A reasonably fast DAG-CBOR parser for Python
Other than speed, a distinguishing feature is that it parses non-recursively. This means you can parse arbitrarily deeply nested objects without running out of call stack (although of course you might still run out of heap)
python3 setup.py develop --user
python3 fuzz.py
This is unstable and almost certainly will be subject to change
import cbrrr
object, parsed_length = cbrrr.parse_dag_cbor(data, cid_ctor)data is a bytes-like object - the data you want to parse.
cid_ctor is a callable that wraps any CID objects encountered. Use lambda x:x if you just want the CIDs as bytes (fastest), but you won't be able to distinguish them from "real" byte strings so you can't safely roundtrip the result. car_parse_benchmark.py shows an (unused) example of a CID wrapper class, but using this kills performance (I think I can mitigate this by creating a wrapper class in C instead)
object is the parsed result, as native python objects (with the exception of CIDs which are wrapped by whatever you supplied as cid_ctor).
parsed_length is the number of bytes that were actually parsed.
On parse failure, an exception is thrown.
- Tests
- Documentation
- Faster CID wrapper class (or similar)
- Round-trip re-serialisation
- DAG-JSON serialisation?
- Set up real fuzzing