As Vasile suggest the future seems to point to cffi. It would work on PyPy (native target) and CPython and would allow prebuild binaries to be packed in the distributions. Maybe it is time to start working on it...
mic 2012/8/23 spiffytech <[email protected]>: > I'd like to cast my vote against implementing something as core as Storage > in C, as that hinders web2py's support for PyPy and other alternate runtimes > that don't play nice with C modules. > > > On Wednesday, August 22, 2012 1:15:45 PM UTC-4, Massimo Di Pierro wrote: >> >> Some more info.... the issue page claims this issue is fixed in pyhton >> 2.7.3 but my tests and Michele's test revealed it is not in fact fixed. >> >> I changed trunk gain: >> - reverted all improvements to dal.py except for lazy tables >> - made the improvement to Storage conditional (it detects the bugs and >> uses the fast or the slow version accordingly). >> >> Problem is that I am not aware of a python version without the bug (2.7.3 >> is the latest in 2.x and has it). >> >> We will try look for other solutions but this does not look good. >> >> There is one more option... implement our own Storage from scratch using >> the C-API. I am sure we can make it very fast and meet our needs. I do not >> have enough experience with this but I will try look into it. Will not >> happen in web2py 2.0. >> >> Massimo >> >> On Wednesday, 22 August 2012 11:50:49 UTC-5, LightDot wrote: >>> >>> There are many hosting companies that won't upgrade OS provided python >>> versions for the lifetime of a server. Good luck requesting a python upgrade >>> on a Red Hat / CentOS / Scientific Linux server, this simply won't happen. >>> >>> This is the situation at the moment: >>> >>> RHEL 6: python 2.6.5 >>> Debian 6: python 2.6.6 >>> Ubuntu 10.4 LTS: python 2.6.5 >>> Ubuntu 12.4 LTS: python 2.7.3 >>> >>> The proper way to handle such a change would be to make this feature >>> optional and file bug reports with all linux distributions that use python < >>> 2.7.3 at the same time. I'm sure Red Hat would backport this into RHEL 6, >>> which takes care of CentOS, SL and others. Can't say about Debian or Ubuntu, >>> but I don't see why not at least try. >>> >>> But, should these kind of major changes even be considered on the eve of >>> the new release..? >>> >>> Regards, >>> Ales >>> >>> >>> On Wednesday, August 22, 2012 6:22:06 PM UTC+2, Khalil KHAMLICHI wrote: >>>> >>>> if we go back the future will not wait for us, and we will become >>>> obsolete. >>>> Python versions are easy to manage on absolutely any platform or OS, I >>>> say we keep it. >>>> >>>> >>>> On Wed, Aug 22, 2012 at 2:52 PM, Anthony <[email protected]> wrote: >>>>> >>>>> If we can't fix the memory leak, can we at least have the code detect >>>>> the Python version and use the faster Storage in case >= 2.7.3, or are >>>>> there >>>>> too many places where code would need to change to make that practical? >>>>> >>>>> Anthony >>>>> >>>>> >>>>> On Wednesday, August 22, 2012 9:28:59 AM UTC-4, Massimo Di Pierro >>>>> wrote: >>>>>> >>>>>> We run into a major problem. The Storage improvements in trunk cause a >>>>>> memory leak. >>>>>> This is a python bug and it was discovered in 2006.: >>>>>> >>>>>> http://bugs.python.org/issue1469629 >>>>>> >>>>>> Apparently this was only fixed in python 3.2 and python 2.7.3. >>>>>> >>>>>> Not sure what to do but at the moment I do not see any solution than >>>>>> removing the improvements. >>>>>> >>>>>> massimo >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> On Wednesday, 22 August 2012 01:44:44 UTC-5, Johann Spies wrote: >>>>>>> >>>>>>> I hope that web2py 2.0 will not be released without these >>>>>>> improvements even if it means that testing it will delay the release. >>>>>>> >>>>>>> Thanks to the developers for their initiatives and ingenuity. >>>>>>> >>>>>>> Regards >>>>>>> Johann >>>>>>> -- >>>>>>> Because experiencing your loyal love is better than life itself, >>>>>>> my lips will praise you. (Psalm 63:3) >>>>>>> >>>>> -- >>>>> >>>>> >>>>> >>>> >>>> > -- > > > --
