Mailman 3 June 2010 - pypy-dev

Adding pure python implementation of datetime to CPython 3.x
by Alexander Belopolsky 17 Jun '10

17 Jun '10

Dear PyPy-Dev, I have started porting PyPy implementation of datetime.py to CPython 3.x with a goal of distributing it together with C implementation. The idea is for C and Python modules to share most of the regression test cases and allow platforms that cannot use C modules use python implementation. The benefit for CPython developers is that this will allow rapid prototyping and experimentation with new features. The benefit for CPython users is that they will be able to consult python implementation as ultimately detailed documentation. Since your project did an admirable job maintaining datetime.py in recent years, your input will be very much appreciated. Please see CPython tracker issue 7989 [1] for details and to leave comments. I already have a patch there which updates PyPy datetime.py to Python 2.7. You should be able to apply it to your tree when you upgrade to 2.7. -- [1] http://bugs.python.org/issue7989

1 0

Re: [pypy-dev] EuroPython Talk on Stackless.py and Questions
by Andrew Francis 17 Jun '10

17 Jun '10

Hi Armin: Message: 3 Date: Wed, 2 Jun 2010 16:19:56 +0200 From: Armin Rigo <arigo(a)tunes.org> Subject: Re: [pypy-dev] EuroPython Talk on Stackless.py and Questions To: Carl Friedrich Bolz <cfbolz(a)gmx.de> Cc: pypy-dev(a)codespeak.net Message-ID: <20100602141956.GA7829(a)code0.codespeak.net> Content-Type: text/plain; charset=us-ascii Hi Andrew, On Wed, Jun 02, 2010 at 03:59:21PM +0200, Carl Friedrich Bolz wrote: > > However I want to get deeper into PyPy. I would like to implement > > Select as a language feature as an exercise rather than an actual > > change to the language. > >> What exactly do you mean by "language feature"? >More precisely, I should say that this kind of changes to the syntax and >bytecode compiler are *really* uninteresting from our point of view. >PyPy is indeed about language implementations, but not about the >syntactic level. There are nice existing tools about generating parsers >in various languages; and as each of them has some issues with the Python language, we just wrote the parser and compiler by hand and >happily forgot about it. Armin, I understand. However I am approaching this from another angle. I view myself as an application programmer. I know the Stackless API relatively well and its algorithms. Enough to implement select (not hard). Outside of selecting the right switches for translate.py or just sticking to greenlets, there really isn't much PyPy to learn, unless I decide to re-write parts specifically in RPython for speed. I may do this so I can better acquaint myself with PyPy. Implementing select is an important exercise because it is a stepping stone for me (and hopefully others) to experiment with more exotic features. For example, I am becoming interested in Complex Event Processing that has pattern matching rules. One may want language features to support those. So I am interested in the low level details of how to alter PyPy's implementation of Python syntax. Maybe to help in the exercise, it would be nice if I can get advice about what exactly what I needed to change. Getting advice from Carl and Stephan about using greenlets *radically* speed up my ability to implement stackless.py - I would never figure that out on my own. Cheers, Andrew

2 1

Re: [pypy-dev] EuroPython Talk on Stackless.py and Questions
by Andrew Francis 17 Jun '10

17 Jun '10

Hi Carl: Message: 2 Date: Wed, 02 Jun 2010 15:59:21 +0200 From: Carl Friedrich Bolz <cfbolz(a)gmx.de> Subject: Re: [pypy-dev] EuroPython Talk on Stackless.py and Questions To: pypy-dev(a)codespeak.net Message-ID: <4C0663B9.8080109(a)gmx.de> Content-Type: text/plain; charset=ISO-8859-1; format=flowed >What exactly do you mean by "language feature"? As in alter the syntax of Python > I assume that select is so far simply a function that lives in >stackless.py, right? Yes. > I don't really see what other form select could take, so I also > don't see in what way you want to change the language. The Newsqueak/Limbo/Go family of languages show select and channels can be implemented as language features. It is from Limbo, where Stackless Python gets channels. >The bytecode interpreter and the parser live in interpreter/, the object >implementations in objspace/std/ and the modules in modules/. Thanks. I found it and started to look at it. > Also I wouldn't > mind learning more about the stackless transform. >Again, the stackless transformation doesn't really need to be touched to >implement select. True. However I would like to learn how the stackless transform works. Cheers, Andrew

1 0

Seeking advice re. implementing an interpreter in RPython
by Tom Locke 17 Jun '10

17 Jun '10

Hi Folks I'm probably in the wrong place, so I'll make this quick : ) I am working on an experimental programming language, and am considering building the next prototype of the interpreter in RPython. I just wanted to ask which mailing-list is the best one to join for help and advice on this topic? Thanks Tom

4 5

using libffi from rpython - a better way?
by Ryan Kelly 17 Jun '10

17 Jun '10

Hi All, First, let me say that PyPy just rocked my world. I develop an auto-update framework for frozen python applications [1] and have started using PyPy to compile parts of the app into stand-alone executables. I got it up and running in under two days, and it shaves several MB off the size of my frozen apps - so thanks for some awesome tech! I'm currently using libffi to dynamically find and load a python DLL from within an RPython program. It works well but the code seems very verbose to someone who's used to working with ctypes. I'm hoping there's a better way... Here's a trimmed-down example of the sort of thing I'm doing:: import ctypes.util libpython = ctypes.util.find_library("python2.6") from pypy.rlib.libffi import * from pypy.rpython.lltypesystem import rffi, lltype def target(*args): def entry_point(argv): # Find the python DLL and extract needed functions py = CDLL(libpython) Py_Initialize = py.getpointer("Py_Initialize",[],ffi_type_void) Py_Finalize = py.getpointer("Py_Finalize",[],ffi_type_void) PyRun_SimpleString = py.getpointer("PyRun_SimpleString",[ffi_type_pointer],ffi_type_sint) # Bootstrap into running a python program Py_Initialize.call(lltype.Void) buf = rffi.str2charp("print 'hello from python!'") PyRun_SimpleString.push_arg(buf) PyRun_SimpleString.call(rffi.INT) rffi.free_charp(buf) Py_Finalize.call(lltype.Void) return 0 return entry_point,None As you can imagine, the real application has a lot more of these boilerplate declarations. Any suggestions on how to do this in less/cleaner code? Thanks, Ryan [1] http://pypi.python.org/pypi/esky/ -- Ryan Kelly http://www.rfk.id.au | This message is digitally signed. Please visit ryan(a)rfk.id.au | http://www.rfk.id.au/ramblings/gpg/ for details

2 1

Re: [pypy-dev] Seeking advice re. implementing an interpreter in RPython
by William Leslie 17 Jun '10

17 Jun '10

This is the right mailing list. You might like to join us on irc for more involved questions, too. What are you working on? On 17/06/2010 6:22 PM, "Tom Locke" <tom(a)tomlocke.com> wrote: Hi Folks I'm probably in the wrong place, so I'll make this quick : ) I am working on an experimental programming language, and am considering building the next prototype of the interpreter in RPython. I just wanted to ask which mailing-list is the best one to join for help and advice on this topic? Thanks Tom _______________________________________________ pypy-dev(a)codespeak.net http://codespeak.net/mailman/listinfo/pypy-dev

1 0

can somebody talk at DZUG/Dresden?
by holger krekel 15 Jun '10

15 Jun '10

Hi all, i have been asked to talk at http://new.zope.de/tagung/Dresden_2010 about PyPy. But i can't make it there. Anybody interested in giving a talk about PyPy? Giving english talks is fine and happens a lot there. I think we have good base material from past conferences so if you are an informed follower of the project and are interested - drop me a note. best, holger

1 0

Re: [pypy-dev] [pypy-svn] r75220 - in pypy/trunk/pypy: annotation annotation/test jit/backend jit/backend/llgraph jit/backend/llgraph/test jit/backend/llsupport jit/backend/llsupport/test jit/backend/test jit/backend/x86 jit/backend/x86/test jit/codewriter jit/codewriter/test jit/metainterp jit/metainterp/test jit/tl jit/tl/spli jit/tl/tla jit/tool jit/tool/test module/pypyjit module/pypyjit/test objspace/flow rpython rpython/lltypesystem rpython/lltypesystem/test rpython/memory/gctransform/test rpython/mem
by Antonio Cuni 13 Jun '10

13 Jun '10

On 08/06/10 23:42, arigo(a)codespeak.net wrote: > The number of changes is a bit huge though. The > format of the static bytecodes used by the jit > changed completely, and codewriter.py is now split > among many files in the new directory pypy.jit.codewriter. > There is also no longer ConstAddr, only ConstInt: prebuilt > addresses are now turned into integers (which are > symbolic, with the new class AddressAsInt, so they can > be rendered by the C translation backend). Various > related changes occurred here and there. I didn't look at the branch deeply, but the last sentence looks suspiciously hard/impossible to implement in ootype. Could you explain why ConstAdrr were bad please? ciao, Anto

2 2

Merging branch/blackhole-improvements
by Armin Rigo 08 Jun '10

08 Jun '10

Hi Fijal, The branch/blackhole-improvement is ready to be merged as far as I can tell. Do you have any issue (as release manager) if I try to merge it now? A bientot, Armin.

1 0

EuroPython Talk on Stackless.py and Questions
by Andrew Francis 02 Jun '10

02 Jun '10

Hi Folks: I understand that stackless is not high on the pypy team priority list but here goes.... My talk "Prototyping Go's Select for Stackles Python in Stackless.py" was accepted. I have never been to Europe. To date, I have been able to implement the Go Select like capability in a number of ways. Most recently I emulated the Plan9 approach which is pretty straight forward. I need to add the code for channel preferences and the channel callback so it is on par with the old stackless.py Questions: Where should I put the new code? Are there regression tests for stackless.py Right now, I am taking Stephan Diehl and Carl Bolz's good advice and using stackless.py with greenlets. However for completeness, I would like to run through the exercise of using the translate tool chain. Right now, select is a function. I figure implementing Select as a language feature would be a good way to familarise myself better with the PyPy framework at a deeper level. However I am not sure where to start. Where is the parser? Do I create new opt codes? Is this doable for a newbie in a month? Cheers, Andrew

4 4