Mailman 3 June 2012 - pypy-dev

PyPy 1.9 on LaunchPad PPA?
by Peter Cock 22 Jun '12

22 Jun '12

Hello all, I'd like to be able to run tests for a Python project on Travis-CI under PyPy 1.9, however to provide PyPy on their virtual machines Travis-CI get PyPy from the PyPy PPA on LaunchPad: http://about.travis-ci.org/docs/user/languages/python/ https://launchpad.net/~pypy/+archive/ppa It appears that currently the latest release on the PPA is still PyPy 1.8 (and sadly for the project I have in mind, I require PyPy 1.9). Could someone update the PPA please? Thanks, Peter

3 3

__format__ support for float64 type in NumPyPy
by Cameron Sparr 21 Jun '12

21 Jun '12

Hi all, I noticed that there is no support for the .__format__() function with float64 data types in NumPyPy... I'm pretty new and was wondering if this is something that I should submit a bug report for? Cameron -- Cameron Sparr SIParCS Intern NCAR \ CISL \ User Services Section (USS) phone: 303-497-2443 mobile: 503-327-5964 email: csparr(a)ucar.edu

3 2

Internship and bug report
by Leonard de HARO 21 Jun '12

21 Jun '12

Hello, I'm a french undergraduated CS. student currently doing an internship in Santigo's University under the supervision of Eric Tanter. I'm working on the meta JIT compiler provided in the pypy project, trying to document it's performance on different kind of interpreters. I started two weeks ago. I have a problem with the translation toolchain. I've been writing an interpreter for a small language inspired by Shriram Krishnamurthi's F1WAE from his book "Programming Languages: Application and Interpretation". My problem is the following: I have an interpreter that seems to work perfecty (at least on my examples) when I use pypy interpreters, but whenever I translate it, execution fails. The fact is: I have identified the source of the problem but I don't understand it. If you look at the parser file, on line 184, there is a "print" instruction. With this instruction on comment, the translated version of RPinterpret executed on test10run10 gets a Stack Overflow. Use a translated version of the same file (RPinterpret.py) with the print instruction in parser.py, you get the normal behavior (which you have anyway using "pypy RPinterpret.py test10runs10"). Looks like a bug, but since I'm a beginer (both on this work and on Python in general), maybe I have missed something. I have the same problem with the supposely-JITing interpreter, but it's worse since the translations takes 20min against 1min, so changing and experiencing on the JITing file seems much harder. Anyway, thanks for your help. I didn't know whether I was supposed to make a bug report or just contact the mailing list, but since I'll be working on this until september, might as well get in touch with you right now. Leonard de HARO PS: As I said, I just learned Python, so maybe my files aren't verry pretty, I apologize for that and I'll welcome any comment on my work you could make. Thanks again.

6 7

A very easy bug for newcomers
by Maciej Fijalkowski 20 Jun '12

20 Jun '12

Hi This is a good start for someone wanting to explore contributing to pypy - https://bugs.pypy.org/issue1180 I'll fix it if noone steps up :) Cheers, fijal

4 13

set with instances very slow from 1.8 to 1.9
by wilk 20 Jun '12

20 Jun '12

Hi, I could isolate a piece of my code that show a big regression from 1.8 to 1.9, event from cpython to 1.9 with operations on set of instances. On the other side set of integer are faster ! Should i open an issue on the bug tracker ? $ python t.py with instances = 0.02350 with integers = 0.02825 $ pypy-1.8 t.py with instances = 0.00935 with integers = 0.02891 $ pypy-1.9 t.py with instances = 0.41753 with integers = 0.00628 pypy-last/bin/pypy t.py with instances = 0.34686 with integers = 0.00628 import random import time import sys class X: pass def bcl(a,b): for i in range(1000): a |= b x1 = set([X() for i in range(1000)]) x2 = set([X() for i in range(1000)]) start = time.time() bcl(x1, x2) print "with instances = %.05f" % (time.time()-start) x1 = set(range(1000)) x2 = set(range(1000)) start = time.time() bcl(x1, x2) print "with integers = %.05f" % (time.time()-start) -- William Dodé Informaticien Indépendant

3 4

CFFI release 0.1
by Armin Rigo 18 Jun '12

18 Jun '12

Hi all, Announcing the release 0.1 of CFFI, the new Python interface to call C libraries done (mostly) by Maciek and myself: http://morepypy.blogspot.ch/2012/06/release-01-of-cffi.html A bientôt, Armin.

1 0

[cpyext] crash in PyThreadState_GetDict()
by Stefan Behnel 18 Jun '12

18 Jun '12

Hi, I'm getting reproducible segfaults in the PyThreadState_GetDict() function with lxml. For that, it's enough to build the latest lxml github master version in PyPy and start the test runner. You'll also need the github master version of Cython for that (with contains the PyPy port). I use these commands for building lxml under Linux: $ CFLAGS=-ggdb make PYTHON=pypy inplace $ pypy test.py -vv -p parser html # no specific test, just not all The stack trace I get starts like this: """ #0 0x0000000000b4fac5 in ?? () #1 0x000000000103a813 in PyThreadState_GetDict () #2 0x00007ffff51b15b0 in __pyx_f_4lxml_5etree__getGlobalErrorLog () at src/lxml/lxml.etree.c:32906 #3 0x00007ffff51a8922 in __pyx_f_4lxml_5etree_13_BaseErrorLog__receive (__pyx_v_self=0x7ffff7f99378, __pyx_v_error=0x3205ed8) at src/lxml/lxml.etree.c:28299 #4 0x00007ffff51fad46 in __pyx_f_4lxml_5etree__forwardParserError (__pyx_v__parser_context=0x3205c80, __pyx_v_error=0x3205ed8) at src/lxml/lxml.etree.c:82070 #5 0x00007ffff51fadc7 in __pyx_f_4lxml_5etree__receiveParserError (__pyx_v_c_context=0x3205c80, __pyx_v_error=0x3205ed8) at src/lxml/lxml.etree.c:82126 """ Background: lxml frees the GIL when it starts parsing XML, then reacquires it to report any errors that occur and, while holding it, looks up the target error log object from thread global storage to find out who wants to know about the error. That's where it crashes above, right when it tries to get at the TLS dict. Any help is appreciated. Stefan

2 5

new benchmark for pypy (without html version)
by Grzegorz Rudnicki 18 Jun '12

18 Jun '12

Hi, I maked new benchmark: import random,string, time t=time.time() def random_string(dl=10): d = [random.choice(string.letters) for x in xrange(dl)] s = "".join(d) return s basic=[] words=[] n=20000 #number of basic words n2=5000 #number of words in text #Create list of words variants for i in range(n): basic_word=random_string() if basic_word in basic: basic_word=random_string() basic.append(basic_word) tmp=[] tmp.append(basic_word) for j in range(4): tmp.append(basic_word+random_string(3)) words.append(tmp) #Create random text text=[] for i in range(n2): r1=random.randint(0,n-1) r2=random.randint(0,3) text.append(words[r1][r2]) #Replace non basic word to basic word for i in range(n2): for j in range(n): if text[i] in words[j]: text[i]=words[j][0] ok=0 for i in range(n2): if text[i] in basic: ok+=1 if ok==n2: print("OK") else: print("Fail") print(time.time()-t) This code tranform non basic word to basic word. For simply I use random strings. I tested this code on Linux and CPython use 7.8 MB RAM, but Pypy use 14.5 MB RAM. Grzegorz

3 2

A proposal about Numpypy missing functions
by Daπid 18 Jun '12

18 Jun '12

Hello, I have seen that there are some missing functions in numpypy that could be implemented, at least naïvely, very easily. An example of this is zeros_like, as we have zeros and shape. Another example is the random module, easy to port using list comprehension and the random module from the standard library. In any case, based on the benchmarks I have seen, I think the pypy version would be even faster than the pure CPython+Numpy. Would you be interested in a sub-optimal port of those functions? In my opinion, it is better to have something not great (that could be improved) that nothing at all. If so, what should I do? Regards, David.

3 3

HTMLParser compatibility with cPython 2.7.3
by Robert Zaremba 18 Jun '12

18 Jun '12

Hi, I would like to import changes from: The problem is that HTMLParser from 2.7.2 is not lenient and likes to throw exceptions, when html document is not well formed: http://bugs.python.org/issue13987 This often involves exception from BeautifoulSoup, which gains great speed up when using from pypy + HTMLParser from stdlib: "RuntimeWarning: Python's built-in HTMLParser cannot parse the given document. This is not a bug in Beautiful Soup. The best solution is to install an external parser (lxml or html5lib), and use Beautiful Soup with that parser. See http://www.crummy.com/software/BeautifulSoup/bs4/doc/#installing- a-parser for help." However lxml is not compatibile with PyPy, and html5lib is slow. Can I port the HTMLParser.py from python 2.7.3 to PyPy? -- Robert Zaremba

2 1