1
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
4
<title>PyPy[project-ideas] </title>
5
<meta content="text/html;charset=ISO-8859-1" name="Content-Type"/>
6
<link href="style.css" media="screen" rel="stylesheet" type="text/css"/></head>
8
<div><a><img alt="PyPy" height="110" id="pyimg" src="http://codespeak.net/pypy/img/py-web1.png" width="149"/></a></div>
10
<div class="project_title">PyPy[project-ideas] </div>
11
<div id="menubar"><a class="menu" href="http://codespeak.net/pypy/dist/pypy/doc/news.html">news</a> <a class="menu" href="getting-started.html">getting-started</a> <a class="menu" href="index.html">documentation</a> <a class="menu" href="https://codespeak.net/viewvc/pypy/dist/">svn</a> <a class="menu" href="https://codespeak.net/issue/pypy-dev/">issues</a><a class="menu" href="contact.html">contact</a> <a class="menu" href="http://pypy.org/">EU/project</a> </div></div>
12
<div id="contentspace">
13
<div id="docinfoline">
14
<div style="float: right; font-style: italic;"> </div></div>
15
<div class="document" id="independent-project-ideas-relating-to-pypy">
16
<h1 class="title">Independent project ideas relating to PyPy</h1>
18
<p>PyPy allows experimentation in many directions -- indeed facilitating
19
experimentation in language implementation was one of the main
20
motivations for the project. This page is meant to collect some ideas
21
of experiments that the core developers have not had time to perform
22
yet and also do not require too much in depth knowledge to get started
24
<p>Feel free to suggest new ideas and discuss them in #pypy on the freenode IRC
25
network or the pypy-dev mailing list (see the <a class="reference" href="contact.html">contact</a> page).</p>
26
<hr class="docutils" />
27
<div class="contents topic">
28
<p class="topic-title first"><a id="contents" name="contents">Contents</a></p>
30
<li><a class="reference" href="#experiment-with-optimizations" id="id1" name="id1">Experiment with optimizations</a></li>
31
<li><a class="reference" href="#start-or-improve-a-back-end" id="id2" name="id2">Start or improve a back-end</a></li>
32
<li><a class="reference" href="#improve-javascript-backend" id="id3" name="id3">Improve JavaScript backend</a></li>
33
<li><a class="reference" href="#improve-one-of-the-jit-back-ends" id="id4" name="id4">Improve one of the JIT back-ends</a></li>
34
<li><a class="reference" href="#write-a-new-front-end" id="id5" name="id5">Write a new front end</a></li>
35
<li><a class="reference" href="#investigate-restricted-execution-models" id="id6" name="id6">Investigate restricted execution models</a></li>
36
<li><a class="reference" href="#experiment-with-distribution-and-persistence" id="id7" name="id7">Experiment with distribution and persistence</a></li>
37
<li><a class="reference" href="#numeric-numpy-numarray-support" id="id8" name="id8">Numeric/NumPy/numarray support</a></li>
38
<li><a class="reference" href="#extension-modules" id="id9" name="id9">Extension modules</a></li>
39
<li><a class="reference" href="#extend-py-execnet-to-a-peer-to-peer-model" id="id10" name="id10">Extend py.execnet to a peer-to-peer model</a></li>
40
<li><a class="reference" href="#or-else" id="id11" name="id11">Or else...</a></li>
44
<h1><a class="toc-backref" href="#id1" id="experiment-with-optimizations" name="experiment-with-optimizations">Experiment with optimizations</a></h1>
45
<p>Although PyPy's Python interpreter is very compatible with CPython, it is not
46
yet as fast. There are several approaches to making it faster, including the
47
on-going Just-In-Time compilation efforts and improving the compilation tool
48
chain, but probably the most suited to being divided into reasonably sized
49
chunks is to play with alternate implementations of key data structures or
50
algorithms used by the interpreter. PyPy's structure is designed to make this
51
straightforward, so it is easy to provide a different implementation of, say,
52
dictionaries or lists without disturbing any other code.</p>
53
<p>As examples, we've got working implementations of things like:</p>
55
<li>lazy string slices (slicing a string gives an object that references a part
56
of the original string).</li>
57
<li>lazily concatenated strings (repeated additions and joins are done
59
<li>dictionaries specialized for string-only keys.</li>
60
<li>dictionaries which use a different strategy when very small.</li>
61
<li>caching the lookups of builtin names (by special forms of
62
dictionaries that can invalidate the caches when they are written to)</li>
64
<p>Things we've thought about but not yet implemented include:</p>
66
<li>create multiple representations of Unicode string that store the character
67
data in narrower arrays when they can.</li>
69
<p>Experiments of this kind are really experiments in the sense that we do not know
70
whether they will work well or not and the only way to find out is to try. A
71
project of this nature should provide benchmark results (both timing and memory
72
usage) as much as code.</p>
73
<p>Some ideas on concrete steps for benchmarking:</p>
75
<li>find a set of real-world applications that can be used as benchmarks
76
for pypy (ideas: docutils, <a class="reference" href="http://hachoir.org/">http://hachoir.org/</a>, moinmoin, ...?)</li>
77
<li>do benchmark runs to see how much speedup the currently written
78
optimizations give</li>
79
<li>profile pypy-c and its variants with these benchmarks, identify slow areas</li>
80
<li>try to come up with optimized implementations for these slow areas</li>
84
<h1><a class="toc-backref" href="#id2" id="start-or-improve-a-back-end" name="start-or-improve-a-back-end">Start or improve a back-end</a></h1>
85
<p>PyPy has complete, or nearly so, back-ends for C, LLVM, CLI/.NET and JVM and partial
86
backends for JavaScript, Common Lisp, Squeak. It would be an
87
interesting project to improve either of these partial backends, or start one
88
for another platform (Objective C comes to mind as one that should not be too
92
<h1><a class="toc-backref" href="#id3" id="improve-javascript-backend" name="improve-javascript-backend">Improve JavaScript backend</a></h1>
93
<p>The JavaScript backend is somehow different from other pypy's backends because
94
it does not try to support all of PyPy (where it might be run then?), but rather
95
to compile RPython programs into code that runs in a browser. Some documents
96
are in <a class="reference" href="js/whatis.html">what is PyPy.js</a> file and <a class="reference" href="js/using.html">using the JavaScript backend</a>. Some project
99
<li>Write some examples to show different possibilities of using the backend.</li>
100
<li>Improve the facilities for testing RPython intended to be translated to
101
JavaScript on top of CPython, mostly by improving existing DOM interface.</li>
102
<li>Write down the mochikit bindings (or any other interesting JS effects library),
104
<li>Write down better object layer over DOM in RPython to make writing applications
108
<div class="section">
109
<h1><a class="toc-backref" href="#id4" id="improve-one-of-the-jit-back-ends" name="improve-one-of-the-jit-back-ends">Improve one of the JIT back-ends</a></h1>
110
<p>PyPy's Just-In-Time compiler relies on two assembler backends for actual code
111
generation, one for PowerPC and the other for i386. Idea would be start a new backend
112
for ie. mobile device.</p>
113
<p>Another idea in a similar vein would be to use LLVM to re-compile functions that
114
are executed particularly frequently (LLVM cannot be used for <em>all</em> code
115
generation, since it can only work on function at a time).</p>
117
<div class="section">
118
<h1><a class="toc-backref" href="#id5" id="write-a-new-front-end" name="write-a-new-front-end">Write a new front end</a></h1>
119
<p>Write an interpreter for <strong>another dynamic language</strong> in the PyPy framework.
120
For example, a Scheme interpreter would be suitable (and it would even be
121
interesting from a semi-academic point of view to see if <tt class="docutils literal"><span class="pre">call/cc</span></tt> can be
122
implemented on top of the primitives the stackless transform provides). Ruby
123
too (though the latter is probably more than two months of work), or Lua, or ...</p>
124
<p>We already have a somewhat usable <a class="reference" href="http://codespeak.net/svn/user/cfbolz/hack/prolog/interpreter">Prolog interpreter</a> and the beginnings of a
125
<a class="reference" href="../../pypy/lang/js">JavaScript interpreter</a>.</p>
127
<div class="section">
128
<h1><a class="toc-backref" href="#id6" id="investigate-restricted-execution-models" name="investigate-restricted-execution-models"><span id="security"></span>Investigate restricted execution models</a></h1>
129
<p>Revive <strong>rexec</strong>: implement security checks, sandboxing, or some similar
130
model within PyPy (which, if I may venture an opinion, makes more sense and is
131
more robust than trying to do it in CPython).</p>
132
<p>There are multiple approaches that can be discussed and tried. One of them is
133
about safely executing limited snippets of untrusted RPython code (see
134
<a class="reference" href="http://codespeak.net/pipermail/pypy-dev/2006q2/003131.html">http://codespeak.net/pipermail/pypy-dev/2006q2/003131.html</a>). More general
135
approaches, to execute general but untrusted Python code on top of PyPy,
136
require more design. The object space model of PyPy will easily allow
137
objects to be tagged and tracked. The translation of PyPy would also be a
138
good place to insert e.g. systematic checks around all system calls.</p>
140
<div class="section">
141
<h1><a class="toc-backref" href="#id7" id="experiment-with-distribution-and-persistence" name="experiment-with-distribution-and-persistence"><span id="persistence"></span><span id="distribution"></span>Experiment with distribution and persistence</a></h1>
142
<p>One of the advantages of PyPy's implementation is that the Python-level type
143
of an object and its implementation are completely independent. This should
144
allow a much more intuitive interface to, for example, objects that are backed
145
by a persistent store.</p>
146
<p>The <a class="reference" href="objspace-proxies.html#tproxy">transparent proxy</a> objects are a key step in this
147
direction; now all that remains is to implement the interesting bits :-)</p>
148
<p>An example project might be to implement functionality akin to the <a class="reference" href="http://www.zope.org/Documentation/Books/ZDG/current/Persistence.stx">ZODB's
149
Persistent class</a>, without the need for the _p_changed hacks, and in pure
150
Python code (should be relatively easy on top of transparent proxy).</p>
152
<div class="section">
153
<h1><a class="toc-backref" href="#id8" id="numeric-numpy-numarray-support" name="numeric-numpy-numarray-support">Numeric/NumPy/numarray support</a></h1>
154
<p>At the EuroPython sprint, some work was done on making RPython's annotator
155
recognise Numeric arrays, with the goal of allowing programs using them to be
156
efficiently translated. It would be a reasonably sized project to finish this
157
work, i.e. allow RPython programs to use some Numeric facilities.
158
Additionally, these facilities could be exposed to applications interpreted by
159
the translated PyPy interpreter.</p>
161
<div class="section">
162
<h1><a class="toc-backref" href="#id9" id="extension-modules" name="extension-modules">Extension modules</a></h1>
163
<p>Rewrite one or several CPython extension modules to be based on <strong>ctypes</strong>
164
(integrated in Python 2.5): this is generally useful for Python
165
developers, and it is now the best path to write extension modules that are
166
compatible with both CPython and PyPy. This is done with the <a class="reference" href="extcompiler.html">extension
167
compiler</a> component of PyPy, which will likely require some attention as
169
<p>Modules where some work is already done:</p>
171
<li><tt class="docutils literal"><span class="pre">_socket</span></tt>, <tt class="docutils literal"><span class="pre">os</span></tt>, <tt class="docutils literal"><span class="pre">select</span></tt> (unfinished yet, feel free to help;
172
see e.g. <a class="reference" href="http://codespeak.net/svn/pypy/dist/pypy/module/_socket/">http://codespeak.net/svn/pypy/dist/pypy/module/_socket/</a>).</li>
173
<li>SSL for socket, <tt class="docutils literal"><span class="pre">bz2</span></tt>, <tt class="docutils literal"><span class="pre">fcntl</span></tt>, <tt class="docutils literal"><span class="pre">mmap</span></tt> and <tt class="docutils literal"><span class="pre">time</span></tt>: part of the
174
Summer of Code project of Lawrence Oluyede
175
(<a class="reference" href="http://codespeak.net/svn/user/rhymes/">http://codespeak.net/svn/user/rhymes/</a>).</li>
177
<p>You are free to pick any other CPython module, either standard or third-party
178
(if relatively well-known, like gtk bindings).
179
Note that some modules exist in a ctypes version
180
already, which would be a good start for porting them to PyPy's extension
183
<div class="section">
184
<h1><a class="toc-backref" href="#id10" id="extend-py-execnet-to-a-peer-to-peer-model" name="extend-py-execnet-to-a-peer-to-peer-model">Extend py.execnet to a peer-to-peer model</a></h1>
186
<li>Work on a P2P model of distributed execution (extending <a class="reference" href="http://codespeak.net/py/current/doc/execnet.html">py.execnet</a>)
187
that allows <a class="reference" href="http://codespeak.net/py/current/doc/test.html">py.test</a> and other upcoming utilities to make use of a
188
network of computers executing python tasks (e.g. tests or PyPy build tasks).</li>
189
<li>Make a client tool and according libraries to instantiate a (dynamic) network
190
of computers executing centrally managed tasks (e.g. build or test ones).
191
(This may make use of a P2P model or not, both is likely feasible).</li>
194
<div class="section">
195
<h1><a class="toc-backref" href="#id11" id="or-else" name="or-else">Or else...</a></h1>
197
<li>Constraint programming: <a class="reference" href="http://codespeak.net/svn/pypy/extradoc/soc-2006/constraints.txt">efficient propagators for specialized
198
finite domains</a> (like numbers, sets, intervals).</li>
199
<li>A <a class="reference" href="http://codespeak.net/svn/pypy/extradoc/soc-2006/code-templating.txt">code templating solution</a> for Python code, allowing to extend
200
the language syntax, control flow operators, etc.</li>
202
<p>...or whatever else interests you!</p>
203
<p>Feel free to mention your interest and discuss these ideas on the <a class="reference" href="http://codespeak.net/mailman/listinfo/pypy-dev">pypy-dev
204
mailing list</a>. You can also have a look around our <a class="reference" href="index.html">documentation</a>.</p>
b'\\ No newline at end of file'