Sender | Message | Time |
---|---|---|
29 Dec 2023 | ||
boltless | making word a single token without children solve the whole problems | 18:27:36 |
.vhyrro | does that mean that, theoretaically, we can the parser even simpler than it is now? | 18:27:47 |
.vhyrro | because I can imagine your current implementation naturally has some "hacks" to make it all work, but maybe we can now have some wiggle room to reduce the parser size | 18:28:15 |
.vhyrro | that's pretty exciting | 18:28:17 |
boltless | I don’t think so because parser size with/without _character isn’t that different | 18:29:13 |
.vhyrro | I meant perhaps there's a configuration of nodes now which allow the parser to still do what we want while being more pedantic and therefore making treesitter consider less possibilities | 18:29:57 |
.vhyrro | but even with how it is now i'm amazed that we got it to work | 18:30:04 |
.vhyrro | good stuff man | 18:30:06 |
boltless | when I successfully put all paragraph syntax inside anchor, we would be good to go | 18:34:14 |
boltless | …and I found a new problem: compiling anchor that can include anchor inside takes forever | 18:44:22 |
boltless | anchor can have bold inside & bold can have anchor inside = anchor can have anchor inside. making tree-sitter explode while compiling parser | 18:46:13 |
tomcek8 | Download image.png | 21:20:17 |
tomcek8 | how do i fix this? running as administrator doesnt work | 21:20:25 |
30 Dec 2023 | ||
boltless | have you tried running that command manually? btw wrong channel. that error seems to be a windows permission problem, not a tree-sitter parser issue. | 02:52:51 |
boltless | v3 parser is near limit that tree-sitter can handle removing one of non-verbatim attached modifier allows implementing anchor | 03:12:47 |
boltless | .vhyrro can we remove subscript and implement as superscript’s variant via attached modifier extension? or remove both and make new syntax for subscript/superscript anchor works fine with up to 6 non-verbatim attached modifiers. from 7, parser compile takes forever (verbatim is fine, removing verbatim doesn’t help) extra reasons: - there are too many attached modifiers. I initially loved this consistency, but it leads making even , as attached modifier opener that user should use carefully- we can prevent nested subscript inside superscript from syntax level - seriously, who want’s to make subscript/superscript bold or italic? - current superscript syntax conflicts with footnote links (both rendered as superscript) - generally superscript/subscript follows right after the word. but as these are attached modifier in Norg, user needs to put link modifier every time | 03:50:15 |
boltless | + removing one of non-verbatim attached modifier reduces parser size to literally half (1.6MiB) | 04:19:05 |
asynthe joined the room. | 04:55:51 | |
boltless | current paragraph grammar seems to be too complex for tree-sitter to handle that I can’t add more markup in paragraph. I should remove two of non-verbatim attached modifiers or linkables to successfully compile without breaking the whole parser some solutions I’m thinking right now: 1. remove subscript & markups inside link destination 2. remove both subscript and superscript 3. find the buggy part from tree-sitter’s generate command and fix itbtw attached modifier extensions are working fine without any problems | 05:36:24 |
boltless | I’m having this problem again with linkables and as grammar.js is already quite simple, I can’t figure out a way to reduce the complexity of grammar except removing some markup syntaxes | 05:38:14 |
boltless | * current paragraph grammar seems to be too complex for tree-sitter to handle that I can’t add more markup in paragraph. I should remove two of non-verbatim attached modifiers or linkables to successfully compile without breaking the whole parser some solutions I’m thinking right now: 1. remove subscript & all markups inside link destination 2. remove both subscript and superscript 3. find the buggy part from tree-sitter’s generate command and fix itbtw attached modifier extensions are working fine without any problems | 05:56:55 |
boltless | * and as grammar.js is already quite simple, I can’t figure out a way to reduce the complexity of grammar except removing some markup syntaxes partially solved. both link & anchor can exist with up to 6 non-verbatim attached modifier (so I should remove only one to successfully compile the parser) | 06:11:01 |
boltless | * I should remove two of non-verbatim attached modifiers or linkables to successfully compile without breaking the whole parser some solutions I’m thinking right now: 1. remove subscript & all markups inside link destination 2. remove both subscript and superscript 3. find the buggy part from tree-sitter’s generate command and fix itbtw attached modifier extensions are working fine without any problems | 06:11:13 |
boltless | summary: - [x] attached modifiers - [x] verbatim attached modifiers - [x] free-form - [x] verbatim free-form - [x] link modifier - [ ] linkables (requires one less non-verbatim attached modifier) - [ ] attached modifier extensions (requires one less non-verbatim attached modifier) | 09:22:10 |
max397 | is this all the syntax, | 09:25:26 |
max397 | * is this all the syntax? | 09:25:30 |
max397 | so we could be finished if we‘d remove one non verbatim attached modifier | 09:25:49 |
boltless | we should remove two of them | 09:28:37 |
boltless | then yes, it’s finished | 09:28:47 |
boltless | but now I’m afraid to add block types in this parser because I have no idea how the bug on parser generation happens, same thing may happen on block types | 09:29:57 |