Olle Fredriksson

@typeswitch Branch displacement optimization, deciding whether a relative jump is short enough to use the smaller instruction encoding, is also messy fun.

@typeswitch I ended up going through a monoidal description record in my assembler because of the damn prefixes. Too much back and forth to go directly to bytes cleanly.

@GabriellaG439 Have you seen hackage.haskell.org/package/Earley…?

I feel like if you have a parser for a type you should be able to derive a generator (e.g. an Arbitrary instance) for the same type for free

@luctielen @OlleF That looks very clean. Nicely done. Btw, you got the wrong Olle on here. :)

Eclair now uses the `rock` build system made by @ollef. Really nice to setup, cleaned up the code a lot! This also opens up the possibility to compile multiple Datalog files! The diff: github.com/luc-tielen/ecl…

@brendanzab @kamatsu8 @wenkokke @cyberglot @jonmsterling Assuming that the implementations and formalisations are distinct... :)

@brendanzab @kamatsu8 @wenkokke @cyberglot @jonmsterling Yes: Sixty does NbE with De Bruijn indices in the syntax and fresh vars in the domain to avoid shifting during pattern matching elaboration, which can make modifications to the middle of the context. But I guess implementations might have different tradeoffs than formalisations.

@julepaj samma fast "melon"

Att slutet på ”citron” uttalas som slutet på ”päron” är självklart, men uttalar ni inledningen med långt eller kort i?

@brendanzab @andrasKovacs6 @jonmsterling It doesn't evaluate further.

@brendanzab @andrasKovacs6 @jonmsterling There it is! Also note that the sentinel Builtin.fail unifies with anything without error in an attempt not to get cascades of errors due to errors.

@brendanzab @andrasKovacs6 @jonmsterling No, just for better inference. Errors in elaboration result in sentinel error terms to be able to keep going.

@brendanzab @andrasKovacs6 @jonmsterling Sixty does postponing!

@jaredweakly @Iceland_jack @JulienTruffaut Careful, those two are not the same.

How do you call these operators? *> and <*

It's a bit crude now, but if you want the fastest parser library in Haskell, which is a) on Hackage b) can be used for real-world programming language parsing with reasonable effort, see this: hackage.haskell.org/package/flatpa…

@tomjaguarpaw That's funny. SQL LIMIT and OFFSET was exactly the context that I found this in: #L120-L135" target="_blank" rel="nofollow noopener">github.com/folq/postgres-…

@ollfredo Especially these two lines #diff-5c499ee77e92cae7e2ec922a422e8607cf13e573191e018f3c4d85942368d913R47-R48" target="_blank" rel="nofollow noopener">github.com/tomjaguarpaw/h…

Has anyone seen this monoid with a cute action on lists before? empty = (∞, 0) (t₁, d₁) ⋄ (t₂, d₂) = (min t₁ (t₂ - d₁), d₁ + d₂) act (t, d) = take t . drop d

act (t₁, d₁) . act (t₂, d₂) = take t₁ . drop d₁ . take t₂ . drop d₂ = take t₁ . take (t₂ - d₁) . drop d₁ . drop d₂ = take (min t₁ (t₂ - d₁)) . drop (d₁ + d₂) = act (min t₁ (t₂ - d₁), d₁ + d₂) = act ((t₁, d₁) ⋄ (t₂, d₂))

London Perf Summit #2 is almost here ! Meet Olle Fredriksson (@ollfredo) as he presents how to build real-time tooling and incremental compilers using a build system library called Rock #londonperfsummit Sign-up here: bit.ly/londonperfsumit youtu.be/NSBfJC-pvmI

@rob_rix Lists seem quite hard to beat for performance at small sizes. I have some benchmarks here: github.com/ollef/environm…

Why are contexts lists? Why not binary trees? I like the idea of pushing an entire recursive pattern onto the context as a single (recursive) entry.

@runarfu Tror syrran är på jakt efter ett, så kanske. Mer detaljer?

Angrer mange musikkinstrumentanskaffelser. Nå er det trekkspillet som skal ut. Noen som er interessert i et rimelig pianotrekkspill med 96 bassknotter?

My best and most important work