00:21 -!- szymon(n=szymon@bzo253.neoplus.adsl.tpnet.pl) has joined #maxima
02:31 -!- Prax01D(n=Prax01D@tor58-23b-95-33.dynamic.rogerstelecom.net) has joined #maxima
06:35 -!- stw__(n=stw@e176155022.adsl.alicedsl.de) has joined #maxima
08:12 <SerajewelKS> HaraldG_log: no, not that i can tell
08:13 <SerajewelKS> HaraldG_log: e.g. if i do something like:   fp:ev(diff(f,x))  then the value in fp is "ev(diff(f,x))", not the derivative of f with respect to x
08:27 <SerajewelKS> at least as near as i can tell
08:27 <SerajewelKS> or i'm on crack, who knows
08:37 -!- Prax01D(n=Prax01D@tor58-5-76-74.dynamic.rogerstelecom.net) has joined #maxima
08:47 -!- cunzhang(n=apzc2529@218.7.43.195) has joined #maxima
12:36 <SerajewelKS> is there a way to put maxima in "double" mode, whereby all math is performed with double precision arithmetic instead of big numbers?
12:37 <SerajewelKS> for doing numerical analysis
12:41 -!- cunzhang(n=apzc2529@218.7.43.195) has joined #maxima
12:43 <SerajewelKS> note that using float() on my answer isn't enough -- every step of the process must use double precision, so if i do at(x^3+x,[x=1/8]) then every step in the evaluation of that formula must be done with double-presicion floating point arithmetic
12:46 <SerajewelKS> will the float evflag do this, or does that again just affect the final answer?
12:46 <SerajewelKS> obviously,  float(float(x)^3)+float(x) would do the right thing but isn't really... well, who wants to do that to a tenth order polynomial
14:07 -!- HaraldG(n=rldprog@g4v1.g4.wien.funkfeuer.at) has joined #maxima
14:11 <HaraldG> SerajewelKS: evflags only work inside ev() AFAIK, but there ist the numer flag. Perhaps numer:true; does what you want.
14:11 <HaraldG> Also note that if any operand is a float, then the operation should return a float too.
14:12 <HaraldG> 1/4+0.1 -> 0.35 thus it shouldn't be too difficult to do all operation with floatingpoint math.
14:19 <HaraldG> Ahh - just got what you want to do with fp. Learn about nouns and verbs. (And give me a few minutes to think about the problem ....)
14:23 <SerajewelKS> numer seems to do the trick, i've been playing with it for a while
14:23 <SerajewelKS> i calculated macheps: the machine epsilon
14:23 <SerajewelKS> is(ev(1+macheps/2-1,numer) = float(0))
14:24 <SerajewelKS> is true over here, where using macheps (not macheps/2) is false
14:24 <SerajewelKS> interestingly,  0 # float(0)  but  0 < float(1)
14:25 <SerajewelKS> apparently = and # always return false when comparing real to floating-point, but the other relational operators work just fine
14:27 <HaraldG> I think this is intentional because 0 means exactly zero while 0.0 is only an approximation.
14:27 <HaraldG> There are some bugs though...
14:28 <SerajewelKS> well as long as i specify 0.0 it works as expected
14:28 <SerajewelKS> when calculating the machine epsilon i want to compare to the approximation specifically anyway
14:28 <SerajewelKS> so i have:   macheps: block([eps: 1], while is(float(1 + eps) # 1.0) do eps: eps / 2, eps * 2);
14:35 -!- Prax01D(n=Prax01D@tor58-7-82-55.dynamic.rogerstelecom.net) has joined #maxima
14:38 <SerajewelKS> is there a way to declare a function local to a block?
14:39 <SerajewelKS> or is that a lambda or whatever
14:40 <SerajewelKS> yup that seems to work
14:41 <HaraldG> To come back to your problem with fp:
14:42 <HaraldG> This wont work because symbols (like fp) are always evaluated once.
14:42 <SerajewelKS> actually i think i figured that part out
14:42 <HaraldG> (Unless you use ev or some other tricky function)
14:42 <SerajewelKS> my issue was that i was defining a function and i wanted to store in the function the result of the right-hand side
14:42 <HaraldG> But if you really insist on your notation you can to the following trick:
14:43 <HaraldG> fp() := ev(diff(f(x), x), nouns);
14:43 <HaraldG> nofix(fp);
14:44 <SerajewelKS> i just did   fp: diff(f, x)
14:44 <SerajewelKS> and i use at(f,[x=...])
14:44 <SerajewelKS> fp rather
14:47 <SerajewelKS> i'm not sure how efficient that is but it works
14:50 <HaraldG> Never did it that way before, but it seems to work. So all you want to do is get the numerical value ot the derivative at some point?
14:50 <SerajewelKS> right, or more specifically, at multiple points, without finding the derivative each time
14:51 <SerajewelKS> let me pastebin an example
14:51 <HaraldG> I think I got it.
14:51 <SerajewelKS> http://pastebin.ca/916254
14:52 <SerajewelKS> this is my implementation of newton's root-finding method after about 40 hours of hacking with maxima :)
14:52 <SerajewelKS> (not 40 hours of hacking *this* function, but with 40 hours of experience)
14:52 <SerajewelKS> ugh, i ev(..., numer) twice
14:53 <SerajewelKS> ok pretend line five is just g:xf(g)
14:56 <HaraldG> Didn't try it, but I think with block([numer:true, ... you should get rid of ev() entirely?
14:56 <SerajewelKS> hmm
14:57 <SerajewelKS> cool
14:59 <SerajewelKS> yep, seems to work
15:00 -!- szymon(n=szymon@cbo220.neoplus.adsl.tpnet.pl) has joined #maxima
15:02 <SerajewelKS> this stuff has a really steep learning curve but is nice when you get into it
15:03 <SerajewelKS> and aren't royally frustrated by simple crap
15:48 -!- cunzhang(n=apzc2529@218.7.43.195) has joined #maxima
15:58 -!- cunzhang(n=apzc2529@218.7.43.195) has left #maxima
16:13 -!- eggauah(n=daniel@20845108.cps.virtua.com.br) has joined #maxima
17:11 -!- stw(n=PC_40@195.80.220.100) has joined #maxima
18:39 -!- stw(n=PC_40@195.80.220.100) has left #maxima
20:36 -!- Prax01D(n=Prax01D@tor58-5-77-38.dynamic.rogerstelecom.net) has joined #maxima
21:09 -!- magni_(n=magni@ip68-14-27-175.ri.ri.cox.net) has joined #maxima
21:10 <magni_> anyone here using imaxima on OS X?
23:31 -!- szymon(n=szymon@cbo220.neoplus.adsl.tpnet.pl) has joined #maxima