loop of 10 billion iterations, just dividing numbers in C and do the same in python, and then import these operations into one another as libraries, does anything change?)

I'm a beginner engineer so please don't judge me if my question is not making perfect sense.

bdbenton52553• 1 year ago

C is many magnitudes faster than Python and you can measure this using nested conditionals. Python is built for a higher level of abstraction and this comes at the cost of speed. It is what makes it very natural and human-like to write in.

xandrius• 1 year ago

Syntax has nothing to do with the speed of the language: python could be "natural" and "human-like" while being much faster and also "unnatural" and "inhuman" while being slower.

rossant5• 1 year ago

My visualization library [1] is written in C and exposes a visualization API in C. It is packaged as a Python wheel using auto-generated ctypes bindings, which includes the shared library (so, dylib, or dll) and a few dependencies. This setup works very well, with no need to compile against each Python version. I only need to build it for the supported platforms, which is handled automatically by GitHub Actions. The library is designed to minimize the number of C calls, making the ctypes overhead negligible in practice.

[1] https://datoviz.org/

Grimblewald6• 1 year ago

This looks awesome, thanks for sharing - do you have any info on how it compares to matplotlib in terms of plotting speed? (even just rough estimates)

fadesibert• 1 year ago

A quick inspection of the article suggests there's a difference of intent.

<snip> Datoviz is a relatively low-level visualization library. It focuses on rendering visual primitives like points, lines, images, and meshes — efficiently and interactively.

Unlike libraries such as Matplotlib, Datoviz does not provide high-level plotting functions like plt.plot(), plt.scatter(), or plt.imshow(). Its goal is not to replace plotting libraries, but to serve as a powerful rendering backend for scientific graphics. </snip>

rossant5• 1 year ago

Good idea, I should do benchmarks on simple plots. It's orders of magnitude faster. Above 60 FPS on hundreds of thousands to even millions of points depending on the type of plot, on common hardware.

lgtx• 1 year ago

How does it compare to VTK, that also provides a Python API?

rossant5• 1 year ago

Datoviz is much lighter and younger than VTK. It compiles in seconds, making it much faster to build, and it's easier to install and distribute.

It has far fewer features since it focuses solely on rendering common primitives. For example, it doesn't handle data file loading, except for a few quick loaders like OBJ used mainly for testing.

There's almost no computational geometry, data processing, or signal processing functionality. Datoviz is solely focused on interactive rendering.

Datoviz also supports fast, high-quality 2D vector graphics, scaling efficiently to millions of points. In contrast, VTK is primarily designed for more complex and heavy 3D rendering tasks.

brcmthrowaway• 1 year ago

How does this compare to pybind11?

dgrunwald• 1 year ago

The article is using the Python C API directly.

pybind11 is a C++ wrapper that makes the Python API more friendly to use from C++ (e.g. smart pointers instead of manual reference counting)

nubinetwork• 1 year ago

Isn't this the whole point to cffi and cython?

softwaredoug• 1 year ago

Definitely though Cython is a layer of abstraction that might feel like Python has all kinds of weirdness you might as well write in a better understood language like C.

pjmlp4• 1 year ago

As someone that routinely kind of talks C down, yet knows it well enough since around 1990, I will acknowledge that for many in the current generation entering the workforce that is definetly not a given, especially the recent trend to treat C as if it was a scripting language, tryingt to avoid learning about the whole compiler toolchain with stuff like header only libraries.

dexzod• 1 year ago

The title of the article is misleading. Making C and python talk to each other implies, calling python from C and calling C from python. The article only covers the former.

hughw• 1 year ago

I realize I'm talking about C++ not C, but coincidentally just today I ported our 7 year old library's Swig/Python interface to nanobind. What a fragile c9k Swig has been all these years (don't touch it!) and the nanobind transformation is so refreshing and clean, and lots of type information suddenly available to Python programs. One day of effort and our tests all pass, and now nanobind seems able to allow us to improve the ergonomics (from the Python pov) of our lib.

rrdharan• 1 year ago

What is c9k short for?

spacechild12• 1 year ago

I guess it's "clusterfuck" :)

hugs• 1 year ago

This is one of the "killer apps" for Nim. Nim makes makes it easy to wrap C and easy to talk to Python (via Nimpy).

nottorp• 1 year ago

Or this?

https://docs.python.org/3/extending/embedding.html

vkoskiv• 1 year ago

I did a lot of this for my raytracer, c-ray [1]. Originally it was just a self-contained C program, but I got tired of writing buggy and limited asset import/export code, so eventually I put together a minimal public C API [2] that I then wrapped with CPython bindings [3] and some additional python code [4] to expose a more 'pythonic' API. It's all still a WIP, but it has already allowed me to write a Blender plugin [5], so now I can play around with my renderer directly in Blender, and test with more complex scenes others have made.

Fun project, and it's really cool to see my little renderer in the interactive viewport in Blender, but I have also learned that I don't particularly enjoy working with non-trivial amounts of Python code.

[1] https://github.com/vkoskiv/c-ray [2] https://github.com/vkoskiv/c-ray/blob/51a742b2ee4d0b570975cd... [3] https://github.com/vkoskiv/c-ray/tree/51a742b2ee4d0b570975cd... [4] https://github.com/vkoskiv/c-ray/tree/51a742b2ee4d0b570975cd... [5] https://github.com/vkoskiv/c-ray/tree/51a742b2ee4d0b570975cd...

UncleEntity• 1 year ago

Just a little nitpick but using Py_BuildValue is much better to generate PyObjects from C if you are doing more than one of them instead of populating a list and converting it to a tuple to pass to the python function.

My general rule (when doing the opposite of passing C values to python as part of an extension) is to use the dedicated function, like PyLong_FromLong, when there is a single return and Py_BuildValue (with its format string for automagic conversion) when the function returns a tuple.

Oh, and if you are checking your return values from python (you are, right?) using Py_XDECREF isn't all that good of an idea since it will mask some flawed logic. I pretty much only use it when a PyObject pointer could validly be NULL (like with an optional function argument) and I'm cleaning up right before throwing an exception. Tracking python reference counts is a whole blog post in itself since some functions steal references and others don't and if you get it wrong you can easily crash the interpreter and/or create memory leaks.

Making C and Python Talk to Each Other

Related Stories