ttr-parser

A pure-Python parser, walker and reference resolver for the TTR modeling language (@grammar-version 2.2). Parses .ttr models into a typed AST and resolves cross-references through the same six-step chain the platform uses — no JVM required at install or runtime (the ANTLR parser is generated and bundled into the wheel).

pip install ttr-parser

Requires CPython 3.13+.

Quickstart

from ttr_parser import parse_file, load_project
from ttr_parser.semantics import ResolutionContext

# Parse a single file into a typed AST.
result = parse_file("model.ttr")
for definition in result.definitions:
    print(definition.kind, definition.name)

# Load a whole project: parses every *.ttr under the root and pre-loads the
# stock CNC vocabulary, all into one symbol table.
project = load_project("path/to/project")

# Resolve a reference the way the platform does (lexical → same-package →
# imports → stock auto-import → fully-qualified).
result = project.resolve(
    "artikl", ResolutionContext(schema_code="er", namespace="entity")
)
print(result)  # Resolved(symbol=..., via_step="same-package") | Unresolved(...)

# Aggregated parse + resolution + validation diagnostics.
for diagnostic in project.diagnostics():
    print(diagnostic.code, diagnostic.message)

What's in the box

• Parser / walker (ttr_parser) — parse_string / parse_file / parse_directory return a typed, frozen AST faithful to grammar v2.2, with accurate source locations on every node.
• Semantics (ttr_parser.semantics) — SymbolTable, the six-step Resolver, the portable Validator subset, the stock CNC vocabulary, and the Project / load_project convenience entry point.

The AST and resolution output are pinned byte-for-byte to the reference TypeScript/Kotlin implementations by a conformance harness, so a Python consumer sees exactly what the platform sees.

License

Apache-2.0.