2026-03-06 10:45:12 <veltas      > Who even farkin knows
2026-03-06 11:59:35 <veltas      > Anyone here used skip lists before?
2026-03-06 12:14:19 <veltas      > Or AA-trees for that matter
2026-03-06 12:14:33 <veltas      > They're relevant to Forth believe it or not, both are basically low code dictionaries
2026-03-06 12:14:47 <veltas      > Less code than red-black trees etc
2026-03-06 12:15:05 <tpbsd       > not me but I did check Mecrisp-Stellaris to see if the dissasembly of a variable varied depending on whether something is stored to the variable or not, and it doesnt
2026-03-06 12:18:46 <veltas      > Likewise in gforth
2026-03-06 12:21:06 <tpbsd       > I do encounter interesting concepts here :)
2026-03-06 12:22:10 <tpbsd       > I always have a STm32F051 running Forth thermometer code here connected to this pc, so it's easy to do any tests I want
2026-03-06 12:22:38 <tpbsd       > 27,0 C  80,7 F
2026-03-06 12:42:29 <cleobuline  > forthBot:  LOAD ini.fth
2026-03-06 12:42:29 <forthBot    > File ini.fth with MOON  loaded
2026-03-06 12:42:40 <cleobuline  > forthBot:  CREDIT
2026-03-06 12:42:40 <forthBot    > Brought to you by Cleobuline updated with hashtable with the help of Grok  https://github.com/cleobuline/forth-bot-gmp-irc-threaded-multi-users/tree/main Site https://labynet.fr
2026-03-06 12:42:52 <cleobuline  > updated ....
2026-03-06 12:45:13 <cleobuline  > forthBot use a hashtable for the dictionnary
2026-03-06 12:45:50 <cleobuline  > at a mement i was plannig to use red black tree
2026-03-06 12:46:13 <cleobuline  > but it cause problem with forget
2026-03-06 12:48:23 <cleobuline  > forthBot:  HELLO
2026-03-06 12:48:23 <forthBot    > Hello cleobuline How are you ?
2026-03-06 12:53:54 <cleobuline  > forget , delete all  the words from a given word to the end of the dictionnary
2026-03-06 12:54:12 <cleobuline  > not possible with a red black tree
2026-03-06 12:54:44 <cleobuline  > unless we use a time stamping
2026-03-06 14:06:27 <veltas      > I don't understand how a hashtable is better for that?
2026-03-06 14:07:10 <veltas      > gforth uses a linked list for neighbour traversal and a hash table for lookup.  You could do same with RB tree instead of hash table surely?
2026-03-06 14:37:49 <cleobuline  > speedest veltas
2026-03-06 14:39:23 <cleobuline  > forthBot use a hashtable for lookup the dictionnary
2026-03-06 14:39:53 <veltas      > But what is the "cause problem with forget"?
2026-03-06 14:40:35 <cleobuline  > veltas:  forget delete entry in the dictionnary from a given word to the end of the dictionnary
2026-03-06 14:41:10 <veltas      > The data to track that is in the entry regardless though right?  The lookup data structure should be irrelevant?
2026-03-06 14:41:21 <cleobuline  > with rb tree you must delete word by word , or add a timestamp tu check the order
2026-03-06 14:43:16 <veltas      > The same would be true of hash table ... unless you've put a list or other mechanism in entries to help you delete older ones
2026-03-06 14:44:09 <veltas      > newer ones rather
2026-03-06 14:44:25 <veltas      > So I'm saying whether it's an RB tree or a hash table is irrelevant, neither is suited to the dictionary's linked-list based forget mechanism, but both can support it with identical workarounds
2026-03-06 14:44:36 <veltas      > Or am I missing something
2026-03-06 14:46:26 <cleobuline  > veltas:  the words are store in a table , but the hashtable is different table
2026-03-06 14:47:06 <cleobuline  > so  i have just to scan the words table and call removeWordFromHash(env, dict_word->name);
2026-03-06 14:47:58 <cleobuline  > with rb tree i do not have the order in wich the words are created
2026-03-06 14:48:23 <forthBot    > Environment for cleobuline inactive, freeing...
2026-03-06 14:48:24 <cleobuline  > unless i had a timestamp with the word created
2026-03-06 14:48:44 <cleobuline  > wich is more complex
2026-03-06 14:48:56 <cleobuline  > and add memory slot
2026-03-06 14:49:51 <cleobuline  > but it can be done
2026-03-06 14:49:58 <veltas      > So the main table the words are stored in would still exist if you used RB tree in place of the hashtable
2026-03-06 14:50:43 <cleobuline  > yes it's a solution , i guess the word where store in the rb tree
2026-03-06 14:51:15 <veltas      > In fact RB tree could be stored totally in-line, so no separate table is needed at all, you can put all words in one table and the tree is within that
2026-03-06 14:51:36 <cleobuline  > yes you are right
2026-03-06 14:52:56 <cleobuline  > i can build the rb tree separated from the word table right
2026-03-06 14:54:41 <cleobuline  > for forthBot i use uthash.h wich os very simple to use
2026-03-06 14:54:50 <cleobuline  > is
2026-03-06 14:56:12 <cleobuline  > i win 50 % performance with the response of the bot
2026-03-06 14:56:26 <veltas      > I expect a hash to be faster on average yeah
2026-03-06 14:56:37 <cleobuline  > yes it will
2026-03-06 14:57:10 <cleobuline  > but use rb tree is good for the sport :)
2026-03-06 14:57:30 <veltas      > But hash tables are limited in use, they don't order the content and are vulnerable to collision attacks
2026-03-06 14:57:41 <cleobuline  > yes
2026-03-06 14:58:00 <veltas      > Self-balancing trees are 'general' solutions so if I'm only going to add one kind of dictionary I'd go for a simple self-balancing tree
2026-03-06 14:58:02 <cleobuline  > collision a rare
2026-03-06 14:58:10 <veltas      > Problem is they aren't simple
2026-03-06 14:58:30 <veltas      > It doesn't really matter for Forth dictionary lookup
2026-03-06 14:58:31 <cleobuline  > i like rb tree :)
2026-03-06 14:59:07 <veltas      > But e.g. in a website you'd need to make sure your hash was safe and maybe use salt, or devolve to self-balanced trees if using closed addressing
2026-03-06 14:59:33 <veltas      > If users can predict what the hash is or whatever, there are all sorts of security considerations
2026-03-06 14:59:40 <veltas      > Just easier to use a tree in security context
2026-03-06 15:00:11 <cleobuline  > yes
2026-03-06 15:01:41 <cleobuline  > https://github.com/cleobuline/some-c-sources/blob/main/red-black-tree.c
2026-03-06 15:01:50 <cleobuline  > i done a test
2026-03-06 15:02:26 <cleobuline  > it is elegant solution rb tree
2026-03-06 15:02:57 <cleobuline  > but a dictionnary wil not contains billions of words !
2026-03-06 15:05:48 <cleobuline  > rb tree is adapted to a large amount of data i think
2026-03-06 15:07:18 <cleobuline  > my next implemantation will use a rb tree i think
2026-03-06 15:08:02 <cleobuline  > i will do it for the sport
2026-03-06 15:09:33 <veltas      > Maybe look up AA-tree though
2026-03-06 15:09:38 <veltas      > That's what I mentioned earlier
2026-03-06 15:10:04 <veltas      > That's what I will implement anyway, although it's not much sport because I'll be basing it on the Pascal code they wrote in their paper
2026-03-06 15:10:17 <veltas      > But it seems more Forthy because it's shorter
2026-03-06 15:10:52 <veltas      > You might find skip lists interesting too
2026-03-06 15:12:53 <cleobuline  > for the dictionnary hashtable is more efficient wich is O
2026-03-06 15:13:33 <cleobuline  > the search is constant
2026-03-06 15:13:55 <veltas      > What hash are you using?
2026-03-06 15:15:11 <cleobuline  > uthash.h
2026-03-06 15:18:02 <veltas      > Default hash function for uthash?
2026-03-06 15:19:57 <veltas      > Jenkins hash which looks okay, quite a simple one
2026-03-06 15:21:41 <cleobuline  > inline
2026-03-06 15:27:35 <cleobuline  > veltas:  https://pastebin.com/FPfNXmTh
2026-03-06 15:28:46 <cleobuline  > juste include and use it
2026-03-06 15:29:40 <cleobuline  > it works for me :)
2026-03-06 15:33:10 <cleobuline  > veltas:  sample use https://pastebin.com/yFpZSbqD
2026-03-06 15:36:07 <veltas      > Yeah that will use Jenkins I think
2026-03-06 15:37:20 <cleobuline  > i'm okay with it with  forthBot
2026-03-06 15:41:05 <veltas      > I'm using an even simpler one, the one that used to be in PHP
2026-03-06 15:41:22 <veltas      > So it's not exactly popular but it is pretty simple and runs well on 8-bit as well as newer machines
2026-03-06 15:41:27 <veltas      > Jenkins would as well I think
2026-03-06 15:42:02 <cleobuline  > ok
2026-03-06 15:55:34 <veltas      > I think it was DJB2, multiply by 33 and XOR ... something like that
2026-03-06 16:02:03 <cleobuline  > https://x.com/rohanpaul_ai/status/2029679074868879852
2026-03-06 16:13:38 <veltas      > Looks like it was made for cdb and cdb is still being developed?
2026-03-06 19:18:11 <veltas      > cleobuline: X post; could be unrelated, or could be super bad news for when the bubble bursts
2026-03-06 19:24:21 <veltas      > SO PROUD of my boys at Apple, proud shareholder right here
2026-03-06 20:08:53 <cleobuline  > release of forthBot for macos https://github.com/cleobuline/forth-bot-gmp-irc-threaded-multi-users/releases/tag/forthbot_macos
2026-03-06 20:09:25 <cleobuline  > tested and approuved :)
2026-03-06 20:16:02 <veltas      > MacOS is a tier 0 platform now
2026-03-06 20:16:18 <veltas      > Move aside MS
2026-03-06 22:12:11 <cleobuline  > forthBot:  S" des tripodes martiens attaquant Paris, style gravure victorienne"S IMAGE
2026-03-06 22:12:55 <forthBot    > https://i.ibb.co/prfXN6sf/zozo-img.png
2026-03-06 22:13:35 <cleobuline  > veltas:  macos is a goog platform
2026-03-06 22:14:23 <cleobuline  > good
2026-03-06 22:22:19 <cleobuline  > veltas:  what you don't like in macos veltas ?
2026-03-06 22:23:14 <cleobuline  > i'm using macos to edit files on  my vps ubuntu it's very sweet to use
2026-03-06 22:25:57 <cleobuline  > i'm confortable with it
2026-03-06 23:59:51 <cleobuline  > forthBot:  LOAD ini.fth
2026-03-06 23:59:52 <forthBot    > File ini.fth with MOON  loaded
2026-03-06 23:59:57 <cleobuline  > forthBot:  fee