- Contributing Guidelines
- Getting involved
- Your first contribution
- Testing After Translation
- Tooling & Utilities
- graphviz & pygame for flow graph viewing (highly recommended)
PyPy is a very large project that has a reputation of being hard to dive into. Some of this fame is warranted, some of it is purely accidental. There are three important lessons that everyone willing to contribute should learn:
- PyPy has layers. There are many pieces of architecture that are very well separated from each other. More about this below, but often the manifestation of this is that things are at a different layer than you would expect them to be. For example if you are looking for the JIT implementation, you will not find it in the implementation of the Python programming language.
- Because of the above, we are very serious about Test Driven Development. It’s not only what we believe in, but also that PyPy’s architecture is working very well with TDD in mind and not so well without it. Often development means progressing in an unrelated corner, one unittest at a time; and then flipping a giant switch, bringing it all together. (It generally works out of the box. If it doesn’t, then we didn’t write enough unit tests.) It’s worth repeating - PyPy’s approach is great if you do TDD, and not so great otherwise.
- PyPy uses an entirely different set of tools - most of them included in the PyPy repository. There is no Makefile, nor autoconf. More below.
The first thing to remember is that PyPy project is very different than most projects out there. It’s also different from a classic compiler project, so academic courses about compilers often don’t apply or lead in the wrong direction. However, if you want to understand how designing & building a runtime works in the real world then this is a great project!
PyPy employs a relatively standard open-source development process. You are encouraged as a first step to join our pypy-dev mailing list and IRC channel, details of which can be found in our contact section. The folks there are very friendly, and can point you in the right direction.
We give out commit rights usually fairly liberally, so if you want to do something with PyPy, you can become a committer. We also run frequent coding sprints which are separately announced and often happen around Python conferences such as EuroPython or PyCon. Upcoming events are usually announced on the blog.
Further Reading: Contact
The first and most important rule how not to contribute to PyPy is “just hacking a feature”. This won’t work, and you’ll find your PR will typically require a lot of re-work. There are a few reasons why not:
- build times are large
- PyPy has very thick layer separation
- context of the cPython runtime is often required
Instead, reach out on the dev mailing list or the IRC channel, and we’re more than happy to help! :)
Some ideas for first contributions are:
- Documentation - this will give you an understanding of the pypy architecture
- Test failures - find a failing test in the nightly builds, and fix it
- Missing language features - these are listed in our issue tracker
PyPy development is based a typical fork/pull request based workflow, centered around Mercurial (hg), hosted on Bitbucket. If you have not used this workflow before, a good introduction can be found here:
The cycle for a new PyPy contributor goes typically like this:
- Make an account on bitbucket.
- Go to https://bitbucket.org/pypy/pypy/ and click “fork” (left icons). You get a fork of the repository, e.g. in https://bitbucket.org/yourname/pypy/.
- Clone your new repo (i.e. the fork) to your local machine with the command
hg clone ssh://firstname.lastname@example.org/yourname/pypy. It is a very slow operation but only ever needs to be done once. See also http://pypy.org/download.html#building-from-source . If you already cloned
https://bitbucket.org/pypy/pypybefore, even if some time ago, then you can reuse the same clone by editing the file
.hg/hgrcin your clone to contain the line
default = ssh://email@example.com/yourname/pypy, and then do
hg pull && hg up. If you already have such a clone but don’t want to change it, you can clone that copy with
hg clone /path/to/other/copy, and then edit
.hg/hgrcas above and do
hg pull && hg up.
- Now you have a complete copy of the PyPy repo. Make a branch
with a command like
hg branch name_of_your_branch.
- Edit things. Use
hg diffto see what you changed. Use
hg addto make Mercurial aware of new files you added, e.g. new test files. Use
hg statusto see if there are such files. Write and run tests! (See the rest of this page.)
- Commit regularly with
hg commit. A one-line commit message is fine. We love to have tons of commits; make one as soon as you have some progress, even if it is only some new test that doesn’t pass yet, or fixing things even if not all tests pass. Step by step, you are building the history of your changes, which is the point of a version control system. (There are commands like
hg upthat you should read about later, to learn how to navigate this history.)
- The commits stay on your machine until you do
hg pushto “push” them back to the repo named in the file
.hg/hgrc. Repos are basically just collections of commits (a commit is also called a changeset): there is one repo per url, plus one for each local copy on each local machine. The commands
hg pullcopy commits around, with the goal that all repos in question end up with the exact same set of commits. By opposition,
hg uponly updates the “working copy” by reading the local repository, i.e. it makes the files that you see correspond to the latest (or any other) commit locally present.
- You should push often; there is no real reason not to. Remember that
even if they are pushed, with the setup above, the commits are (1)
bitbucket.org/yourname/pypy, and (2) in the branch you named. Yes, they are publicly visible, but don’t worry about someone walking around the thousands of repos on bitbucket saying “hah, look at the bad coding style of that guy”. Try to get into the mindset that your work is not secret and it’s fine that way. We might not accept it as is for PyPy, asking you instead to improve some things, but we are not going to judge you.
- The final step is to open a pull request, so that we know that you’d
like to merge that branch back to the original
pypy/pypyrepo. This can also be done several times if you have interesting intermediate states, but if you get there, then we’re likely to proceed to the next stage, which is…
- Get a regular account for pushing directly to
bitbucket.org/pypy/pypy(just ask and you’ll get it, basically). Once you have it you can rewrite your file
default = ssh://firstname.lastname@example.org/pypy/pypy. Your changes will then be pushed directly to the official repo, but (if you follow these rules) they are still on a branch, and we can still review the branches you want to merge.
- If you get closer to the regular day-to-day development, you’ll notice
that we generally push small changes as one or a few commits directly
to the branch
default. Also, we often collaborate even if we are on other branches, which do not really “belong” to anyone. At this point you’ll need
hg mergeand learn how to resolve conflicts that sometimes occur when two people try to push different commits in parallel on the same branch. But it is likely an issue for later
PyPy has layers. Just like ogres or onions. Those layers help us keep the
respective parts separated enough to be worked on independently and make the
complexity manageable. This is, again, just a sanity requirement for such
a complex project. For example writing a new optimization for the JIT usually
does not involve touching a Python interpreter at all or the JIT assembler
backend or the garbage collector. Instead it requires writing small tests in
rpython/jit/metainterp/optimizeopt/test/test_* and fixing files there.
After that, you can just compile PyPy and things should just work.
Further Reading: architecture
PyPy is made from parts that are relatively independent of each other. You should start looking at the part that attracts you most (all paths are relative to the PyPy top level directory). You may look at our directory reference or start off at one of the following points:
- pypy/interpreter contains the bytecode interpreter: bytecode dispatcher in pypy/interpreter/pyopcode.py, frame and code objects in pypy/interpreter/eval.py and pypy/interpreter/pyframe.py, function objects and argument passing in pypy/interpreter/function.py and pypy/interpreter/argument.py, the object space interface definition in pypy/interpreter/baseobjspace.py, modules in pypy/interpreter/module.py and pypy/interpreter/mixedmodule.py. Core types supporting the bytecode interpreter are defined in pypy/interpreter/typedef.py.
- pypy/interpreter/pyparser contains a recursive descent parser, and grammar files that allow it to parse the syntax of various Python versions. Once the grammar has been processed, the parser can be translated by the above machinery into efficient code.
- pypy/interpreter/astcompiler contains the compiler. This contains a modified version of the compiler package from CPython that fixes some bugs and is translatable.
- pypy/objspace/std contains the
Standard object space. The main file
is pypy/objspace/std/objspace.py. For each type, the file
xxxobject.pycontains the implementation for objects of type
xxx, as a first approximation. (Some types have multiple implementations.)
For building PyPy, we recommend installing a pre-built PyPy first (see Downloading and Installing PyPy). It is possible to build PyPy with CPython, but it will take a lot longer to run – depending on your architecture, between two and three times as long.
Further Reading: Build
Instead, we practice a lot of test driven development. This is partly because of very high quality requirements for compilers and partly because there is simply no other way to get around such complex project, that will keep you sane. There are probably people out there who are smart enough not to need it, we’re not one of those. You may consider familiarizing yourself with pytest, since this is a tool we use for tests. This leads to the next issue:
The py.test testing tool drives all our testing needs.
We use the py library for filesystem path manipulations, terminal writing, logging and some other support functionality.
You don’t necessarily need to install these two libraries because we also ship them inlined in the PyPy source tree.
The PyPy source tree comes with an inlined version of
which you can invoke by typing:
python pytest.py -h
This is usually equivalent to using an installed version:
If you encounter problems with the installed version
make sure you have the correct version installed which
you can find out with the
You will need the build requirements to run tests successfully, since many of them compile little pieces of PyPy and then run the tests inside that minimal interpreter. The cpyext tests also require pycparser, and many tests build cases with hypothesis.
Now on to running some tests. PyPy has many different test directories and you can use shell completion to point at directories or files:
py.test pypy/interpreter/test/test_pyframe.py # or for running tests of a whole subdirectory py.test pypy/interpreter/
See py.test usage and invocations for some more generic info on how you can run tests.
Beware trying to run “all” pypy tests by pointing to the root
directory or even the top level subdirectory
pypy. It takes
hours and uses huge amounts of RAM and is not recommended.
To run CPython regression tests you can point to the
This will usually take a long time because this will run the PyPy Python interpreter on top of CPython. On the plus side, it’s usually still faster than doing a full translation and running the regression test with the translated PyPy Python interpreter.
While the usual invocation of pytest translates a piece of RPython code and runs it, we have a test extension to run tests without translation, directly on the host python. This is very convenient for modules such as cpyext, to compare and contrast test results between CPython and PyPy. Untranslated tests are invoked by using the -A or –runappdirect option to pytest:
python2 pytest.py -A pypy/module/cpyext/test
where python2 can be either python2 or pypy2. On the py3 branch, the collection phase must be run with python2 so untranslated tests are run with:
cpython2 pytest.py -A pypy/module/cpyext/test --python=path/to/pypy3
If you are interested in the inner workings of the PyPy Python interpreter, there are some features of the untranslated Python interpreter that allow you to introspect its internals.
To start interpreting Python with PyPy, install a C compiler that is supported by distutils and use Python 2.7 or greater to run PyPy:
cd pypy python bin/pyinteractive.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 “>”.
If you press
<Ctrl-C> on the console you enter the interpreter-level console, a
usual CPython console. You can then access internal objects of PyPy
(e.g. the object space) and any variables you have created on the PyPy
prompt with the prefix
>>>> a = 123 >>>> <Ctrl-C> *** Entering interpreter-level console *** >>> w_a W_IntObject(123)
The mechanism works in both directions. If you define a variable with the
w_ prefix on the interpreter-level, you will see it on the app-level:
>>> w_l = space.newlist([space.wrap(1), space.wrap("abc")]) >>> <Ctrl-D> *** Leaving interpreter-level console *** KeyboardInterrupt >>>> l [1, 'abc']
Note that the prompt of the interpreter-level console is only ‘>>>’ since it runs on CPython level. If you want to return to PyPy, press <Ctrl-D> (under Linux) or <Ctrl-Z>, <Enter> (under Windows).
Also note that not all modules are available by default in this mode (for
_continuation needed by
greenlet) , you may need to use one of
--withmod-... command line options.
You may be interested in reading more about the distinction between interpreter-level and app-level.
To list the PyPy interpreter command line options, type:
cd pypy python bin/pyinteractive.py --help
pyinteractive.py supports most of the options that CPython supports too (in addition to a large amount of options that can be used to customize pyinteractive.py). As an example of using PyPy from the command line, you could type:
python pyinteractive.py --withmod-time -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 pyinteractive.py --withmod-time ../../lib-python/2.7/test/pystone.py 10
--withmod-xxx option enables the built-in module
default almost none of them are, because initializing them takes time.
If you want anyway to enable all built-in modules, you can use
See our configuration sections for details about what all the commandline options do.
You can use a simple tracing mode to monitor the interpretation of
bytecodes. To enable it, set
__pytrace__ = 1 on the interactive
>>>> __pytrace__ = 1 Tracing enabled >>>> x = 5 <module>: LOAD_CONST 0 (5) <module>: STORE_NAME 0 (x) <module>: LOAD_CONST 1 (None) <module>: RETURN_VALUE 0 >>>> x <module>: LOAD_NAME 0 (x) <module>: PRINT_EXPR 0 5 <module>: LOAD_CONST 0 (None) <module>: RETURN_VALUE 0 >>>>
The example-interpreter repository contains an example interpreter written using the RPython translation toolchain.
graphviz and pygame are both necessary if you want to look at generated flow graphs: