I don't care about ORM performance which is worse both in term of bare performance and in development speed, but using large databases and doing profiling I find that most of the time that is "spent on DB IO" is ispent inside python code unless executesql is used. There is still a lot of room for improvement even if a lot has been done already.
Our target should be that in the future: rows = db(query).select(...) vs. sql=db(query)._select(...) rows = db.exequtesql(sql) have a not a difference of one order of magnitude anymore. 2013/7/13 Massimo Di Pierro <[email protected]> > It really depends on what one compares. > > I do not think we ever published benchmarks claiming web2py is faster than > other frameworks. If one want speed raw WSGI (perhaps with gunicorn or > gevent) without templates and without databases beats every framework. The > more a framework does, the slower it gets. I do claim web2py is better than > other frameworks because it is easier to use and because it does more for > you than other frameworks do. > > According to my tests: > - the DAL is faster in generating SQL queries than ORM based on > metaclasses but this is pointless since all the time is in DB IO, not in > generating queries. That time is the same. Django in some cases > (postgresql) has the ability of returning an iterators while web2py always > returns the data. The Django approach is better from a memory management > way and faster is only plan to loop over a subset of the selected records. > If you plan to loop over all of the records selected the web2py approach is > better. > - The web2py template language is faster than the Django templates because > uses the python interpreter and not its own. Yet it depends on the details > of the view page. > - Apart for templates and DAL, the execution of models and controllers is > slower in web2py. In Django models are executed only once (at startup) > while in web2py at every request. We do this for a reason. It is a small > performance penalty to pay for a greater flexibility (multi-app support, > hot plug and play, no library conflicts, etc.). This only hits in hello > world kind of tests and in the case of very very long models (that is why > some of use prefer to put code in modules and not in models) and then > import the modules in models. > > Massimo > > On Friday, 12 July 2013 18:44:11 UTC-5, Arnon Marcus wrote: > >> I seem to remember vividly seeing a benchmark about 4 years ago that >> showed that web2py was indeed the fastest. >> >> Either that benchmark was misleading, or things have changed. >> >> In either case, it would still remain important to explain to the public >> the reasons for the trade-offs that were chosen. If "speed-of-development" >> or "security" had to come with a performance-penalty, and still it was a >> conscious-choice to for-go with performance, than this trade-off choice >> should be out there in the open and explicit. It's a form of >> expectation-management and quality public-relation. >> >> If people out-there understand that "hello-world" benchmarks are >> meaningless for real-world use-cases, as the bottleneck always end-up being >> the database whenever one is used (which is practically 100% of real-world >> use-cases), they would be more apt to trying-out web2py, and have less >> irrelevant/erroneous-biases against it. >> >> And if the DAL is really faster than an ORM, than the bottleneck should >> theoretically be less severe in web2py, since less queries would be >> executed to the database. Meaning, that any performance-loss of >> non-database-related executions, should be grossly overshadows and >> nullified by virtue of having fewer database accesses. Meaning, that for >> real-world production use - web2py ultimately should be faster for >> database-intensive applications. >> >> >> On Fri, Jul 12, 2013 at 5:11 AM, Ovidio Marinho <[email protected]>wrote: >> >>> Web2py can not be the fastest, but it is the most simple and functional >>> for everyone. >>> >>> >>> >>> >>> Ovidio Marinho Falcao Neto >>> ITJP.NET.BR >>> [email protected] >>> 83 8826 9088 - Oi >>> 83 9336 3782 - Claro >>> Brasil >>> >>> >>> >>> 2013/7/12 Massimo Di Pierro <[email protected]> >>> >>>> Because I think it is pointless for various reasons: >>>> >>>> 1) I am biased and people outside the community would not trust it >>>> 2) Code changes so it would become obsolete quick >>>> 3) One can produce benchmarks to produce almost any result one wants >>>> 4) People who are concerned about 2x factors in speed are not web2py >>>> audience. The point of web2py is security and speed of development. For >>>> example we can make web2py more than 2x faster moving sessions management >>>> to the app level (like Flask) and not creating the request.env object. Yet >>>> we do not do it. What matters is not speed but scalability and web2py >>>> scales no better or worse than any other major Python framework. >>>> 5) Every time I published any comparison between web2py and other >>>> frameworks in the past, somebody got upset and attacked us. Not worth it. >>>> >>>> On Thursday, 11 July 2013 20:44:07 UTC-5, viniciusban wrote: >>>> >>>>> Massimo, how about you writing an article about this subject and share >>>>> with us? >>>>> >>>>> So, this could be spread. >>>>> >>>>> On Thu, Jul 11, 2013 at 2:12 PM, Massimo Di Pierro >>>>> <[email protected]> wrote: >>>>> > I agree. I will do. >>>>> > >>>>> > On Thursday, 11 July 2013 11:51:39 UTC-5, Arnon Marcus wrote: >>>>> >> >>>>> >> I see. >>>>> >> In that case, I think it would be advisable to note that in >>>>> presentations, >>>>> >> as peope might get the wrong impression... >>>>> >> >>>>> >> On Thursday, July 11, 2013, Massimo Di Pierro < >>>>> [email protected]> >>>>> >> wrote: >>>>> >> > It is true but not an issue. Django is faster only in hello world >>>>> >> > examples because does not perform as many header >>>>> validation/conversions as >>>>> >> > web2py does and because you cannot turn off sessions in web2py. >>>>> As soon as >>>>> >> > one uses templates, web2py is faster. If you use databases the >>>>> speed is >>>>> >> > about the same because that becomes the bottle neck. >>>>> >> > Massimo >>>>> >> > >>>>> >> > On Monday, 8 July 2013 17:08:53 UTC-5, Arnon Marcus wrote: >>>>> >> >> >>>>> >> >> BTW, is it really true that web2py is twice as slower than >>>>> django >>>>> >> >> nowadays? >>>>> >> >> How can that be? >>>>> >> >> Didn't it used to be twice as fast? >>>>> >> >> When I first evaluated it 3 years ago, it was by-far the fastest >>>>> - what >>>>> >> >> changed? >>>>> >> >> You said that one of the core principles of accepting changes to >>>>> >> >> web2py, is that they should always make it run faster - never >>>>> slower. Has >>>>> >> >> that principle been broken? >>>>> >> >> And what about the whole ORM-vs-DAL fiasco? Didn't you guys >>>>> always say >>>>> >> >> that a DAL is always faster than an ORM? Given that it makes >>>>> sense, how come >>>>> >> >> Django is faster? Shouldn't it's ORM make it slower? >>>>> >> >> And as for executing-vs-importing - didn't you say before that >>>>> it >>>>> >> >> should be a non-issue in terms of performance? What changed? >>>>> >> > >>>>> >> > -- >>>>> >> > >>>>> >> > --- >>>>> >> > You received this message because you are subscribed to a topic >>>>> in the >>>>> >> > Google Groups "web2py-users" group. >>>>> >> > To unsubscribe from this topic, visit >>>>> >> > https://groups.google.com/d/**to**pic/web2py/kWwIuQjeatQ/** >>>>> unsubsc**ribe<https://groups.google.com/d/topic/web2py/kWwIuQjeatQ/unsubscribe>. >>>>> >>>>> >> > To unsubscribe from this group and all its topics, send an email >>>>> to >>>>> >> > [email protected]. >>>>> >> > For more options, visit https://groups.google.com/**grou** >>>>> ps/opt_out <https://groups.google.com/groups/opt_out>. >>>>> >> > >>>>> >> > >>>>> >> > >>>>> > >>>>> > -- >>>>> > >>>>> > --- >>>>> > You received this message because you are subscribed to the Google >>>>> Groups >>>>> > "web2py-users" group. >>>>> > To unsubscribe from this group and stop receiving emails from it, >>>>> send an >>>>> > email to web2py+un...@**googlegroups.com. >>>>> > For more options, visit >>>>> > https://groups.google.com/**grou**ps/opt_out<https://groups.google.com/groups/opt_out>. >>>>> >>>>> > >>>>> > >>>>> >>>> -- >>>> >>>> --- >>>> You received this message because you are subscribed to the Google >>>> Groups "web2py-users" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to web2py+un...@**googlegroups.com. >>>> >>>> For more options, visit >>>> https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/opt_out> >>>> . >>>> >>>> >>>> >>> >>> -- >>> >>> --- >>> You received this message because you are subscribed to a topic in the >>> Google Groups "web2py-users" group. >>> To unsubscribe from this topic, visit https://groups.google.com/d/** >>> topic/web2py/kWwIuQjeatQ/**unsubscribe<https://groups.google.com/d/topic/web2py/kWwIuQjeatQ/unsubscribe> >>> . >>> To unsubscribe from this group and all its topics, send an email to >>> web2py+un...@**googlegroups.com. >>> For more options, visit >>> https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/opt_out> >>> . >>> >>> >>> >> >> -- > > --- > You received this message because you are subscribed to the Google Groups > "web2py-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > -- --- You received this message because you are subscribed to the Google Groups "web2py-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
