Newest at the top
| 2026-06-15 22:26:53 +0000 | __monty__ | (~toonn@user/toonn) (Quit: leaving) |
| 2026-06-15 22:25:37 +0000 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 272 seconds) |
| 2026-06-15 22:22:49 +0000 | <EvanR> | smh |
| 2026-06-15 22:22:37 +0000 | <EvanR> | even well behaved float math ends up wrong |
| 2026-06-15 22:22:04 +0000 | Sgeo | (~Sgeo@user/sgeo) Sgeo |
| 2026-06-15 22:21:54 +0000 | <lambdabot> | False |
| 2026-06-15 22:21:52 +0000 | <probie> | > 2.0^53 + 1 + 1 == 2.0^53 + 2 |
| 2026-06-15 22:21:37 +0000 | <probie> | The downside is that if you add two half bits, you don't get a bit, because no information + no information doesn't gives you information, but that's nothing new |
| 2026-06-15 22:20:51 +0000 | <EvanR> | a byte has 256 values so a half byte has 128 values? |
| 2026-06-15 22:20:31 +0000 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-06-15 22:20:08 +0000 | <probie> | A bit has two values, so half a bit only has one value, and there for contains no information, so (9/2) bits = 4.5 bits = 4 bits |
| 2026-06-15 22:19:29 +0000 | <EvanR> | o_O |
| 2026-06-15 22:19:20 +0000 | <probie> | 4 or 3 bits |
| 2026-06-15 22:19:01 +0000 | <EvanR> | what's a half-byte when byte = 9bits or 7bits |
| 2026-06-15 22:18:29 +0000 | <probie> | How about a compromise. We can call it a half-bite and then absolutely everyone is equally unhappy |
| 2026-06-15 22:17:45 +0000 | <int-e> | Well Wikipedia is wrong here (authority: mine) ;-) |
| 2026-06-15 22:17:07 +0000 | <EvanR> | must use upsilon to make it clear what we're doing |
| 2026-06-15 22:16:45 +0000 | <EvanR> | crumb -> crυmb |
| 2026-06-15 22:16:15 +0000 | <probie> | I was taught it as "nibble", and wikipedia also seems to prefer nibble over nybble https://en.wikipedia.org/wiki/Nibble |
| 2026-06-15 22:16:05 +0000 | <geekosaur> | right, that |
| 2026-06-15 22:16:03 +0000 | <int-e> | that's what I've been saying |
| 2026-06-15 22:15:52 +0000 | <geekosaur> | isn't it "nybble" anyway?) |
| 2026-06-15 22:14:52 +0000 | <int-e> | (I'll always read "nibble" as the verb first. Sure, I can auto-correct but it's an extra step :P) |
| 2026-06-15 22:14:02 +0000 | <EvanR> | y for the short i sound, egregious |
| 2026-06-15 22:13:31 +0000 | <int-e> | nybble : nibble :: byte : bite |
| 2026-06-15 22:12:59 +0000 | <EvanR> | array of nibbles at least won't require any real parsing using 8bit bytes |
| 2026-06-15 22:12:56 +0000 | <probie> | The need for Storable is because I'm interacting with foreign code, but I think from the Haskell side, just treating it as a Vector Word8 is what I'm going to do |
| 2026-06-15 22:12:53 +0000 | <geekosaur> | (the PDP11/VAX11 used bytes, but various others used 9, 18, 36, etc. bits. PDP10s' CPUs had 3-bit sub-register addressing, even) |
| 2026-06-15 22:11:58 +0000 | <int-e> | probie: Though the conclusion would still be that `Storable` isn't the right abstraction. Working with Vector Word8 and having your own adapter around it is compatible with that claim. |
| 2026-06-15 22:11:51 +0000 | <geekosaur> | well, these days it's not like it's relevant unless someone has an old DEC mini or mainframe in storage or something |
| 2026-06-15 22:11:13 +0000 | <EvanR> | .oO( how would you distinguish bytes and octets like that in french ... ) |
| 2026-06-15 22:10:25 +0000 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 272 seconds) |
| 2026-06-15 22:10:06 +0000 | <geekosaur> | anything that isn't based on the fundamental unit for the storage medium (bytes for memory on pretty much everything these days, octets for network streams and disks, etc.) requires extra work |
| 2026-06-15 22:09:51 +0000 | CiaoSen | (~Jura@dynamic-046-114-167-087.46.114.pool.telefonica.de) CiaoSen |
| 2026-06-15 22:09:39 +0000 | machinedgod | (~machinedg@d172-219-48-230.abhsia.telus.net) machinedgod |
| 2026-06-15 22:09:19 +0000 | <int-e> | I can't read. I skipped the "Data.Vector" part :-/ |
| 2026-06-15 22:09:03 +0000 | <int-e> | Ohh |
| 2026-06-15 22:08:38 +0000 | <int-e> | probie: Ah, so you're not doing this for marshalling then. |
| 2026-06-15 22:08:34 +0000 | <geekosaur> | which is the fundamental problem |
| 2026-06-15 22:07:37 +0000 | CiaoSen | (~Jura@2a02:3035:eef:857d:4e50:ddff:fe9b:8922) (Ping timeout: 248 seconds) |
| 2026-06-15 22:07:34 +0000 | <geekosaur> | the streams they operate on are octet based, so "yes". anything else must be done by, essentially, parsing |
| 2026-06-15 22:07:14 +0000 | <probie> | int-e: That actually gave me a genius idea; I can have a Vector Word8, and just abuse pattern synonyms |
| 2026-06-15 22:05:09 +0000 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-06-15 22:04:56 +0000 | <int-e> | geekosaur: aren't those still byte based though? |
| 2026-06-15 22:04:40 +0000 | Pozyomka | (~pyon@user/pyon) (Quit: brb) |
| 2026-06-15 22:03:47 +0000 | vms14 | (~vms14@user/vms14) vms14 |
| 2026-06-15 22:03:33 +0000 | <geekosaur> | I think you're supposed to use binary/cereal or attoparsec for this kind of thing |
| 2026-06-15 21:59:36 +0000 | <int-e> | s/Storable types/Storable instances/ |
| 2026-06-15 21:59:21 +0000 | <int-e> | probie: Oh I guess you can do something awful (from a usability perspective): Have HighNybble and LowNybble types with Storable types that only access the low/high 4 bits of a byte. |
| 2026-06-15 21:55:40 +0000 | <lambdabot> | GHC.Internal.Ptr.Ptr a -> GHC.Internal.Ptr.Ptr b |