Update schema to correctly reflect supported missing value strategies…

… for different feature types (#3053)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

ludwig/constants.py

@@ -100,7 +100,6 @@
 MISSING_VALUE_STRATEGY_OPTIONS = [
     FILL_WITH_CONST,
     FILL_WITH_MODE,
-    FILL_WITH_MEAN,
     BFILL,
     FFILL,
     DROP_ROW,

ludwig/schema/features/preprocessing/bag.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import BAG, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
+from ludwig.constants import BAG, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -26,7 +26,7 @@ class BagPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a set column",
         parameter_metadata=FEATURE_METADATA[BAG][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/binary.py

@@ -1,7 +1,7 @@
 from typing import Union
 
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import BINARY, DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
+from ludwig.constants import BINARY, DROP_ROW, FILL_WITH_FALSE, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -17,8 +17,8 @@ class BinaryPreprocessingConfig(BasePreprocessingConfig):
     """BinaryPreprocessingConfig is a dataclass that configures the parameters used for a binary input feature."""
 
     missing_value_strategy: str = schema_utils.StringOptions(
-        MISSING_VALUE_STRATEGY_OPTIONS + ["fill_with_false"],
-        default="fill_with_false",
+        MISSING_VALUE_STRATEGY_OPTIONS + [FILL_WITH_FALSE],
+        default=FILL_WITH_FALSE,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a binary column",
         parameter_metadata=FEATURE_METADATA[BINARY][PREPROCESSING]["missing_value_strategy"],
@@ -63,7 +63,7 @@ class BinaryPreprocessingConfig(BasePreprocessingConfig):
 @ludwig_dataclass
 class BinaryOutputPreprocessingConfig(BinaryPreprocessingConfig):
     missing_value_strategy: str = schema_utils.StringOptions(
-        MISSING_VALUE_STRATEGY_OPTIONS + ["fill_with_false"],
+        MISSING_VALUE_STRATEGY_OPTIONS + [FILL_WITH_FALSE],
         default=DROP_ROW,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a binary output feature",

ludwig/schema/features/preprocessing/category.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import CATEGORY, DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
+from ludwig.constants import CATEGORY, DROP_ROW, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -17,7 +17,7 @@ class CategoryPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a category column",
         parameter_metadata=FEATURE_METADATA[CATEGORY][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/date.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DATE, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
+from ludwig.constants import BFILL, DATE, DROP_ROW, FFILL, FILL_WITH_CONST, PREPROCESSING
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -12,8 +12,8 @@
 @ludwig_dataclass
 class DatePreprocessingConfig(BasePreprocessingConfig):
     missing_value_strategy: str = schema_utils.StringOptions(
-        MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        [FILL_WITH_CONST, BFILL, FFILL, DROP_ROW],
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a date column",
         parameter_metadata=FEATURE_METADATA[DATE][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/h3.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import H3, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
+from ludwig.constants import FILL_WITH_CONST, H3, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -13,7 +13,7 @@
 class H3PreprocessingConfig(BasePreprocessingConfig):
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in an h3 column",
         parameter_metadata=FEATURE_METADATA[H3][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/number.py

@@ -1,5 +1,12 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, NUMBER, PREPROCESSING
+from ludwig.constants import (
+    DROP_ROW,
+    FILL_WITH_CONST,
+    FILL_WITH_MEAN,
+    MISSING_VALUE_STRATEGY_OPTIONS,
+    NUMBER,
+    PREPROCESSING,
+)
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -14,8 +21,8 @@ class NumberPreprocessingConfig(BasePreprocessingConfig):
     """NumberPreprocessingConfig is a dataclass that configures the parameters used for a number input feature."""
 
     missing_value_strategy: str = schema_utils.StringOptions(
-        MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        MISSING_VALUE_STRATEGY_OPTIONS + [FILL_WITH_MEAN],
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a number column",
         parameter_metadata=FEATURE_METADATA[NUMBER][PREPROCESSING]["missing_value_strategy"],
@@ -50,7 +57,7 @@ class NumberPreprocessingConfig(BasePreprocessingConfig):
 @ludwig_dataclass
 class NumberOutputPreprocessingConfig(NumberPreprocessingConfig):
     missing_value_strategy: str = schema_utils.StringOptions(
-        MISSING_VALUE_STRATEGY_OPTIONS,
+        MISSING_VALUE_STRATEGY_OPTIONS + [FILL_WITH_MEAN],
         default=DROP_ROW,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a number output feature",

ludwig/schema/features/preprocessing/sequence.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, SEQUENCE
+from ludwig.constants import DROP_ROW, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, SEQUENCE
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -75,7 +75,7 @@ class SequencePreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a text column",
         parameter_metadata=FEATURE_METADATA[SEQUENCE][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/set.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, SET
+from ludwig.constants import DROP_ROW, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, SET
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -24,7 +24,7 @@ class SetPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a set column",
         parameter_metadata=FEATURE_METADATA[SET][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/text.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, TEXT
+from ludwig.constants import DROP_ROW, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, TEXT
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -86,7 +86,7 @@ class TextPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a text column",
         parameter_metadata=FEATURE_METADATA[TEXT][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/timeseries.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, TIMESERIES
+from ludwig.constants import FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, TIMESERIES
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -44,7 +44,7 @@ class TimeseriesPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a text column",
         parameter_metadata=FEATURE_METADATA[TIMESERIES][PREPROCESSING]["missing_value_strategy"],

ludwig/schema/features/preprocessing/vector.py

@@ -1,5 +1,5 @@
 from ludwig.api_annotations import DeveloperAPI
-from ludwig.constants import DROP_ROW, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, VECTOR
+from ludwig.constants import DROP_ROW, FILL_WITH_CONST, MISSING_VALUE_STRATEGY_OPTIONS, PREPROCESSING, VECTOR
 from ludwig.schema import utils as schema_utils
 from ludwig.schema.features.preprocessing.base import BasePreprocessingConfig
 from ludwig.schema.features.preprocessing.utils import register_preprocessor
@@ -20,7 +20,7 @@ class VectorPreprocessingConfig(BasePreprocessingConfig):
 
     missing_value_strategy: str = schema_utils.StringOptions(
         MISSING_VALUE_STRATEGY_OPTIONS,
-        default="fill_with_const",
+        default=FILL_WITH_CONST,
         allow_none=False,
         description="What strategy to follow when there's a missing value in a vector column",
         parameter_metadata=FEATURE_METADATA[VECTOR][PREPROCESSING]["missing_value_strategy"],