Commit Graph
4 Commits
Author SHA1 Message Date
lennylxx 4c2939d66e fix(python): guard against None before .decode() on split_names metadata key (#3229)
`.get(b"split_names", None).decode()` was called unconditionally in both
Permutations.__init__ and Permutation.from_tables(), crashing with
AttributeError when schema metadata existed but lacked the split_names
key. Guard the decode behind a None check and add regression tests.
2026-04-08 16:04:13 -07:00
lennylxx 9f85d4c639 fix(embeddings): add missing urllib.request import in url_retrieve (#3190)
url_retrieve() calls urllib.request.urlopen() but only urllib.error was
imported, causing AttributeError for any HTTP URL input. This affects
open-clip, siglip, and jinaai embedding functions when processing image
URLs.

The bug has existed since the embeddings API refactor (#580) but was
masked because most users pass local file paths or bytes rather than
HTTP URLs.
2026-03-30 12:03:44 -07:00
lennylxx 1d1cafb59c fix(python): don't assign dict.update() return value in _sanitize_data (#3198)
dict.update() mutates in place and returns None. Assigning its result
caused with_metadata(None) to strip all schema metadata when embedding
metadata was merged during create_table with embedding_functions.
2026-03-30 10:15:45 -07:00
lennylxx 74f457a0f2 fix(rust): handle Mutex lock poisoning gracefully across codebase (#3196)
Replace ~30 production `lock().unwrap()` calls that would cascade-panic
on a poisoned Mutex. Functions returning `Result` now propagate the
poison as an error via `?` (leveraging the existing `From<PoisonError>`
impl). Functions without a `Result` return recover via
`unwrap_or_else(|e| e.into_inner())`, which is safe because the guarded
data (counters, caches, RNG state) remains logically valid after a
panic.
2026-03-30 09:25:18 -07:00