fastparse 0.1.0rc20

Python binding for the FastParse Tree-sitter parser library with bundled native runtime

FastParse Python Binding

Small ctypes binding for the FastParse native C library.

The binding keeps the same design boundary as the native library:

• Python owns file I/O.
• Python passes bytes already loaded in RAM.
• FastParse returns JSON, CSV, MessagePack binary, diagnostics, or stats in RAM.
• Python copies the returned native buffer and frees the native result.

Install

From PyPI, once the package is published:

pip install --pre fastparse

From a downloaded wheel:

pip install fastparse-0.1.0rc17-py3-none-macosx_11_0_arm64.whl

The wheel includes the native library for its platform and py.typed markers for type-aware tooling, so normal users do not need FASTPARSE_LIBRARY_PATH.

Quick Use

from fastparse import FastParse, Field, OutputFormat, ParseOptions

parser = FastParse()
source = b"class Demo { void run() {} }"

result = parser.parse_bytes(
    source,
    ParseOptions(
        language="java",
        output_format=OutputFormat.JSON,
        fields=Field.RULE | Field.TEXT | Field.BYTE_RANGE,
    ),
)

print(result.node_count)
print(result.json())

Binary MessagePack output:

result = parser.parse_bytes(
    source,
    language="java",
    output_format="binary",
    fields=["rule", "text", "byte_range"],
)

print(result.node_count)
print(result.data)  # MessagePack bytes

Decode binary output into Python dataclasses:

document = result.binary_document()
print(document.nodes[0].rule)
print(document.nodes[0].text)

If the caller only needs node/output counts and does not need to copy the generated JSON/CSV into Python:

summary = parser.parse_bytes_summary(
    source,
    language="java",
    output_format="json",
)

print(summary.node_count)
print(summary.output_length)

Use FASTPARSE_LIBRARY_PATH to point at a specific native library:

FASTPARSE_LIBRARY_PATH=/path/to/libfastparse.dylib python your_script.py

Examples

From the repository root:

python3 examples/python/03_binary_decode/binary_decode.py
python3 examples/python/04_inventory_to_sqlite/inventory_to_sqlite.py --workers 12
python3 examples/python/05_diagnostics_scan/diagnostics_scan.py /path/to/java/root --glob "*.java" --workers 12

04_inventory_to_sqlite is the enterprise pattern for high-throughput work: the parent Python app reads files, uses threads, decodes binary output into dataclasses, and writes SQLite tables outside the native library.