dynamicsyntax 0.2.1

DyLan, the Dynamic Syntax parser in Python

dynamicsyntax

dynamicsyntax is the python implementation of the DyLan the Dynamic Syntax parser. It has been (vibe) translated from Java to Python by @incrementaliser using Cursor, and is still under development.

Installation

You need uv and a Python that satisfies >=3.11 (see pyproject.toml).

1. Create a virtual environment with a fixed Python version (run this in the directory where you want .venv, e.g. the repo root for a clone):

   uv venv --python 3.12
   

2. Activate it. On Linux or macOS:

   source .venv/bin/activate
   

   On Windows, use .venv\Scripts\activate (cmd) or .\.venv\Scripts\Activate.ps1 (PowerShell).

3. Install into that environment, using the published package from PyPI:

   uv pip install dynamicsyntax
   

   or editable install from a git clone:

   uv pip install -e .
   

Examples

dynamicsyntax.parse returns a ParseResult (with .semantics, .ok, .tree, .vis(), and .address_order). This replaces older versions that returned TTRRecordType | None directly—use .semantics for the final TTR record.

List bundled grammars and (placeholder) datasets:

import dynamicsyntax as ds

print(ds.get_grammars())   # e.g. ['2015-english-ttr', 'ttr', ...]
print(ds.get_datasets()) # [] until datasets ship with the package

One-shot parse with a grammar id or alias ("ttr" maps to 2015-english-ttr):

import dynamicsyntax as ds

p = ds.parse("a man arrives", "ttr")
print(p.ok, p.semantics)
p.vis()  # prints the same address-order tree view as the GUI

Load a grammar once, then parse without repeating the grammar argument:

import dynamicsyntax as ds

ds.load_grammar("ttr")  # or ds.load_grammar("2015-english-ttr")
p = ds.parse("a man arrives")
print(p.semantics)
p.vis()

Use a filesystem path to a grammar directory (as in the GUI “load folder” flow):

import dynamicsyntax as ds
from pathlib import Path

ds.load_grammar(Path("/path/to/grammar-dir"))
p = ds.parse("go to the red box")

Testing

See testing.md for how to run pytest with uv, show per-test pass/fail output, and useful options (-v, --lf, subsets, and CI).

Get Involved

Contributions (issues, PRs, donations) are very welcome! Since this project has been translated using Cursor and tested only on Windows, there is a small chance errors exist. I would be grateful if you could report them to me through the above channels. Although, it is important to mention that some of the most critical methods here have been verified via unit tests, so there is a high chance the migration from Java has been corrrect so far.