I'm impressed. Muse Spark 1.3, from few Doknil code examples, was able to extend my previous code with: new syntax; new AST; new unparsing rules for generating Doknil code from AST. tangled.org/dokmelody.bs...

Updated the syntax of Doknil according the new documentation/spec · dokmelody.bsky.social/scratch-02-jastadd@89632d9

Implement using JustAdd attribute grammar approach and JCR content repository

tangled.org

I added the specification of the #Doknil language: tangled.org/dokmelody.bs... Doknil is a simple knowledge-base language, for linking and navigating in all the parts of a project.

stages/stage-03/docs/DSL-specs/Doknil.md at c94392c6336808c1199f4308805a597f824a81a9 · dokmelody.bsky.social/scratch-02-jastadd

Implement using JustAdd attribute grammar approach and JCR content repository

tangled.org

I started a blog about the boostrapping of DokMelody.

The long journey towards Dok AST

Now I can parse Dok (a first version of it) and generate its Abstract Syntax Tree (Dok AST). commit/a43417024115fa586dd7b4be2a27f92b87a91c0b The followed approach is far from pragmatic, but I'm experimenting some paradigms that I would use also in the final project: I have the Dok grammar, written using the tree-sitter formalism. Tree-Sitter tools are used for generating a JSON description of the Dok grammar.

bootstrapping-dokmelody.pckt.blog

I added the generation of JastAdd AST of a language, from its tree-sitter grammar.

I switched from Aider/Cecli (i.e. AI pair programming) to omp.sh/docs/vibe that is also an AI vibe-coding tool. I started adding some "slop" AI code. The hope is to use Dok, in future, for having understandable AI generated code, because the code can be linked to docs and high-level views.

omp — a coding agent with the IDE wired in

Subagents, plan mode, LSP, DAP, hindsight memory, hashline edits, time-traveling rules — with a native Rust engine doing the heavy lifting.

omp.sh

I can convert a generic tree-sitter grammar into a JastAdd representation, using JastAdd attribute grammars equations. I can show: the JSON source text; the JSON graph; the resulting graph.

If I select a text, the corresponding AST node is highlighted, and vice-versa. This feature is inspired to CodeProber behavior codeprober.org . An AST tree can become very big, but this feature it is easy to see the represented part.

After trying pi.dev for AI coding harness, I decided to use the more conservative aider-chat (aider.chat): it is mainly an AI-pair-programming tool; it reads `AI!` annotations near the source code positions I need to change; it uses the tokens in a very efficient way, and many of them are cached.

Pi Coding Agent

A terminal-based coding agent

pi.dev

If I select a node on the tree, it shows the associated source text. Useful for debugging the grammar, because a tree can be rather complex, while the source text is linear.

I can draw (in a ugly way) a text parsed from JastAdd to a tree, managed from the ELK library (eclipse.dev/elk/). For the conversion I take advantage of the Java annotations generated by JastAdd, and added to JVM Classes.

Despite Tangled works out of the box with Jujutsu(www.jj-vcs.dev), I started using Sapling (sapling-scm.com), and I like it very much: - it can push and pull commits from normal Git repo - it has a very elegant and simple semantic Maybe in future I will study better Pijul, and its patch semantic.

The JastAdd formalism is more compact than Java classes. This is the AST of Dok tangled.org/dokmelody.bs...

src/main/jastadd/org/dokmelody/scratch2/ast/dok/dok.ast at main · dokmelody.bsky.social/scratch-02-jastadd

Implement using JustAdd attribute grammar approach and JCR content repository

tangled.org

I added the repo with the current implementation attempt: tangled.org/dokmelody.bs... I'm using these tools: - Tree-Sitter for the grammar - JastAdd Attribute Grammars for the semantic - Java Jackrabbit JCR for the storage of hierarchical documents - Java H2 for Doknil facts

dokmelody.bsky.social/scratch-02-jastadd

Implement using JustAdd attribute grammar approach and JCR content repository

tangled.org

I dediced to use the European Union Public Licence (EUPL) for DokMelody, for these reasons: dokmelody.tngl.io/scratch-02-d... #eupl

I published the temporary (scratch) documentation website. I'm using DITA, because it allows to navigate easily in related concepts of the documentation. dokmelody.tngl.io/scratch-02-d...

I put here an old implementation attempt, based on Common Lisp: tangled.org/dokmelody.bs... I decided to change the name from Dokmelody to DokMelody, because it makes more sense.

dokmelody.bsky.social/scratch-01-common-lisp

An archived attempt

tangled.org