Newest at the top
| 2024-05-01 04:48:03 +0200 | waldo | (~waldo@user/waldo) (Quit: waldo) |
| 2024-05-01 04:45:22 +0200 | peterbecich | (~Thunderbi@syn-047-229-123-186.res.spectrum.com) (Ping timeout: 268 seconds) |
| 2024-05-01 04:45:19 +0200 | terrorjack | (~terrorjac@2a01:4f8:c17:87f8::) |
| 2024-05-01 04:43:24 +0200 | terrorjack | (~terrorjac@2a01:4f8:c17:87f8::) (Quit: The Lounge - https://thelounge.chat) |
| 2024-05-01 04:43:21 +0200 | gues54302 | (~username@3.184.70.115.static.exetel.com.au) |
| 2024-05-01 04:42:35 +0200 | Square3 | (~Square4@user/square) (Ping timeout: 268 seconds) |
| 2024-05-01 04:29:44 +0200 | td_ | (~td@i53870927.versanet.de) |
| 2024-05-01 04:27:50 +0200 | td_ | (~td@i53870902.versanet.de) (Ping timeout: 245 seconds) |
| 2024-05-01 04:21:27 +0200 | cashew | (~cashewsta@65.17.175.150) |
| 2024-05-01 04:17:49 +0200 | Axman4765 | Bynbo7 |
| 2024-05-01 04:17:42 +0200 | Axman4765 | (~Axman6@user/axman6) |
| 2024-05-01 04:13:50 +0200 | Katarushisu1 | (~Katarushi@finc-20-b2-v4wan-169598-cust1799.vm7.cable.virginm.net) |
| 2024-05-01 04:12:09 +0200 | <c_wraith> | no, I mean that if a problem like this is hard, there are just a *lot* of distinct subproblems. |
| 2024-05-01 04:11:54 +0200 | <talismanick> | It's probably be none-too-expensive to preface the interpreter with the reduction rules, so long as the free monad/binary decision tree is built and consumed incrementally, right? |
| 2024-05-01 04:11:39 +0200 | Katarushisu1 | (~Katarushi@finc-20-b2-v4wan-169598-cust1799.vm7.cable.virginm.net) (Read error: Connection reset by peer) |
| 2024-05-01 04:09:58 +0200 | <talismanick> | The reduction of a Boolean function is canonical (up to variable ordering), so maybe it's not a problem in this case... |
| 2024-05-01 04:08:09 +0200 | <talismanick> | c_wraith: the cost of bookkeeping is prohibitive for free monads in general, you mean? |
| 2024-05-01 04:05:28 +0200 | ski | only heard someone mention Soufflé the other week, recognized the mention of BDDBDDB, but hasn't looked at it |
| 2024-05-01 04:02:45 +0200 | <c_wraith> | Ah, not quite. Patricia tries have a fixed fanout of 2 |
| 2024-05-01 04:01:17 +0200 | <c_wraith> | Which... sounds sorta like Patricia Tries, actually. |
| 2024-05-01 04:01:01 +0200 | <c_wraith> | B-trees are already lexicographically ordered. I'm betting it's more like adjusting the number of bits in a node in order keep the fan-out within particular bounds. |
| 2024-05-01 03:59:41 +0200 | <talismanick> | found it: https://souffle-lang.github.io/pdf/pmam19.pdf |
| 2024-05-01 03:59:28 +0200 | <talismanick> | ski: I think Souffle now uses a concurrent B-tree/trie hybrid (lexicographically-ordered B-tree?) |
| 2024-05-01 03:50:54 +0200 | wroathe | (~wroathe@user/wroathe) (Ping timeout: 252 seconds) |
| 2024-05-01 03:49:59 +0200 | <ski> | (not clear how would apply something like that to BDDs, though) |
| 2024-05-01 03:49:07 +0200 | <ski> | (what i was thinking about was encoding the sharing in the type, or rather, in the decomposition process, making it mandatory, thereby making sure it will happen in the result as well) |
| 2024-05-01 03:47:32 +0200 | <ski> | there's also a problem that when traversing a structure with sharing, producing a new parallel structure, that would normally lose all sharing |
| 2024-05-01 03:46:30 +0200 | ski | . o O ( "BDD-Based Deductive Databasee" <https://bddbddb.sourceforge.net/> ; "Soufflé - A Datalog Synthesis Tool for Static Analysis" <https://souffle-lang.github.io> ) |
| 2024-05-01 03:43:11 +0200 | waleee | (~waleee@h-176-10-144-38.NA.cust.bahnhof.se) (Quit: WeeChat 4.1.2) |
| 2024-05-01 03:38:18 +0200 | <c_wraith> | well, no. The problem is sharing only happens when explicitly introduced, and doing that is going to require bookkeeping. And the memory cost of that bookkeeping is ridiculous. |
| 2024-05-01 03:36:16 +0200 | <talismanick> | Is the problem is that laziness memoizes and shares for you without asking, making every tree equivalent to a DAG but not the one you want? |
| 2024-05-01 03:32:04 +0200 | <talismanick> | kind of like SAT solving |
| 2024-05-01 03:31:46 +0200 | <talismanick> | https://en.wikipedia.org/wiki/Binary_decision_diagram |
| 2024-05-01 03:31:22 +0200 | <ski> | s/re/are/ |
| 2024-05-01 03:31:19 +0200 | ski | doesn't recall what those re |
| 2024-05-01 03:31:09 +0200 | <talismanick> | yeah |
| 2024-05-01 03:30:28 +0200 | <ski> | hm, is "BDD" Binary Decision Diagram ? |
| 2024-05-01 03:29:18 +0200 | <talismanick> | is it the kind of thing where it becomes obvious when you write it down? |
| 2024-05-01 03:29:12 +0200 | otto_s | (~user@p4ff27c65.dip0.t-ipconnect.de) |
| 2024-05-01 03:29:11 +0200 | <ski> | or, in terms of composition, if `Compose (Compose ar0 ar1a) (Compose ar1b ar2) = ar012', then `ar1a = ar1b'. so we have something like `Compose :: (ar0 :: Slice n) -> (ar1 :: Slice n) -> Agree ar0 ar1 => Slice (1+n)', where `Agree (Singleton x) (Singleton y)' as well as `(ar1a = ar1b) => Agree (Compose ar0 ar1a) (Compose ar1b ar2)' (where `Agree :: Slice n -> Slice n -> Constraint') |
| 2024-05-01 03:27:48 +0200 | otto_s | (~user@p4ff27e40.dip0.t-ipconnect.de) (Ping timeout: 252 seconds) |
| 2024-05-01 03:26:03 +0200 | <talismanick> | because, if one can assume "maximal sharing", I feel like I can see in my head how a BDD might be implemented with a (real) free monad |
| 2024-05-01 03:24:25 +0200 | <ski> | basically, we have, if `Just (ar01,ar12) = decompose ar012', `Just (ar0,ar1a) = decompose ar01' and `Just (ar1b,ar2) = decompose ar12', then `ar1a = ar1b' |
| 2024-05-01 03:23:41 +0200 | <talismanick> | I'm feeling a little stupid, then, trying to understand https://hackage.haskell.org/package/zsdd/docs/Data-Diagram.html |
| 2024-05-01 03:18:35 +0200 | <ski> | advantage of the sharing |
| 2024-05-01 03:18:29 +0200 | <ski> | given an array/vector with indices from `0' to `n-1' (incl.), we could decompose this into one slice from `0' to `n-2' and one from `1' to `n-1', guaranteed to overlap on the `1' to `n-2' part. eventually, we'd get down to singleton slices, which could be the base case of a dynamic programming, then results percolating back up in a "lattive"-like structure, rather than a tree, proper, thereby taking |
| 2024-05-01 03:15:53 +0200 | waldo | (~waldo@user/waldo) |
| 2024-05-01 03:15:04 +0200 | <ski> | iow, i'm thinking of something like a catamorphism |
| 2024-05-01 03:15:01 +0200 | <c_wraith> | you can embed basically anything you want into f |
| 2024-05-01 03:14:50 +0200 | <c_wraith> | pick f carefully. |