search

agrafix / superrecord

Haskell: Supercharged anonymous records
BSD 3-Clause "New" or "Revised" License
83 stars 16 forks source link

Array# support #5

Open dredozubov opened 7 years ago

dredozubov commented 7 years ago

An exact number of fields is statically known, so there's no other reason that performance to choose SmallArray# over Array#. The issue with SmallArray# is an implicit maximum bound on the number of fields(<=128). Happily, it's possible to use multiple backends for the records of different size. Tests pass, but I want to add some new ones for the huge records specifically. Benchmarking won't hurt either.

This change is Reviewable

dredozubov commented 7 years ago

Also I should note it's a breaking change. This will force some new constraints on the user, but I consider it a lesser evil.

LeanderK commented 7 years ago

I have actually a use case for this. I am working on a DataFrames library and will probably choose superrrecord for as a backend for some parts. Having CSVs with > 128 columns is not super uncommon.

agrafix commented 7 years ago

We will want to support this, but at the moment there are some problems with larger records that cause massive compile time slow downs. We have to investigate those first before we can drill down on this.

LeanderK commented 7 years ago

okay, no rush. It's not even close to being finished.

dredozubov commented 6 years ago

I've created an issue to dig into the compilation time blow-up, in a sense it blocks this PR(but maybe it isn't?) https://github.com/agrafix/superrecord/issues/12