octorules-wirefilter 0.4.1

Wirefilter expression parser FFI bindings for octorules

octorules-wirefilter

Rust FFI bindings for Cloudflare's wirefilter expression parser, exposed to Python via PyO3. When installed, octorules lint uses the real wirefilter parser for authoritative expression analysis instead of the built-in regex fallback.

Installation

# Install octorules with wirefilter support
pip install octorules[wirefilter]

# Or install standalone
pip install octorules-wirefilter

How it works

octorules lint
    │
    ▼
expression_bridge.py          Python-side routing layer
    │
    ├─► octorules_wirefilter   (if installed)
    │       │
    │       ├── lib.rs         PyO3 parse_expression(expr, phase=None)
    │       ├── scheme.rs      Phase-aware field/function schemes
    │       └── visitor.rs     AST walker → fields, functions, operators, literals
    │       │
    │       ▼
    │   wirefilter-engine      Cloudflare's Rust expression parser
    │
    └─► regex fallback         Built-in patterns (always available)

octorules tries to import octorules_wirefilter at module load time. If available, expressions are parsed by the real Cloudflare wirefilter engine. On import failure or parse error, the bridge transparently falls back to regex extraction. Either path returns the same ExpressionInfo dataclass consumed by the linter.

Scheme

A single wirefilter scheme is built at startup and cached:

• 169 fields exposed via get_schema_info(), 36 functions.
• Named list support — expressions like ip.src in $my_list parse without error. AlwaysList is registered for Int, Ip, and Bytes types so any $name reference is accepted. Actual list validation (existence, type compatibility) is handled by the Python linter (CF102: unresolved list reference, CF104: field type incompatible with list kind).
• Wildcard limit — ParserSettings enforces a maximum of 10 * metacharacters per wildcard pattern to prevent catastrophic backtracking.

The phase parameter is accepted for API compatibility but currently unused — all expressions are parsed against the same scheme. Transform-phase function-call syntax (where http.request.uri.path is callable) is handled on the Python side.

Building from source

Prerequisites

• Rust toolchain >= 1.86, edition 2024 (stable, via rustup)
• Python >= 3.10 with venv
• maturin (pip install maturin)

Development build

maturin develop

Builds the Rust crate and installs the resulting Python extension module into the active virtualenv.

Wheel build

maturin build --release

Produces a wheel in target/wheels/.

Testing

# Install test dependencies
pip install pytest

# Run FFI tests (requires octorules-wirefilter to be installed via maturin develop)
pytest tests/

Tests skip gracefully if the native extension is not installed.

API

This package exposes two functions:

parse_expression(expr, phase=None)

from octorules_wirefilter import parse_expression

# Parse an expression against the default scheme
result = parse_expression('http.host eq "example.com"')
# {'fields': ['http.host'], 'operators': ['eq'], 'string_literals': ['example.com'], ...}

# Phase parameter is accepted for forward compatibility but currently unused —
# all expressions parse against the same scheme regardless of phase value.
result = parse_expression('lower(http.host) eq "test"', phase="url_rewrite_rules")

# Parse errors return an error key (all list keys present but empty)
result = parse_expression('bogus_field eq "x"')
# {'error': '...', 'fields': [], 'functions': [], 'operators': [], ...}

Returns a dict with keys fields, functions, operators, string_literals, regex_literals, ip_literals, int_literals, regex_field_pairs (all lists), plus:

• On success: lists populated with extracted values. regex_field_pairs holds [field, regex] pairs showing which field each regex is matched against (e.g. [["http.request.uri.path", "^/api"]]). If AST nesting exceeded the depth limit, depth_exceeded: true is included.
• On failure: error (string) with all list keys present but empty.

Expressions exceeding 1 MiB are rejected with an error dict before parsing. Nesting depth is capped at 100 levels to prevent stack overflow on pathological input.

get_schema_info()

from octorules_wirefilter import get_schema_info

info = get_schema_info()
# {'fields': [{'name': 'http.host', 'type': 'STRING'}, ...],
#  'functions': ['lower', 'upper', ...]}

Returns schema metadata for automated synchronization with the Python linter schemas. Field types use the Python FieldType enum names (STRING, INT, BOOL, IP, ARRAY_STRING, etc.).

Contributing

Important: Field and function registries exist in two places: src/scheme.rs (Rust — used by wirefilter for parsing and type checking) and octorules_cloudflare/linter/schemas/ in the octorules-cloudflare repo (Python — used by the regex fallback parser and lint rules). A pre-commit hook in octorules-cloudflare auto-regenerates schemas.json when overlay.toml or pyproject.toml is modified. Rust-side changes here must still be made manually.

Adding fields

When Cloudflare adds new fields, update src/scheme.rs — add the field to register_common_fields() and to the COMMON_FIELD_NAMES array.

Then in the octorules-cloudflare repo, run python scripts/sync_schemas.py to regenerate schemas.json. If the field needs Python-only metadata (requires_plan, is_response), add it to overlay.toml in that repo first.

Adding functions

Update src/scheme.rs — register in register_common_functions() and add the name to the COMMON_FUNCTION_NAMES array.

Then in the octorules-cloudflare repo, run python scripts/sync_schemas.py to regenerate schemas.json. If the function needs restricted_phases or requires_plan, add it to overlay.toml in that repo first.

Design decisions

• Separate PyPI package. The Rust build requires a toolchain and takes longer to compile. Users who want fast installs get pip install octorules; those who want authoritative parsing opt in with pip install octorules[wirefilter].
• Git dependency pinning. wirefilter-engine is pinned to a specific commit because the required APIs (SchemeBuilder, function registration) are not in the published crates.io version.
• Stub function implementations. Functions are registered with correct type signatures but no-op execution. Expressions parse and extract correctly; runtime evaluation is not supported.
• cdylib crate type. Required by PyO3's extension-module feature for Python to load the native extension.

License

octorules-wirefilter is licensed under the Apache License 2.0.