Not for 'proper' production (still missing a few libraries and C++/swig related stuff), but I have a collection of scripts that I use all the time that reads in data from various sources, does stuff to that data and then writes out that data to new files. Typing "pypy foo.py" instead of "python foo.py" gives me a 3-7 times speed up without having to make any changes to the source code. Can't really argue with that.

Quora ran PyPy briefly before switching to Scala in production, see: http://www.quora.com/Alex-Gaynor/Posts/Quora-is-now-running-...

Was it established that Quora stopped using PyPy? My impression was that more parts of the stack are now being written in Scala:

http://www.quora.com/Is-the-Quora-team-considering-adopting-...

"Yes, we're using Scala for some projects at Quora, both big and small. As a programming language, it offers the following advantages: fast, concise, type-safe, object-oriented yet functional, easy parallelism, extensible and built on the JVM. It is also gaining more adoption in the industry and thus getting more support. At Quora, we'll choose the best tools and stack for the product we're building, and that includes our language choices at times."

From talking to Quora engineers, my understanding is that they rewrote the majority of their stack in Scala. They might still be running Tornado (under PyPy) for their realtime connections to clients.

My coworkers and I use it instead of C Python for practically every script we use in our bioinformatics research. It works seamlessly, but "research scripts" are probably a perfect niche market for pypy.

Mike Tigas made a PyPy buildpack for Heroku:

http://github.com/mtigas/heroku-buildpack-pypy/

Here's his blog post about it:

http://v3.mike.tig.as/blog/2012/02/13/pypy-on-heroku/

I haven't used it on a serious project yet, but it was pretty easy to work with.

I'd be curious to hear any comments on the memory consumption of PyPy, and if has become an issue in any real usage.

Check out seatgeek, I know they're using it in production. I think they do most of their string analysis with it when they are scraping other sites to collect ticket data

I'm waiting for PyPy's NumPy support to get a bit more complete.

www.pylive.codespeaks.in uses pypy for sandboxing.

Not yet, but I'm considering it (== too lazy to implement).