Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions py/src/braintrust/span_identifier_v3.py
Original file line number Diff line number Diff line change
Expand Up @@ -12,10 +12,16 @@


def _try_make_uuid(s):
# Only compress canonical (lowercase, dashed) UUID strings. `UUID()` also
# accepts 32-char hex, braces, `urn:uuid:` and uppercase, but decoding always
# yields the canonical form, so compressing those would silently change the ID
# (e.g. a hex trace ID would come back with dashes). Anything else is stored
# verbatim in the JSON section.
try:
ret = UUID(s).bytes
assert len(ret) == 16
return ret, True
u = UUID(s)
if str(u) != s:
return None, False
return u.bytes, True
except Exception:
return None, False

Expand Down
51 changes: 40 additions & 11 deletions py/src/braintrust/test_span_components.py
Original file line number Diff line number Diff line change
Expand Up @@ -50,27 +50,56 @@ def test_with_metadata(self):
assert imported.object_id == components.object_id
assert imported.propagated_event == components.propagated_event

def test_otel_ids_fail_roundtrip(self):
"""Test that V3 fails to preserve OTEL hex strings for 16-byte IDs (converts to UUID format)."""
def test_otel_ids_roundtrip_unchanged(self):
"""V3 must preserve non-canonical IDs such as OTEL hex strings exactly.

Python's ``UUID()`` also parses 32-char hex, so V3 used to compress these
and decode them back in dashed form, silently changing the ID.
"""
otel_gen = OTELIDGenerator()
trace_id = otel_gen.get_trace_id() # 32-char hex (16 bytes)
span_id = otel_gen.get_span_id() # 16-char hex (8 bytes)

# Use 16-byte hex strings for object_id and root_span_id to see UUID conversion
components = SpanComponentsV3(
object_type=SpanObjectTypeV3.PROJECT_LOGS,
object_id=trace_id, # 16-byte hex should get converted to UUID format
object_id=trace_id,
row_id="test-row-id",
span_id=span_id, # 8-byte hex might be preserved
root_span_id=trace_id, # 16-byte hex should get converted to UUID format
span_id=span_id,
root_span_id=trace_id,
)

exported = components.to_str()
imported = SpanComponentsV3.from_str(exported)
imported = SpanComponentsV3.from_str(components.to_str())

assert imported.object_id == trace_id
assert imported.span_id == span_id
assert imported.root_span_id == trace_id

@pytest.mark.parametrize(
"value_fn",
[
lambda u: u.upper(),
lambda u: "{" + u + "}",
lambda u: "urn:uuid:" + u,
lambda u: u.replace("-", ""),
],
ids=["uppercase", "braced", "urn", "undashed"],
)
def test_non_canonical_uuid_strings_roundtrip_unchanged(self, value_fn):
"""Alternate spellings of a UUID must round-trip verbatim, not be canonicalized."""
value = value_fn(str(uuid4()))
components = SpanComponentsV3(
object_type=SpanObjectTypeV3.PROJECT_LOGS,
object_id=str(uuid4()),
row_id=value,
span_id=value,
root_span_id=value,
)

imported = SpanComponentsV3.from_str(components.to_str())

# V3 should convert 16-byte hex strings to UUID format (with dashes)
# Note: span_id (8 bytes) may or may not be converted depending on whether UUID parsing succeeds
assert imported.root_span_id != trace_id # 16-byte should have dashes added
assert imported.row_id == value
assert imported.span_id == value
assert imported.root_span_id == value


class TestSpanComponentsV4:
Expand Down