| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Comments:
<0> RadSurfer, the RFC.
<1> and just where might I find that ?
<1> grrrr....
<2> hopefully urlparse module knows them
<1> research! every programmers needs to do research
<1> that just breaks a url down...
<1> I need to find 'em first
<1> fascinating
<1> it lists the RFC's
<1> how handy is that !!
<3> would creating an object in the __init__ method of a cl*** increase stability? or no?
<4> RadSurfer: someone wrote a Perl program to *generate* a regep which recognises URL's: http://www.foad.org/~abigail/Perl/url2.html
<1> thanks.
<4> RadSurfer: click the "program" link for the source code
<2> reikon: what kind of "stability" are you thinking about?
<2> python is unlikely to crash any more or less
<3> drewp, I'm not used to python, so I'm not sure of all the issues it's prone to...mainly slowdowns and performance issues
<5> reikon: Do not worry about speed or performance.
<5> reikon: Not unless you know what you are doing anyhow.
<5> And at the moment neither you (nor I :) know what we are doing with python. :)
<3> haha
<3> I know what I'm doing, just not what I'm doing with python as you said :P
<5> For the vast majority of uses python will always be more than fast enough.
<0> reikon, I wouldn't see why nested construction wouldn't work as expected, but from C++ I have adopted the rule that no constructor call (such as __init__()) should ever fail.
<0> reikon, I would think it a bit more iffy and in need of careful testing if you were constructing more items of the SAME cl*** branch, thus recursing.
<3> hari`, __init__() it's self can't fail, only the funciton it calls if the user supplies an argument.
<0> reikon, sure, but what do you do then? Raise an exception? Very ugly.
<3> yes...actually..i do :P
<0> I would do two-stage construction then.
<2> exceptions aren't ugly
<0> exceptions from __init__ can be. You have this aborted half-initialized object out there. It would be uglier if there was a finalizer to have to contend with the mess.
<3> my object isn't half initialized..i have it to where only 1 variable is created, before it's p***ed to the function that raises the exception
<3> so no major processing has been done
<0> You can do what you like.
<2> hari`: how do you have a half-initialized object?
<3> i'm just trying to make it as fast as possible
<3> now that it actually works :P
<5> drewp: The object is created but the various attributes or whatever are not completely set?
<2> reikon: does your thing even work yet?
<2> tessier_: nah
<3> drewp, it works very well
<3> now i'm optimizing
<0> drewp, what if the function it was calling was sequestering resources of some kind. The finalizer may have to step carefully if some of the attributes aren't even there yet.
<2> reikon: oh- then what's the slow part? don't just ask if python features are fast or slow; find your bottleneck and address it
<0> I'm doomsaying and you're dismissing. The real level of concern is somewhere in between.
<4> if you ***ign to __slots__ instead of "dynamic" attributes then they will all exist (but be None) I think
<2> hari`: then __init__, like any other function, should release resources that it's not going to keep. that probably happens automatically in most cases
<3> drewp, there are no bottlenecks as i know of...I guess i'm just too used to other languages :P
<2> reikon: if it's too slow, there's a bottleneck. It might be very large, but that's unlikely
<2> have you profiled?
<3> nope...
<6> RELATION_OPTIONS = [('relate_work', 'Work'),]
<0> drewp, sure, but it's a lot less messy if you separate construction from all that double-blind-extra-safe-try-catch-everything code in the ctor and dtor.
<6> How do I extract 'relate_work'
<7> I'm looking to write something that you would use in the same way as twisted, but could also be incorporated as part of a larger framework containing multiple instances - any ideas?
<6> From a list with two tuple pairs like taht?
<7> an rtfm is fine if you suggest one
<2> Patrick`: what are "multiple instances" of a twisted-like lib?
<7> 2hmm
<0> brian, RELATION_OPTIONS[0][0]
<6> I mean like, a function....
<6> The list has more than one thing in it.
<7> well, it would abstract "being a client" for a given service
<7> but there could be a larger program that allows multiple instances
<7> I guess it wouldn't work
<6> Usually there is a better thing than a loop
<7> since it's event driven from that one socket/whatever
<0> brian, uh, I don't follow.
<3> drewp, i'll show you my code if you like and you can point out anything stupid i'm doing...
<0> Your example isn't two tuple pairs, either.
<2> reikon: please present output from a profiling p*** too, and then we can suggest actual improvements
<2> reikon: without that, we can probably make style suggestions only
<3> can you suggest a profiler?
<7> ok, my fundemental question, looking over these twisted examples, is when they do "from a import b" then "def Cl***Name(a)", how does that work?
<7> is that cl*** extension or something?
<7> and I'd like to write something that lets someone do the same thing with my code as a backend
<2> Patrick`: that wouldn't work, since "from a import b" doesn't put 'a' in the namespace
<7> drewp: you know what I mean, I just used the wrong temporary variables in my example
<7> from foo import a
<6> hari`: Like there's zip...
<6> hari`: Is there an unzip function?
<0> Did you look?
<6> hari`: I searched google
<6> hari`: didn't really find anything
<6> I'm finding stuff about compression
<2> Patrick`: then what's the question? Cl***Name is subcl***ing from 'a', which was in module foo
<7> subcl***ing, right
<7> I wanted to know what bit of OO programming I needed to learn about next
<0> Uh, 'def Cl***Name(a)'?
<0> def?
<6> How do I merge two lists?
<2> brian: the opposite of x=zip(a,b) is zip(*x)
<7> cl***
<7> meh
<3> drewp, heres the source: http://paste.insanityflows.net/viewpaste.php?id=49359 I'll look for a profiler.. i heard pylint was good
<7> I can't read
<7> but yes, that's what I mean
<2> reikon: stdlib profile is probabyl fine
<0> Patrick`, if you ask a question, get your words right or nobody can follow what you're asking. :)
<7> hari`: anyway, point is I need to read about subcl***ing
<6> How do I merge two lists?
<7> so I can build something that mimics twisted in the way you'd use it
<7> brian: depends
<7> list3 = list1 + list2
<2> brian: what does merge mean? x+y will concatenate them
<7> will append list2 to list1
<2> reikon: well, this is so short that maybe you already know which function you want to optimize. are your input strings long, or just numerous?
<3> possibly both, depending on who's using it
<2> reikon: well that's informative. What function, and for what type of input, would you like me to make optimization suggests for first?
<2> also, why do you make those local functions hex_string and bin_string? they seem to just make the code a bit harder to read
<3> drewp, engine_crypt or it's transparent methods preferably
<2> reikon: it may help on very long inputs to preallocate self.out, like "self.out = [None] * len(input)"
<3> to keep the code clean, and because I had already created them in other scripts, and it made more sense to make them nested functions
<3> ahhh
<3> i suppose i should do the same for state_array?
<2> reikon: but my move in a case like this would be to compile the program with pyrex. A few type hints would make this run about as fast as C-- you probably wouldn't even have to rewrite any of the algorithm
<2> reikon: no! a loop of length 256 is not worth complicating your code over
<2> reikon: my goal, and i hope it's your goal too, is to make the code run fast with the least amount of added junk
<3> it is, however it's also for easy usability
<2> _rc4_init is probably running very rarely, and probably takes no time
<3> hence the hex_string and bin_string methods
<2> how is the usability going to change? we're not messing with the cl*** API
<3> it reduced the code in my test vectors script
<3> it's not i was just stating my goals :P
<2> how do hex_string and bin_string make any difference to the users of the RC4 cl***?
<3> it allows them to get the output of the cipher in hex or p***ed through chr() without having to create the code themselves
<3> i've already made use of it in them each in two examples, and the test vectors cl***
<2> wait, i see the point of engine_crypt_hex and engine_crypt_bin, but why does each one have a nested function that does the work?
<8> hey, anyone knows why Python needs the new with clause, rather than just use a simple @run @with_stuff decorators?
<3> drewp, copy and paste from my other scripts. I created those methods and decided it'd be easier to implement them in the RC4 cl*** than to keep putting them in scripts
<3> that used the RC4 cl***
<3> but I see what you mean...
<2> reikon: you're making me suffer because you copy/pasted?
<3> hehe
<3> well....i see your point :P
<8> anyone here took a look at subtext?
<2> reikon: anyway, i have changed my mind again about the profiling. Is the majority of the time being spent in engine_crypt, or in one of the output-formatter functions?
<3> output formatter
<3> i believe...
<3> *check profiler*
<8> might as well use print profiles..
<8> or a statistic profiler..
<2> Peaker: i'd love a statistical profiler
<8> drewp: I always use one.. KeyboardInterrupt :-)
<2> Peaker: that is a fine one
<2> but i cannot help but wonder if there coudl be something, i don't know, a bit *richer*
<8> heh
<8> shouldn't be too tough to do
Return to
#python
or
Go to some related
logs:
#perl
pythonfu+win32
#php
perl win32 process send sigint
pam-login block shadow
undelete files knoppix ext3
swftool script
how to determine location of bash script
#gentoo
XFburn+burn dvd
|
|