IRC channel logs
2017-03-24.log
back to list of logs
<sirgazil>ACTION is using Guile 2.2 now. Thanks to all people who work on it :) <MoronicAcid>How would I go about making a typed array containing real numbers? <lloda-home>e.g. (make-typed-array 'f64 0. 10 10) makes a 10x10 2d array filled with 0. <amz3>catonano: if culturia, works, don't do 'git pull' because master is broken <amz3>the state of the project is not very nice <amz3>I am not sure what do to next <amz3>especially since the code I've done yesterday is broken today <amz3>My plan was to parse wiktionary <amz3>mbuf: what do you seek exactly? <mbuf>amz3, something like ruby on rails guides <amz3>I am not familiar with rails guides, what are they? <mbuf>amz3, if I have a Rails application, how can I re-implement the same using artanis, for example <amz3>nalaginrut: is the maintainer of artanis <mbuf>amz3, the present document is like a reference manual <mbuf>I would like to see some performance metrics as well <amz3>I've done 5 years of Django, I never seen any performance metrics about it <amz3>wait... I saw some performance metrics actually <amz3>anyway, it would not be fair to compare rails/django to artanis since they don't implement the same feature <amz3>as of right now, you can't be as productive using guile web infrastructures as you would be in ruby or python <amz3>there is still a lot of plumbing to do <amz3>I mean, for instance, there is a no form validation library <amz3>mbuf: if you have no answers here, please send a mail to guile-user mailling list or artanis mailling list <OrangeShark>mbuf, amz3: artanis is under heavy development, I know I had some issues the last time I was trying to use it. <davexunit>wingo: for some reason I have an easier time unboxing f64s than u64s. I'm currently writing a merge sort to order sprites by their Y coordinate and there's allocation due to u64->scm and scm->u64 calls. <davexunit>I have an f32vector of sprite coordinates and I'm filling up a u64vector with indices into that coordinate vector. to reference the y coordinate, I need to do (f32vector-ref coordinates (+ (* (u64vector-ref indices i) 2) 1)) <davexunit>guile is boxing the u64 to do arithmetic then unboxing it for the f32vector-ref call, it seems. <davexunit>I guess either I'm doing something wrong that makes the compiler do this, or in order to eliminate allocation I need to split the f32vector of coordinate pairs into 2 f32vectors of x coords and y coords. <davexunit>hmm, seems that splitting up the vectors makes things more convenient for me, but not more efficient. the scm->u64 and u64->scm ops remain <davexunit>,x (lambda () (f32vector-ref (f32vector 1.0 2.0 3.0) (u64vector-ref (u64vector 0) 0))) <davexunit>is this an optimization to be added or user error? <janneke>mescc-compiled mes now loads modules and passes 30 scheme tests <janneke>now stuck on: expected: (#\\a #\\b #\\c #\\n unbound variable: ewline <lloda>davexunit: it also looks like the compiler pevals vector/vector-ref but not u64vector/u64vector-ref (if you look at ,x with u64vector -> vector) <davexunit>yeah, I'm not concerned about that right now, but yeah. <davexunit>if it did I would need a bigger example to demonstrate my problem <civodul>janneke: oh so you're about to close the loop! <janneke>civodul: ...almost...almost self-hosting <taylan>did anyone raise the issue of some syntax-rules macros desiring unhygienic (by-name) matching of literals yet, since 2.2 is now released and matches them all hygienically (as is conformant BTW)? <taylan>e.g. guix has the 'modify-phases' macro which uses literals like 'add' and 'remove' which are really intended to be matched by name not variable-identity. <taylan>I suspect it'll bite a few people in the back eventually if the issue is ignored :) <civodul>i just rediscovered it and found it impressive <dsmith-work>wingo: Did I hear something about userland code being more performant than kernel code from you? <dsmith-work>The last few devices we have done here at work use Python. A lot. I really hate it. <dsmith-work>But I hate that I can't just look at the code and see the data structures. Any random funtion can modify any random object at runtime. <dsmith-work>How does lua compare? (probably the wrong chan for this question) <dsmith-work>Sure wish we had something more staticaly and strongly typed.