nitely/v8-cffi

Embed the V8 Javascript Interpreter into Python

V8-CFFI

Embed the V8 Javascript engine into Python.

Note: The initial purpose of this library is to render React components server side.

Compatibility

• GCC +4.8
• G++ +4.8
• Python 2.7, 3.4, 3.5 and PyPy 5.3

Note: Linux-x64 is the only (officially) supported platform. To build the binaries for other platforms, the ./dev steps must be adapted (probably to vagrant instead of docker) accordingly. PRs are welcome.

Usage

from v8cffi import shortcuts
shortcuts.set_up()

ctx = shortcuts.get_context()
ctx.load_libs(['./foo_bundled.js'])
ctx.run_script('foo.render("hola mundo");')
# "hola mundo"

Read the docs.

Notes

• Currently ships with V8 4.9.385.33 (stable).
• This repo contains V8 static files for Linux-x64, built with CentOS 6.7 (glibc 2.12), it's known to work in Ubuntu 12.04/14.04/16.04.

Resources

• How does NodeJS works?
• Beautiful Native Libraries or alt
• V8 Custom Snapshots
• V8 Cache Code
• Latest stable V8
• V8 Wiki
• V8 Embedder's Guide
• V8 users group
• CFFI user group
• CFFI - Distribute including a (pre-)compiled library

Devs

Devs

Benchmarks

This will run some silly benchmarks. It runs about ~110k ops/s on a 1.8GHz CPU.

$ make benchmarks

License

MIT

v8-cffi's Issues

Compile with custom v8 (archives) source

Should allow to compile v8cffi with an external v8 static-library source.

• Add ENV['external_v8'] to use as path for the archives when compiling.

Support compile with clang

Thread pool of pre-loaded VMs

Would allow to run JS code in parallel. Should support asyncio. Should be transparent (ie: same API as for managing one VM)

Python 2/PyPy support

I checked your .travis.yml and it seems you only test with Python 3.
Is that on purpose?

C layer platform should be a singleton

This is just for correctness. The underlying V8 platform is a singleton, so the C layer should be implemented as a singleton as well, nothing else should be refactored.

Priority is very low.

Compile with custom v8 shared library

Should allow to compile and link a v8 shared library. Also, this way it should be possible to distribute the binaries for OSX and Windows (static libraries are too big ~600MB).

• Support for custom rpath