Skip to content

feat: Add Set as feature type #5888

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
franciscojavierarceo merged 11 commits into from
Jan 29, 2026
Merged

feat: Add Set as feature type #5888

Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
821b1f2
Add Set types to Feast type system
franciscojavierarceo Jan 27, 2026
8ba622e
Fix Milvus integration for Set types
franciscojavierarceo Jan 27, 2026
a56b774
Fix item_id type mismatch in milvus test
franciscojavierarceo Jan 27, 2026
325edb7
Apply ruff formatting
franciscojavierarceo Jan 27, 2026
947bfb7
Merge branch 'master' into copilot/add-set-as-feature-type
franciscojavierarceo Jan 27, 2026
191782b
refactor: Extract Set type conversion to separate function
franciscojavierarceo Jan 27, 2026
45251de
Apply ruff formatting to refactored code
franciscojavierarceo Jan 27, 2026
2ad1eec
Delete test_registry
franciscojavierarceo Jan 28, 2026
4f7128a
Merge branch 'master' into copilot/add-set-as-feature-type
franciscojavierarceo Jan 28, 2026
5e87fc9
Merge branch 'master' into copilot/add-set-as-feature-type
franciscojavierarceo Jan 28, 2026
4a45b75
Merge branch 'master' into copilot/add-set-as-feature-type
franciscojavierarceo Jan 28, 2026
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
Prev Previous commit
Next Next commit
refactor: Extract Set type conversion to separate function 
Extract the long set type handling code from _python_value_to_proto_value
into a dedicated _python_set_to_proto_values function for better readability
and maintainability.

Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
  • Loading branch information
@franciscojavierarceo @claude
franciscojavierarceo and claude committed Jan 27, 2026
commit 191782be6e0bdbe23e51a6991d25a26d17b4bca5
228 changes: 123 additions & 105 deletions sdk/python/feast/type_map.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -483,6 +483,128 @@ def _python_datetime_to_int_timestamp(
return int_timestamps


def _python_set_to_proto_values(
feast_value_type: ValueType, values: List[Any]
) -> List[ProtoValue]:
"""
Converts Python set values to Feast Proto Values.

Args:
feast_value_type: The target set value type
values: List of set values that will be converted

Returns:
List of Feast Value Proto
"""
# Feature can be set but None is still valid
if feast_value_type not in PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE:
return []

set_proto_type, set_field_name, set_valid_types = (
PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE[feast_value_type]
)

# Convert set to list for proto (proto doesn't have native set type)
# We store unique values in a repeated field
def convert_set_to_list(value):
if value is None:
return None
# If it's already a set, convert to list
if isinstance(value, set):
return list(value)
# If it's a list/tuple/ndarray, remove duplicates
elif isinstance(value, (list, tuple, np.ndarray)):
return list(set(value))
else:
return value

converted_values = [convert_set_to_list(v) for v in values]
sample = next(filter(_non_empty_value, converted_values), None)

# Bytes to array type conversion
if isinstance(sample, (bytes, bytearray)):
# Bytes of an array containing elements of bytes not supported
if feast_value_type == ValueType.BYTES_SET:
raise _type_err(sample, ValueType.BYTES_SET)

json_sample = json.loads(sample)
if isinstance(json_sample, list):
json_values = [
json.loads(value) if value is not None else None
for value in converted_values
]
if feast_value_type == ValueType.BOOL_SET:
json_values = [
[bool(item) for item in list_item]
if list_item is not None
else None
for list_item in json_values
]
return [
(
ProtoValue(**{set_field_name: set_proto_type(val=v)}) # type: ignore
if v is not None
else ProtoValue()
)
for v in json_values
]
raise _type_err(sample, set_valid_types[0])

if sample is not None and not all(
type(item) in set_valid_types for item in sample
):
for item in sample:
if type(item) not in set_valid_types:
if feast_value_type in [
ValueType.INT32_SET,
ValueType.INT64_SET,
]:
if not any(np.isnan(item) for item in sample):
logger.error(
"Set of Int32 or Int64 type has NULL values."
)
raise _type_err(item, set_valid_types[0])

if feast_value_type == ValueType.UNIX_TIMESTAMP_SET:
result = []
for value in converted_values:
if value is not None:
result.append(
ProtoValue(
unix_timestamp_set_val=Int64Set(
val=_python_datetime_to_int_timestamp(value) # type: ignore
)
)
)
else:
result.append(ProtoValue())
return result
if feast_value_type == ValueType.BOOL_SET:
result = []
for value in converted_values:
if value is not None:
result.append(
ProtoValue(
**{
set_field_name: set_proto_type(
val=[bool(e) for e in value] # type: ignore
)
}
)
)
else:
result.append(ProtoValue())
return result
return [
(
ProtoValue(**{set_field_name: set_proto_type(val=value)}) # type: ignore
if value is not None
else ProtoValue()
)
for value in converted_values
]


def _python_value_to_proto_value(
feast_value_type: ValueType, values: List[Any]
) -> List[ProtoValue]:
Expand Down Expand Up @@ -600,111 +722,7 @@ def _python_value_to_proto_value(

# Detect set type and handle separately
if "set" in feast_value_type.name.lower():
# Feature can be set but None is still valid
if feast_value_type in PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE:
set_proto_type, set_field_name, set_valid_types = (
PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE[feast_value_type]
)

# Convert set to list for proto (proto doesn't have native set type)
# We store unique values in a repeated field
def convert_set_to_list(value):
if value is None:
return None
# If it's already a set, convert to list
if isinstance(value, set):
return list(value)
# If it's a list/tuple/ndarray, remove duplicates
elif isinstance(value, (list, tuple, np.ndarray)):
return list(set(value))
else:
return value

converted_values = [convert_set_to_list(v) for v in values]
sample = next(filter(_non_empty_value, converted_values), None)

# Bytes to array type conversion
if isinstance(sample, (bytes, bytearray)):
# Bytes of an array containing elements of bytes not supported
if feast_value_type == ValueType.BYTES_SET:
raise _type_err(sample, ValueType.BYTES_SET)

json_sample = json.loads(sample)
if isinstance(json_sample, list):
json_values = [
json.loads(value) if value is not None else None
for value in converted_values
]
if feast_value_type == ValueType.BOOL_SET:
json_values = [
[bool(item) for item in list_item]
if list_item is not None
else None
for list_item in json_values
]
return [
(
ProtoValue(**{set_field_name: set_proto_type(val=v)}) # type: ignore
if v is not None
else ProtoValue()
)
for v in json_values
]
raise _type_err(sample, set_valid_types[0])

if sample is not None and not all(
type(item) in set_valid_types for item in sample
):
for item in sample:
if type(item) not in set_valid_types:
if feast_value_type in [
ValueType.INT32_SET,
ValueType.INT64_SET,
]:
if not any(np.isnan(item) for item in sample):
logger.error(
"Set of Int32 or Int64 type has NULL values."
)
raise _type_err(item, set_valid_types[0])

if feast_value_type == ValueType.UNIX_TIMESTAMP_SET:
result = []
for value in converted_values:
if value is not None:
result.append(
ProtoValue(
unix_timestamp_set_val=Int64Set(
val=_python_datetime_to_int_timestamp(value) # type: ignore
)
)
)
else:
result.append(ProtoValue())
return result
if feast_value_type == ValueType.BOOL_SET:
result = []
for value in converted_values:
if value is not None:
result.append(
ProtoValue(
**{
set_field_name: set_proto_type(
val=[bool(e) for e in value] # type: ignore
)
}
)
)
else:
result.append(ProtoValue())
return result
return [
(
ProtoValue(**{set_field_name: set_proto_type(val=value)}) # type: ignore
if value is not None
else ProtoValue()
)
for value in converted_values
]
return _python_set_to_proto_values(feast_value_type, values)

# Handle scalar types below
else:
Expand Down