Re: [Python-Dev] standard library mimetypes module pathologically broken?
Jacob Rus wrote: Brett Cannon wrote: Jacob Rus wrote: At the very least, I think some changes can be made to this code without altering its basic function, which would clean up the actual mime types it returns, comment the exceptions to Apache and explain why they're there, and make the code flow understandable to someone reading the code. That all sounds reasonable. Okay, as a start, I did a simple code cleanup that I think fixes some potential bugs (any code using its own instance of the MimeTypes class should now be insulated from other same-process users of the module), chops out 80 or 90 lines, removes some redundant code paths, clarifies some of the micro level behavior of some chunks of code, adds a bit more to the docstring at the top of the file, and makes the program flow somewhat clearer … *without* changing the semantics of the module or its included list of MIME types. Here is a somewhat more substantively changed version. This one does away with the 'inited' flag and the 'init' function, which might be impossible given that their documented (though I would be extremely surprised if anyone calls them in third-party code), and makes the behavior of the code much clearer, I think, by making it very obvious how the singleton instance is actually working. Additionally, this version brings the lazy loading of Apache mime.types files to every MimeTypes instance, and makes the read_mime_types() function behave as expected (only getting the mapping from an apache mime.types file rather than including some extra types as the current code does). In this version, tests would want to call the _init_singleton() function to reset to defaults. http://pastie.textmate.org/568399 http://pastie.textmate.org/568400 To reiterate: this should still behave identically to the current module in all reasonable conditions. I still haven't made any changes to the set of MIME types included in the file, or the behavior of the module. Some such changes should be made as well, but the changes so far should be relatively uncontroversial, I hope. Cheers, Jacob Rus ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
Jacob Rus wrote: Jacob Rus wrote: Brett Cannon wrote: Jacob Rus wrote: At the very least, I think some changes can be made to this code without altering its basic function, which would clean up the actual mime types it returns, comment the exceptions to Apache and explain why they're there, and make the code flow understandable to someone reading the code. That all sounds reasonable. Okay, as a start, I did a simple code cleanup that I think fixes some potential bugs (any code using its own instance of the MimeTypes class should now be insulated from other same-process users of the module), chops out 80 or 90 lines, removes some redundant code paths, clarifies some of the micro level behavior of some chunks of code, adds a bit more to the docstring at the top of the file, and makes the program flow somewhat clearer … *without* changing the semantics of the module or its included list of MIME types. Here is a somewhat more substantively changed version. This one does away with the 'inited' flag and the 'init' function, which might be impossible given that their documented (though I would be extremely surprised if anyone calls them in third-party code), and makes the behavior of the code much clearer, I think, by making it very obvious how the singleton instance is actually working. Additionally, this version brings the lazy loading of Apache mime.types files to every MimeTypes instance, and makes the read_mime_types() function behave as expected (only getting the mapping from an apache mime.types file rather than including some extra types as the current code does). In this version, tests would want to call the _init_singleton() function to reset to defaults. http://pastie.textmate.org/568399 http://pastie.textmate.org/568400 To reiterate: this should still behave identically to the current module in all reasonable conditions. I still haven't made any changes to the set of MIME types included in the file, or the behavior of the module. Some such changes should be made as well, but the changes so far should be relatively uncontroversial, I hope. Please post the patches to the Python bug tracker: http://bugs.python.org Thanks Michael Foord Cheers, Jacob Rus ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/fuzzyman%40voidspace.org.uk -- http://www.ironpythoninaction.com/ http://www.voidspace.org.uk/blog ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
2009/8/2 Michael Foord fuzzy...@voidspace.org.uk: [...] In this version, tests would want to call the _init_singleton() function to reset to defaults. [...] Please post the patches to the Python bug tracker: http://bugs.python.org Thanks The patch you post should also patch the test suite to use your replacement initialisation function where needed (if you didn't already do that). Paul. ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
On Sat, 01 Aug 2009 23:37:18 -0700, Jacob Rus wrote: Here is a somewhat more substantively changed version. This one does away with the 'inited' flag and the 'init' function, which might be impossible given that their documented (though I would be extremely surprised if anyone calls them in third-party code) [snip] There seem to be quite a bunch of high-profile third-party modules relying on this interface, eg. Zope, Plone, TurboGears, and CherryPy. See http://www.google.com/codesearch?q=mimetypes.init+lang%3Apython for a more thorough listing. Given that most of them aren't ported to Python 3 yet, I guess, changing the semantics in 3.x seems not-too-bad to me. HTH, -- Robert Stargaming Lehmann ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] [regex] memory leak
John Machin wrote: Hi Matthew, Your post in c.l.py about your re rewrite didn't mention where to report bugs etc so I dug this address out of Google Groups ... Environment: Python 2.6.2, Windows XP SP3, your latest (29 July) regex from the Python bugtracker. Problem is repeated calls of e.g. compiled_pattern.search(some_text) -- Task Manager performance panel shows increasing memory usage with regex but not with re. It appears to be cumulative i.e. changing to another pattern or text doesn't release memory. Example: 8-- regex_timer.py import sys import time if sys.platform == 'win32': timer = time.clock else: timer = time.time module = __import__(sys.argv[1]) count = int(sys.argv[2]) pattern = sys.argv[3] expected = sys.argv[4] text = 80 * '~' + 'qwerty' rx = module.compile(pattern) t0 = timer() for i in xrange(count): assert rx.search(text).group(0) == expected t1 = timer() print %d iterations in %.6f seconds % (count, t1 - t0) 8--- Here are the results of running this (plus observed difference between peak memory usage and base memory usage): dos-prompt\python26\python regex_timer.py regex 100 ~ ~ 100 iterations in 3.811500 seconds [60 Mb] dos-prompt\python26\python regex_timer.py regex 200 ~ ~ 200 iterations in 7.581335 seconds [128 Mb] dos-prompt\python26\python regex_timer.py re 200 ~ ~ 200 iterations in 2.549738 seconds [3 Mb] This happens on a variety of patterns: w, wert, [a-z]+, [a-z]+t, ... Thanks for that, John. I've should've kept an eye on the Task Manager! :-) Now fixed. It's surprising how much time and effort is needed just to manage the memory! ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] pylinting the stdlib
On Sat, Aug 1, 2009 at 11:40 PM, Vincent Legollvincent.leg...@gmail.com wrote: Hello, I've fed parts of the stdlib to pylint and after some filtering there appears to be some things that looks strange, I've filled a few bugs to the tracker for them. buglist snipped Is this useless and taking reviewer's time for nothing ? Please advise, if this is deemed useful, I'll continue further I think this is valuable work---please do continue! Just out of interest, how many false positives did you have to filter out in finding the 5 cases above? Mark ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
[Python-Dev] standard library mimetypes module pathologically broken?
[It may be worth creating a patch; I think most of these comments would be better on the bug-tracker.] (1) In a few cases, it looked like you were changing parameter names between files and filenames. This might break code that was calling it with keyword arguments -- as I typically would for this type of function. (1a) If you are going to change the .sig, you might as well do it right, and make the default be knownfiles rather than the empty tuple. (2) The comment about why inited was set true at the beginning of the function instead of the end should probably be kept, or at least reworded. (3) Default values: (3a) Why the list of known files going back to Apache 1.2, in that order? Is there any risk in using too *new* of a MimeTypes file? I would assume that the goal is to pick up whatever changes the user has made locally, but in that case, it still makes sense to have the newest file be the last one read, in case Apache has made bugfixes. (3b) Also, this would improve cross-platform consistency; if I read that correctly, the Apache files will override the python defaults on unix or a mac, but not on windows. That will change the results on the majority of items in _common_types. (application vs text, whether to put an x- in front of the word pict.) (3c) rtf is listed in non-standard, but http://www.iana.org/assignments/media-types/ does define it. (Though whether to guess application vs text is not defined, and python chooses differently from apache.) (3d) jpg is listed as non-standard. It turns out that this is just for the inverse mapping, where image/jpg is non-standard (for image/jpeg) but that is worth a comment. (see #5) (3e) In _types_map, the lines marked duplicates are duplicate keys, not duplicate values; it would be more clear to also comment out the (first) line itself, instead of just marking it a duplicate. (Or better yet, to mention that it is just being added for the inverse mapping, if that is the case.) (4) Why bother to lazyinit?Is there any sane usecase for a MimeTypes that hasn't been inited? I see value in not reading the default files, but none in not reading at least the files that were asked for. I could see value in only partial initialization if there were several long steps, but right now, initialization is all-or-nothing. If the thing is useless without an init, then it makes sense to just get done it immediately and skip the later checks; anyone who could have actually saved time should just remove the import. -jJ ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
Jim Jewett wrote: [It may be worth creating a patch; I think most of these comments would be better on the bug-tracker.] I'm going to do that shortly. (1) In a few cases, it looked like you were changing parameter names between files and filenames. This might break code that was calling it with keyword arguments -- as I typically would for this type of function. Sorry, that was a mistake. (1a) If you are going to change the .sig, you might as well do it right, and make the default be knownfiles rather than the empty tuple. Seems reasonable. (2) The comment about why inited was set true at the beginning of the function instead of the end should probably be kept, or at least reworded. (3) Default values: (3a) Why the list of known files going back to Apache 1.2, in that order? Is there any risk in using too *new* of a MimeTypes file? I would assume that the goal is to pick up whatever changes the user has made locally, but in that case, it still makes sense to have the newest file be the last one read, in case Apache has made bugfixes. I did not change this in my patch, but I completely agree. Indeed, I think it makes more sense to grab the newest Apache mime.types and just include them with the standard library, either as an in-code python object, or as a mime.types file to be parsed. (3b) Also, this would improve cross-platform consistency; if I read that correctly, the Apache files will override the python defaults on unix or a mac, but not on windows. That will change the results on the majority of items in _common_types. (application vs text, whether to put an x- in front of the word pict.) Quite possibly true. It actually seems (3c) rtf is listed in non-standard, but http://www.iana.org/assignments/media-types/ does define it. (Though whether to guess application vs text is not defined, and python chooses differently from apache.) (3d) jpg is listed as non-standard. It turns out that this is just for the inverse mapping, where image/jpg is non-standard (for image/jpeg) but that is worth a comment. (see #5) (3e) In _types_map, the lines marked duplicates are duplicate keys, not duplicate values; it would be more clear to also comment out the (first) line itself, instead of just marking it a duplicate. (Or better yet, to mention that it is just being added for the inverse mapping, if that is the case.) I completely agree that this whole section should be considered carefully. Just any changes might have more impact on backwards compatibility than the code flow changes I made, so I thought they could be in a separate patch. (4) Why bother to lazyinit? Is there any sane usecase for a MimeTypes that hasn't been inited? Only because the original was written that way, back in 1997 or whatever. I don't think there's necessarily any need for it these days: reading the default files even should be blazingly fast, unless the disk is otherwise thrashing: each is about a a 37k file, and there are at most going to be 3 or 4 of them installed on one machine for different versions of Apache. Cheers, Jacob Rus ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
Robert Lehmann wrote: Jacob Rus wrote: Here is a somewhat more substantively changed version. This one does away with the 'inited' flag and the 'init' function, which might be impossible given that their documented (though I would be extremely surprised if anyone calls them in third-party code) [snip] There seem to be quite a bunch of high-profile third-party modules relying on this interface, eg. Zope, Plone, TurboGears, and CherryPy. See http://www.google.com/codesearch?q=mimetypes.init+lang%3Apython for a more thorough listing. Given that most of them aren't ported to Python 3 yet, I guess, changing the semantics in 3.x seems not-too-bad to me. Ooh, okay. Well I guess we can’t get rid of those then! Michael Foord wrote: Please post the patches to the Python bug tracker: I made a new issue on the bug tracker, http://bugs.python.org/issue6626, and added a new patch which should hopefully be fairly reasonable. I still haven't addressed the issue of which MIME types should be included by default, and how precisely the logic should work for setting those up. But again, hopefully this at least makes it clear what the code is trying to do, so that it's relatively readable for someone trying to use the module. (For instance, so they'll be warned off of using init() and breaking each-other's code) Paul Moore wrote: The patch you post should also patch the test suite to use your replacement initialisation function where needed (if you didn't already do that). Done. The tests still pass, though to be honest this test suite isn't really testing any edge cases. Cheers, Jacob Rus ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] standard library mimetypes module pathologically broken?
On Sun, Aug 2, 2009 at 4:17 PM, Jacob Rus jacobo...@gmail.com wrote: Robert Lehmann wrote: Jacob Rus wrote: Here is a somewhat more substantively changed version. This one does away with the 'inited' flag and the 'init' function, which might be impossible given that their documented (though I would be extremely surprised if anyone calls them in third-party code) [snip] There seem to be quite a bunch of high-profile third-party modules relying on this interface, eg. Zope, Plone, TurboGears, and CherryPy. See http://www.google.com/codesearch?q=mimetypes.init+lang%3Apython for a more thorough listing. Given that most of them aren't ported to Python 3 yet, I guess, changing the semantics in 3.x seems not-too-bad to me. No, it's bad. If I may quote Guido: http://www.artima.com/weblogs/viewpost.jsp?thread=227041 So, once more for emphasis: *Don't change your APIs at the same time as porting to Py3k!* Please follow this policy as much as possible in the standard library; the language transition is going to be hard enough. Put a different way: please don't change the library unless you're *also*going to write a 2to3 fixer that somehow updates all calling code, too. Ooh, okay. Well I guess we can’t get rid of those then! Indeed not. ___ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com