Improving Test Coverage (#171)

* Test for end of record as a start event

* Test parsing time from None

* Tests for query edge cases

* Test for missing static predicates

* Ignore __init__

* Test label reference

* Tests for error messages for loading and initializing functions

* Test for missing references in nested derived predicates

* Explicitly include __main__.py

* Fix --cov-include

* Erase Codecov first

* Reset by deleting __main__.py

* Readd __main__.py

* Reset tests

* Re-add all tests

* Test for invalid args

* Ignore logging

* Tests for tuple and invalid endpoint_expr

* Remove unnecessary imports

* Tests for right and left inclusive and removed extra else from coverage

* Resolve lazyframe warnings

* Test missing right and left inclusive options

* Add codecov config

* No cover in __main__

* Change __main__ to run

* Fix numpy version and run.py tests

* No cover for main script

* No cover for main execution either

* ESGPT import error patch test

Author: justin13601

pyproject.toml

@@ -19,6 +19,7 @@ classifiers = [
 ]
 dependencies = [
     "polars >= 1.0.0, <= 1.17.1",
+    "numpy < 1.29.0",
     "bigtree == 0.18.*",
     "ruamel.yaml == 0.18.*",
     "hydra-core == 1.3.*",
@@ -37,7 +38,7 @@ your_package = ["*.pyi", "py.typed"]
 [tool.setuptools_scm]
 
 [project.scripts]
-aces-cli = "aces.__main__:main"
+aces-cli = "aces.run:main"
 expand_shards = "aces.expand_shards:main"
 
 [project.optional-dependencies]

src/aces/__init__.py

@@ -7,5 +7,5 @@
 __package_name__ = "es-aces"
 try:
     __version__ = version(__package_name__)
-except PackageNotFoundError:
+except PackageNotFoundError:  # pragma: no cover
     __version__ = "unknown"

src/aces/aggregate.py

@@ -370,6 +370,22 @@ def aggregate_event_bound_window(
         │ 2          ┆ 1989-12-08 16:22:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
         │ 2          ┆ 1989-12-10 03:07:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
         └────────────┴─────────────────────┴─────────────────────┴─────────────────────┴──────┴──────┴──────┘
+        >>> aggregate_event_bound_window(df, (True, "is_C", True, timedelta(days=3)))
+        shape: (8, 7)
+        ┌────────────┬─────────────────────┬─────────────────────┬─────────────────────┬──────┬──────┬──────┐
+        │ subject_id ┆ timestamp           ┆ timestamp_at_start  ┆ timestamp_at_end    ┆ is_A ┆ is_B ┆ is_C │
+        │ ---        ┆ ---                 ┆ ---                 ┆ ---                 ┆ ---  ┆ ---  ┆ ---  │
+        │ i64        ┆ datetime[μs]        ┆ datetime[μs]        ┆ datetime[μs]        ┆ i64  ┆ i64  ┆ i64  │
+        ╞════════════╪═════════════════════╪═════════════════════╪═════════════════════╪══════╪══════╪══════╡
+        │ 1          ┆ 1989-12-01 12:03:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
+        │ 1          ┆ 1989-12-03 13:14:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
+        │ 1          ┆ 1989-12-05 15:17:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
+        │ 2          ┆ 1989-12-02 12:03:00 ┆ 1989-12-05 12:03:00 ┆ 1989-12-06 15:17:00 ┆ 1    ┆ 1    ┆ 1    │
+        │ 2          ┆ 1989-12-04 13:14:00 ┆ 1989-12-07 13:14:00 ┆ 1989-12-10 03:07:00 ┆ 0    ┆ 2    ┆ 1    │
+        │ 2          ┆ 1989-12-06 15:17:00 ┆ 1989-12-09 15:17:00 ┆ 1989-12-10 03:07:00 ┆ 0    ┆ 1    ┆ 1    │
+        │ 2          ┆ 1989-12-08 16:22:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
+        │ 2          ┆ 1989-12-10 03:07:00 ┆ null                ┆ null                ┆ 0    ┆ 0    ┆ 0    │
+        └────────────┴─────────────────────┴─────────────────────┴─────────────────────┴──────┴──────┴──────┘
     """
     if not isinstance(endpoint_expr, ToEventWindowBounds):
         endpoint_expr = ToEventWindowBounds(*endpoint_expr)
@@ -845,6 +861,24 @@ def boolean_expr_bound_sum(
         │ 2          ┆ 1989-12-08 16:22:00 ┆ 1989-12-05 16:22:00 ┆ 1989-12-10 03:07:00 ┆ 1    ┆ 3    ┆ 2    │
         │ 2          ┆ 1989-12-10 03:07:00 ┆ 1989-12-07 03:07:00 ┆ 1989-12-10 03:07:00 ┆ 0    ┆ 2    ┆ 1    │
         └────────────┴─────────────────────┴─────────────────────┴─────────────────────┴──────┴──────┴──────┘
+
+        >>> boolean_expr_bound_sum(df, pl.col("idx").is_in([1, 4, 7]), "invalid_mode", "right",
+        ...     offset = timedelta(days=-3))
+        Traceback (most recent call last):
+            ...
+        ValueError: Mode 'invalid_mode' invalid!
+        >>> boolean_expr_bound_sum(df, pl.col("idx").is_in([1, 4, 7]), "row_to_bound", "invalid_closed",
+        ...     offset = timedelta(days=-3))
+        Traceback (most recent call last):
+            ...
+        ValueError: Closed 'invalid_closed' invalid!
+
+        >>> boolean_expr_bound_sum(df, pl.col("idx").is_in([1, 4, 7]), mode="row_to_bound",
+        ...         closed="right", offset=timedelta(days=1)).columns
+        ['subject_id', 'timestamp', 'timestamp_at_start', 'timestamp_at_end', 'idx', 'is_A', 'is_B', 'is_C']
+        >>> boolean_expr_bound_sum(df, pl.col("idx").is_in([1, 4, 7]), mode="row_to_bound",
+        ...         closed="left", offset=timedelta(days=-1)).columns
+        ['subject_id', 'timestamp', 'timestamp_at_start', 'timestamp_at_end', 'idx', 'is_A', 'is_B', 'is_C']
     """
     if mode not in ("bound_to_row", "row_to_bound"):
         raise ValueError(f"Mode '{mode}' invalid!")
@@ -1010,7 +1044,8 @@ def agg_offset_fn(c: str) -> pl.Expr:
         def agg_offset_fn(c: str) -> pl.Expr:
             return pl.col(c) + pl.col(f"{c}_in_offset_period")
 
-    else:
+    # Might not need as mode is already checked above (line 888)
+    else:  # pragma: no cover
         raise ValueError(f"Mode '{mode}' and offset '{offset}' invalid!")
 
     return with_at_boundary_events.join(

src/aces/config.py

@@ -205,6 +205,7 @@ def ESGPT_eval_expr(self, values_column: str | None = None) -> pl.Expr:
             >>> expr = PlainPredicateConfig("BP//systole", other_cols={"chamber": "atrial"}).ESGPT_eval_expr()
             >>> print(expr) # doctest: +NORMALIZE_WHITESPACE
             [(col("BP")) == (String(systole))].all_horizontal([[(col("chamber")) == (String(atrial))]])
+
             >>> expr = PlainPredicateConfig("BP//systolic", value_min=120).ESGPT_eval_expr()
             Traceback (most recent call last):
                 ...
@@ -522,6 +523,7 @@ class WindowConfig:
                             offset=datetime.timedelta(0))
         >>> target_window.root_node
         'end'
+
         >>> invalid_window = WindowConfig(
         ...     start="gap.end gap.start",
         ...     end="start -> discharge_or_death",
@@ -1014,7 +1016,6 @@ class TaskExtractorConfig:
         Traceback (most recent call last):
             ...
         FileNotFoundError: Cannot load missing configuration file /foo/non_existent_file.yaml!
-
         >>> import tempfile
         >>> with tempfile.NamedTemporaryFile(mode="w", suffix=".txt") as f:
         ...     config_path = Path(f.name)
@@ -1059,7 +1060,6 @@ class TaskExtractorConfig:
         Traceback (most recent call last):
             ...
         ValueError: Unrecognized keys in configuration file: 'foo, trigger'
-
         >>> predicates = {"foo bar": PlainPredicateConfig("foo")}
         >>> trigger = EventConfig("foo")
         >>> config = TaskExtractorConfig(predicates=predicates, trigger=trigger, windows={})
@@ -1083,7 +1083,6 @@ class TaskExtractorConfig:
         Traceback (most recent call last):
             ...
         KeyError: "Missing 1 relationships: Derived predicate 'foobar' references undefined predicate 'bar'"
-
         >>> predicates = {"foo": PlainPredicateConfig("foo")}
         >>> trigger = EventConfig("foo")
         >>> windows = {"foo bar": WindowConfig("gap.end", "start + 24h", True, True)}
@@ -1119,7 +1118,6 @@ class TaskExtractorConfig:
             ...
         ValueError: Only the 'start'/'end' of one window can be used as the index timestamp, found
         2 windows with index_timestamp: foo, bar
-
         >>> predicates = {"foo": PlainPredicateConfig("foo")}
         >>> trigger = EventConfig("bar")
         >>> config = TaskExtractorConfig(predicates=predicates, trigger=trigger, windows={})
@@ -1255,7 +1253,8 @@ def load(
             ...     "windows": {
             ...         "start": {
             ...             "start": None, "end": "trigger + 24h", "start_inclusive": True,
-            ...             "end_inclusive": True, "has": {"abnormal_labs": "(1, None)"},
+            ...             "end_inclusive": True, "label": "abnormal_labs",
+            ...             "has": {"abnormal_labs": "(1, None)"},
             ...         }
             ...     },
             ... }
@@ -1285,6 +1284,33 @@ def load(
                 ...
             ValueError: Predicate 'admission' is not defined correctly in the configuration file. Currently
             defined as the string: invalid. Please refer to the documentation for the supported formats.
+            >>> predicates_dict = {
+            ...     "predicates": {'adm': {"code": "admission"}},
+            ... }
+            >>> with (tempfile.NamedTemporaryFile(mode="w", suffix=".yaml") as config_fp,
+            ...      tempfile.NamedTemporaryFile(mode="w", suffix=".yaml") as pred_fp):
+            ...     config_path = Path(config_fp.name)
+            ...     pred_path = Path(pred_fp.name)
+            ...     yaml.dump(no_predicates_config, config_fp)
+            ...     yaml.dump(predicates_dict, pred_fp)
+            ...     cfg = TaskExtractorConfig.load(config_path, pred_path) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            KeyError: "Something referenced predicate 'admission' that wasn't defined in the configuration."
+            >>> config_dict = {
+            ...     "predicates": {"A": {"code": "A"}, "B": {"code": "B"}, "A_or_B": {"expr": "or(A, B)"},
+            ...                  "A_or_B_and_C": {"expr": "and(A_or_B, C)"}},
+            ...     "trigger": "_ANY_EVENT",
+            ...     "windows": {"start": {"start": None, "end": "trigger + 24h", "start_inclusive": True,
+            ...             "end_inclusive": True, "has": {"A_or_B_and_C": "(1, None)"}}},
+            ... }
+            >>> with tempfile.NamedTemporaryFile(mode="w", suffix=".yaml") as f:
+            ...     config_path = Path(f.name)
+            ...     yaml.dump(config_dict, f)
+            ...     cfg = TaskExtractorConfig.load(config_path) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            KeyError: "Predicate 'C' referenced in 'A_or_B_and_C' is not defined in the configuration."
         """
         if isinstance(config_path, str):
             config_path = Path(config_path)
@@ -1351,7 +1377,7 @@ def load(
         all_predicates = {**final_predicates, **final_demographics}
 
         logger.info("Parsing windows...")
-        if windows is None:
+        if windows is None:  # pragma: no cover
             windows = {}
             logger.warning(
                 "No windows specified in configuration file. Extracting only matching trigger events."
@@ -1432,6 +1458,23 @@ def _initialize_predicates(self) -> None:
 
         Raises:
             ValueError: If the predicate name is not valid.
+
+        Examples:
+            >>> import networkx as nx
+            >>> TaskExtractorConfig(
+            ...     predicates={
+            ...         "A": DerivedPredicateConfig("and(A, B)"),  # A depends on B
+            ...         "B": DerivedPredicateConfig("and(B, C)"),  # B depends on C
+            ...         "C": DerivedPredicateConfig("and(A, C)"),  # C depends on A (Cyclic dependency)
+            ...     },
+            ...     trigger=EventConfig("A"),
+            ...     windows={},
+            ... ) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            ValueError: Predicate graph is not a directed acyclic graph!
+            Cycle found: [('A', 'A')]
+            Graph: None
         """
 
         dag_relationships = []
@@ -1479,6 +1522,42 @@ def _initialize_windows(self) -> None:
 
         Raises:
             ValueError: If the window name is not valid.
+
+            Examples:
+            >>> TaskExtractorConfig(  # doctest: +NORMALIZE_WHITESPACE
+            ...     predicates={"A": PlainPredicateConfig("A")},
+            ...     windows={
+            ...         "win1": WindowConfig(None, "trigger", True, False, has={"B": "(1, 0)"}) # B undefined
+            ...     },
+            ...     trigger=EventConfig("_ANY_EVENT"),
+            ... ) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            KeyError: "Window 'win1' references undefined predicate 'B'.
+            Window predicates: B;
+            Defined predicates: A"
+            >>> TaskExtractorConfig(
+            ...     predicates={"A": PlainPredicateConfig("A")},
+            ...     windows={
+            ...         "win1": WindowConfig(None, "event_not_trigger", True, False)
+            ...     },
+            ...     trigger=EventConfig("_ANY_EVENT"),
+            ... ) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            KeyError: "Window 'win1' references undefined trigger event
+            'event_not_trigger' -- must be trigger!"
+            >>> TaskExtractorConfig(
+            ...     predicates={"A": PlainPredicateConfig("A")},
+            ...     windows={
+            ...         "win1": WindowConfig("win2.end", "start -> A", True, False)
+            ...     },
+            ...     trigger=EventConfig("_ANY_EVENT"),
+            ... ) # doctest: +NORMALIZE_WHITESPACE
+            Traceback (most recent call last):
+                ...
+            KeyError: "Window 'win1' references undefined window 'win2' for event 'end'.
+            Allowed windows: win1"
         """
 
         for name in self.windows:
@@ -1559,8 +1638,8 @@ def _initialize_windows(self) -> None:
             for predicate in window.referenced_predicates - {ANY_EVENT_COLUMN}:
                 if predicate not in self.predicates:
                     raise KeyError(
-                        f"Window '{name}' references undefined predicate '{predicate}'.\n"
-                        f"Window predicates: {', '.join(window.referenced_predicates)}\n"
+                        f"Window '{name}' references undefined predicate '{predicate}'. "
+                        f"Window predicates: {', '.join(window.referenced_predicates)}; "
                         f"Defined predicates: {', '.join(self.predicates.keys())}"
                     )
 
@@ -1580,15 +1659,17 @@ def _initialize_windows(self) -> None:
                         f"Window '{name}' references undefined window '{referenced_window}' "
                         f"for event '{referenced_event}'. Allowed windows: {', '.join(self.windows.keys())}"
                     )
-                if referenced_event not in {"start", "end"}:
+                # Might not be needed as valid window event references are already checked (line 660)
+                if referenced_event not in {"start", "end"}:  # pragma: no cover
                     raise KeyError(
                         f"Window '{name}' references undefined event '{referenced_event}' "
                         f"for window '{referenced_window}'. Allowed events: 'start', 'end'"
                     )
 
                 parent_node = f"{referenced_window}.{referenced_event}"
                 window_nodes[f"{name}.{window.root_node}"].parent = window_nodes[parent_node]
-            else:
+            # Might not be needed as valid window event references are already checked (line 660)
+            else:  # pragma: no cover
                 raise ValueError(
                     f"Window '{name}' references invalid event '{window.referenced_event}' "
                     "must be of length 1 or 2."

src/aces/constraints.py

@@ -166,6 +166,10 @@ def check_static_variables(patient_demographics: list[str], predicates_df: pl.Da
         │ 1          ┆ 1989-12-02 12:03:00 ┆ 1    ┆ 0    ┆ 1    │
         │ 1          ┆ 1989-12-06 11:00:00 ┆ 0    ┆ 0    ┆ 0    │
         └────────────┴─────────────────────┴──────┴──────┴──────┘
+        >>> check_static_variables(['female'], predicates_df)
+        Traceback (most recent call last):
+            ...
+        ValueError: Static predicate 'female' not found in the predicates dataframe.
     """
     for demographic in patient_demographics:
         if demographic not in predicates_df.columns:

src/aces/expand_shards.py

@@ -48,6 +48,11 @@ def expand_shards(*shards: str) -> str:
         ...     result = expand_shards(tmpdirname)
         ...     sorted(result.split(","))
         ['1', '3', 'evens/0/file_0', 'evens/0/file_2']
+
+        >>> expand_shards("train.invalid")
+        Traceback (most recent call last):
+            ...
+        ValueError: Invalid shard format: train.invalid
     """
 
     result = []
@@ -71,9 +76,9 @@ def expand_shards(*shards: str) -> str:
     return ",".join(result)
 
 
-def main() -> None:
+def main() -> None:  # pragma: no cover
     print(expand_shards(*sys.argv[1:]))
 
 
-if __name__ == "__main__":
+if __name__ == "__main__":  # pragma: no cover
     main()