[Che4ter]

string.__new__ decodes bytes (with errors="surrogateescape"), but uri.__init__ passed the original argument to urlparse. urlparse decodes bytes as ASCII, so a uri created from bytes had two problems:

• A non-ASCII byte raised a UnicodeDecodeError, which stops a producer that builds records in a generator. For example the browser history plugins of dissect.target pass the UTF-8 bytes returned by try_idna to a uri field, see iexplore.history raises UnicodeDecodeError on URLs that are not IDNA-encodable (try_idna returns bytes for a uri field) dissect.target#1947.
• For ASCII bytes the parts (hostname, path, ...) were bytes, although they are annotated -> str.

from flow.record.fieldtypes import uri

uri(b"http://example.com/\xc3\x8c")
# UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 19

uri(b"http://example.com/a").hostname
# b'example.com'

Change

uri.__init__ parses str(self), the string that string.__new__ already built, instead of the argument. Input that is a str is unchanged. Input that is bytes now gives str parts, and bytes that are not valid UTF-8 are kept the way the string type keeps them.

Behaviour change: uri(<bytes>).hostname, .path and the other parts are str and no longer bytes. I found nothing in flow.record or dissect.target that reads them as bytes.

Tests

test_uri_type_bytes in tests/fieldtypes/test_fieldtypes.py: ASCII bytes, UTF-8 bytes with a non-ASCII character, and bytes that are not valid UTF-8. Without the change it fails (hostname is b'www.google.com'); with it the uri tests pass. The rest of the suite has the same failures with and without the change in my environment (17 in tests/tools/test_rdump.py and 1 in tests/test_regressions.py, and three adapter test files whose optional packages are not installed). ruff check and ruff format --check are clean.

Related to fox-it/dissect.target#1947