Getting Started with PyPy's Python Interpreter
PyPy's Python interpreter is a very compliant Python interpreter implemented in Python. When translated to C, it passes most of CPythons core language regression tests and comes with many of the extension modules included in the standard library including ctypes. It can run large libraries such as Django and Twisted. There are some small behavioral differences with CPython and some missing extensions, for details see CPython differences.
To actually use PyPy's Python interpreter, the first thing to do is to download a pre-built PyPy for your architecture.
(Note: for some hints on how to translate the Python interpreter under Windows, see the windows document)
You can translate the whole of PyPy's Python interpreter to low level C code, or CLI code. If you intend to build using gcc, check to make sure that the version you have is not 4.2 or you will run into this bug.
First download a pre-built PyPy for your architecture which you will use to translate your Python interpreter. It is, of course, possible to translate with a CPython 2.6 or later, but this is not the preferred way, because it will take a lot longer to run -- depending on your architecture, between two and three times as long.
Install build-time dependencies. On a Debian box these are:
[user@debian-box ~]$ sudo apt-get install \ gcc make python-dev libffi-dev pkg-config \ libz-dev libbz2-dev libncurses-dev libexpat1-dev \ libssl-dev libgc-dev python-sphinx python-greenlet
On a Fedora-16 box these are:
[user@fedora-or-rh-box ~]$ sudo yum install \ gcc make python-devel libffi-devel pkgconfig \ zlib-devel bzip2-devel ncurses-devel expat-devel \ openssl-devel gc-devel python-sphinx python-greenlet
The above command lines are split with continuation characters, giving the necessary dependencies first, then the optional ones.
- pkg-config (to help us locate libffi files)
- libz-dev (for the optional zlib module)
- libbz2-dev (for the optional bz2 module)
- libncurses-dev (for the optional _minimal_curses module)
- libexpat1-dev (for the optional pyexpat module)
- libssl-dev (for the optional _ssl module)
- libgc-dev (for the Boehm garbage collector: only needed when translating with --opt=0, 1 or size)
- python-sphinx (for the optional documentation build. You need version 1.0.7 or later)
Translation is time-consuming -- 45 minutes on a very fast machine -- and RAM-hungry. As of March 2011, you will need at least 2 GB of memory on a 32-bit machine and 4GB on a 64-bit machine. If your memory resources are constrained, or your machine is slow you might want to pick the optimization level 1 in the next step. A level of 2 or 3 or jit gives much better results, though. But if all you want to do is to test that some new feature that you just wrote translates, level 1 is enough.
Let me stress this again: at --opt=1 you get the Boehm GC, which is here mostly for historical and for testing reasons. You really do not want to pick it for a program you intend to use. The resulting pypy-c is slow.
cd pypy/translator/goal python translate.py --opt=jit targetpypystandalone.py
possibly replacing --opt=jit with another optimization level of your choice like --opt=2 if you do not want to include the JIT compiler, which makes the Python interpreter much slower.
If everything works correctly this will create an executable pypy-c in the current directory. Type pypy-c --help to see the options it supports - mainly the same basic options as CPython. In addition, pypy-c --info prints the translation options that where used to produce this particular executable. The executable behaves mostly like a normal Python interpreter:
$ ./pypy-c Python 2.7.2 (341e1e3821ff, Jun 07 2012, 15:40:31) [PyPy 1.9.0 with GCC 4.4.3] on linux2 Type "help", "copyright", "credits" or "license" for more information. And now for something completely different: ``RPython magically makes you rich and famous (says so on the tin)'' >>>> 46 - 4 42 >>>> from test import pystone >>>> pystone.main() Pystone(1.1) time for 50000 passes = 0.220015 This machine benchmarks at 227257 pystones/second >>>> pystone.main() Pystone(1.1) time for 50000 passes = 0.060004 This machine benchmarks at 833278 pystones/second >>>>
Note that pystone gets faster as the JIT kicks in. This executable can be moved around or copied on other machines; see Installation below.
The translate.py script takes a very large number of options controlling what to translate and how. See translate.py -h. The default options should be suitable for mostly everybody by now. Find a more detailed description of the various options in our configuration sections.
It is possible to have non-standard features enabled for translation, but they are not really tested any more. Look, for example, at the objspace proxies document.
To create a standalone .NET executable using the CLI backend:
./translate.py --backend=cli targetpypystandalone.py
The executable and all its dependencies will be stored in the ./pypy-cli-data directory. To run pypy.NET, you can run ./pypy-cli-data/main.exe. If you are using Linux or Mac, you can use the convenience ./pypy-cli script:
$ ./pypy-cli Python 2.7.0 (61ef2a11b56a, Mar 02 2011, 03:00:11) [PyPy 1.6.0] on linux2 Type "help", "copyright", "credits" or "license" for more information. And now for something completely different: ``distopian and utopian chairs'' >>>>
Moreover, at the moment it's not possible to do the full translation using only the tools provided by the Microsoft .NET SDK, since ilasm crashes when trying to assemble the pypy-cli code due to its size. Microsoft .NET SDK 2.0.50727.42 is affected by this bug; other versions could be affected as well: if you find a version of the SDK that works, please tell us.
Windows users that want to compile their own pypy-cli can install Mono: if a Mono installation is detected the translation toolchain will automatically use its ilasm2 tool to assemble the executables.
To try out the experimental .NET integration, check the documentation of the clr module.
A prebuilt pypy-c can be installed in a standard location like /usr/local/bin, although some details of this process are still in flux. It can also be copied to other machines as long as their system is "similar enough": some details of the system on which the translation occurred might be hard-coded in the executable.
PyPy dynamically finds the location of its libraries depending on the location of the executable. The directory hierarchy of a typical PyPy installation looks like this:
./bin/pypy ./include/ ./lib_pypy/ ./lib-python/2.7 ./site-packages/
The hierarchy shown above is relative to a PREFIX directory. PREFIX is computed by starting from the directory where the executable resides, and "walking up" the filesystem until we find a directory containing lib_pypy and lib-python/2.7.
The archives (.tar.bz2 or .zip) containing PyPy releases already contain the correct hierarchy, so to run PyPy it's enough to unpack the archive, and run the bin/pypy executable.
To install PyPy system wide on unix-like systems, it is recommended to put the whole hierarchy alone (e.g. in /opt/pypy1.9) and put a symlink to the pypy executable into /usr/bin or /usr/local/bin
If the executable fails to find suitable libraries, it will report debug: WARNING: library path not found, using compiled-in sys.path and then attempt to continue normally. If the default path is usable, most code will be fine. However, the sys.prefix will be unset and some existing libraries assume that this is never the case.
To start interpreting Python with PyPy, install a C compiler that is supported by distutils and use Python 2.5 or greater to run PyPy:
cd pypy python bin/py.py
After a few seconds (remember: this is running on top of CPython), you should be at the PyPy prompt, which is the same as the Python prompt, but with an extra ">".
Now you are ready to start running Python code. Most Python modules should work if they don't involve CPython extension modules. This is slow, and most C modules are not present by default even if they are standard! Here is an example of determining PyPy's performance in pystones:
>>>> from test import pystone >>>> pystone.main(10)
The parameter is the number of loops to run through the test. The default is 50000, which is far too many to run in a non-translated PyPy version (i.e. when PyPy's interpreter itself is being interpreted by CPython).
To list the PyPy interpreter command line options, type:
cd pypy python bin/py.py --help
py.py supports most of the options that CPython supports too (in addition to a large amount of options that can be used to customize py.py). As an example of using PyPy from the command line, you could type:
python py.py -c "from test import pystone; pystone.main(10)"
Alternatively, as with regular Python, you can simply give a script name on the command line:
python py.py ../../lib-python/2.7/test/pystone.py 10
See our configuration sections for details about what all the commandline options do.