Expose imapSM, correct imapIM2, and change to alter

[GregorySchwartz]

imapSM is needed for many processes and is much better than flattening then building the matrix again to map over indices. I fixed the imapIM2 issue referenced in #74 and changed insert to alter for more speedup and better memory usage.

[GregorySchwartz]

Also strict foldl' is necessary for faster transposing.

[GregorySchwartz]

Can these changes be included in the next upload, along with the toRowsL / fromRowsL fix (which is already committed I believe)? I have some code relying on my fork where I would prefer to have it rely on the main Hackage release.

[GregorySchwartz]

Is it possible to get this merged and released relatively soon?

[ocramz]

👍 thanks you !

[ocramz]

Sorry for the slow feedback and thank you for looking into performance issues! 👍

[GregorySchwartz]

Thanks! Will there be a release on hackage?

[ocramz]

@GregorySchwartz yes, but first I'd like to bring back master to feature parity with the current latest hackage release. I commented out a few things (see https://github.com/ocramz/sparse-linear-algebra/blob/master/src/Numeric/LinearAlgebra/Sparse.hs) in order to get rid of the MonadIO constraint and while settling on a decent effect type (something like Iterative here : https://github.com/ocramz/sparse-linear-algebra/blob/master/src/Control/Iterative/Internal.hs ) . After this is done I indend to push out a new release, I just don't know when I will have the bandwidth to do this; could be a few more weeks.