@ocramz Spoke too soon! But it’s working now. CHOLMOD can be…subtle. Anyway, I think I’d like to write a cholmod-sla library, just an interface to CHOMOD using SLA matrix types for input and output. I think I’ll also dis-intermediate hcholmod since it’s not on hackage and I’ve had to add a bunch of FFI stuff anyway. I might as well do all of it and then it won’t rely on a non-hackage project. Not quite sure when I’ll get to splitting it all off and re-doing the hcholmod bits, but sometime in September is likely.
I see you have a permutation matrix constructor and a swap rows function. But I don’t see
permuteRowsSM :: [IxRow] -> SpMatrix a -> SpMatrix a
permuteColsSM :: [IxCol] -> SpMatrix a -> SpMatrix a
permuteSV :: [IxRow] -> SpVector a -> SpVector a
which I think will be useful. Once I get to doing all this right, should I put those in cholmod-sla or submit a PR to SLA?
I see you have a permutation matrix constructor and a swap rows function. But I don’t see
permuteRowsSM :: [IxRow] -> SpMatrix a -> SpMatrix a
permuteColsSM :: [IxCol] -> SpMatrix a -> SpMatrix a
permuteSV :: [IxRow] -> SpVector a -> SpVector a
which I think will be useful. Once I get to doing all this right, should I put those in cholmod-sla or submit a PR to SLA?
While
stack build
I installed it using
stack build
then created another file and imported it, showed me an error which stated that I need to add it to package.yaml
even after adding it we get the above error
The error seems to be coming from hpack, which makes sense, since that is an invalid syntax for dependencies. It should be analyze == 0.1.0.0, or simply analyze, but for stack.yamlyou are expected to pin the exact version, so that would be correct: analyze-0.1.0.0
Error while parsing $.dependencies[1] - invalid dependency "analyze-0.1.0.0"
@shubhamkumar13 to expand on that: if you're a package developer, and are working in the project (say,
analyze) directory, you'd run stack build to check your work. If you are a package user, and you only want to import a library into your project (provided the library is on Hackage already), you'd first add it in the build-depends section of the .cabal file
(btw hi @lehins , good to have you here! :thumbsup: )
btw @/all I've set up a GitLab clone of the github organization : https://gitlab.com/datahaskell . I'd like to have it as an alternative git remote
@ocramz, is Gitlab and Github now going to be automatically in a two-way? So any commits to master in either of them is going to be reflected in the other automatically (presumably in, say a short time)? Or Gitlab is just a read-only mirror of Github, and Github still remains the main development system?
*two-way sync...?
but I haven't tried so myself. I only know it is possible to have multiple git remotes
https://www.cl.cam.ac.uk/~nk480/numlin.pdf
This paper seems cool is something like this possible in haskell?
This paper seems cool is something like this possible in haskell?
@ocramz you have to either do it manually or probably with a git post-hook. you can add many remotes with "git remote add ....", but each branch is configured to follow one remote by default (i.e. if you do "git push"). If you want to push to everything you would have to "git push origin master" "git push otherRemote master" "...."
It could be possible to automate that with git-hooks (either on client or server-side) - but they can cause other problems in case remotes have diverged.. you get all of the nice race-conditions if 2 people happen do push at the same time .. :)
It could be possible to automate that with git-hooks (either on client or server-side) - but they can cause other problems in case remotes have diverged.. you get all of the nice race-conditions if 2 people happen do push at the same time .. :)
http://www.ofb.net/~frederik/stla/draft-r2.pdf - similar things were already done in 2006 but did not catch on..
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.296.7937&rep=rep1&type=pdf - that is more to what i think is a killer-feature of haskell.. Automagic streaming/recycling & synthesizing an efficient runtime-kernel for a specific problem.
(but i am a fan of Roman Leshchinskiy anyway ;) )
money quote from that paper: "Clearly “properly”-written C++ can outperform Haskell. The
challenge is in figuring out what “proper” means. A Haskell programmer can write straightforward, declarative code and expect the
compiler to handle fusion. The C++ programmer must worry about
the performance implications of abstraction."
challenge is in figuring out what “proper” means. A Haskell programmer can write straightforward, declarative code and expect the
compiler to handle fusion. The C++ programmer must worry about
the performance implications of abstraction."
but you can then have bad surprises that you change some code, some optimization stops and you get a 10x+ slowdown ...
to prevent that one can use inspection testing: https://arxiv.org/abs/1803.07130 (paper) / https://www.youtube.com/watch?v=h0pCKIFP_iQ (30 min youtube-presentation)
but it is still easy to mess up performance with little changes.. but GHC is evolving and it is better than some years ago :)
If you know anyone who wants to learn Haskell but is only willing to spend less than 10 seconds configuring a working Haskell environment, then please send them this link to the Jupyter adaptation of Learn You a Haskell for Great Good! https://mybinder.org/v2/gh/jamesdbrock/learn-you-a-haskell-notebook/master?urlpath=lab/tree/learn_you_a_haskell/00-preface.ipynb
@acowley and other vinyl people -- just thought you would appreciate hearing, i'm planning on releasing a library soon that "generalizes" vinyl in a way to other foldables besides lists. It doesn't really intend to replace or substitute vinyl in any way, but it heavily depends on it and it's been something i've found useful :)
let me know if you have any reservations about me presenting the library in this way