Post, new method, and rolling deprecation (#43)

* script

* finalize image

* Add wrapper for the series

* address rolling deprecation

* up the version

* run ruff

latent_calendar/extensions.py

@@ -113,6 +113,58 @@ class SeriesAccessor:
     def __init__(self, pandas_obj: pd.Series):
         self._obj = pandas_obj
 
+    def aggregate_events(
+        self,
+        minutes: int = 60,
+        as_multiindex: bool = True,
+    ) -> pd.Series:
+        """Transform event level Series to row of wide format.
+
+        Args:
+            minutes: The number of minutes to discretize by.
+            as_multiindex: whether to use MultiIndex columns
+
+        Returns:
+            Series that would be row of wide format
+
+        Examples:
+            Discretize datetime Series to 30 minutes
+
+            ```python
+            import pandas as pd
+
+            import matplotlib.pyplot as plt
+
+            from latent_calendar.datasets import load_chicago_bikes
+
+            df_trips = load_chicago_bikes()
+
+            start_times = df_trips["started_at"]
+
+            agg_start_times = start_times.cal.aggregate_events(minutes=30)
+            agg_start_times.cal.plot_row()
+            plt.show()
+
+
+            ```
+
+
+        """
+        name = self._obj.name or "timestamp"
+        return (
+            self._obj.rename(name)
+            .to_frame()
+            .assign(tmp=1)
+            .cal.aggregate_events(
+                by="tmp",
+                timestamp_col=name,
+                minutes=minutes,
+                as_multiindex=as_multiindex,
+            )
+            .iloc[0]
+            .rename(name)
+        )
+
     def timestamp_features(
         self, discretize: bool = True, minutes: int = 60, create_vocab: bool = True
     ) -> pd.DataFrame:

latent_calendar/segments/convolution.py

@@ -16,6 +16,7 @@ def sum_next_hours(df: pd.DataFrame, hours: int) -> pd.DataFrame:
     00 00 column would be 06 06 23
 
     TODO: Consider if negative hours should be allowed
+    TODO: Handle when minutes are not 60
 
     Arguments:
         df: DataFrame of probabilities or counts in wide format
@@ -35,9 +36,9 @@ def sum_next_hours(df: pd.DataFrame, hours: int) -> pd.DataFrame:
     return (
         pd.concat([df, df.iloc[:, :hours]], axis=1)
         .pipe(_reverse_columns)
-        .rolling(hours + 1, axis=1)
+        .T.rolling(hours + 1)
         .sum()
-        .iloc[:, hours:]
+        .T.iloc[:, hours:]
         .pipe(_reverse_columns)
     )
 

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "latent-calendar"
-version = "1.2.0"
+version = "1.3.0"
 description = "Analyzing and modeling weekly calendar distributions using latent components"
 authors = ["Will Dean <wd60622@gmail.com>"]
 readme = "README.md"

scripts/plot-post.py

@@ -0,0 +1,36 @@
+import pandas as pd 
+
+import matplotlib.pyplot as plt
+
+import latent_calendar
+
+
+if __name__ == "__main__": 
+    # More information on the dataset: 
+    # https://posit-dev.github.io/great-tables/reference/data.pizzaplace.html#great_tables.data.pizzaplace
+    file = "https://raw.githubusercontent.com/posit-dev/great-tables/main/great_tables/data/05-pizzaplace.csv"
+
+    df = pd.read_csv(file)
+
+    # Create a datetime column
+    datetime_column = pd.to_datetime(df["date"].str.cat(df["time"], sep=" "))
+
+
+    fig, axes = plt.subplots(nrows=1, ncols=2, sharey=True)
+    fig.suptitle("Plotting datetime column from posit-dev/great-tables pizza place dataset")
+
+    ax = axes[0]
+    datetime_column.sample(n=100, random_state=0).cal.plot(ax=ax)
+    ax.set_title("Continuous Series cal.plot()")
+
+    ax = axes[1]
+    (
+        datetime_column
+        .cal.aggregate_events(minutes=30)
+        .cal.plot_row(ax=ax)
+    )
+    ax.set_title("Discretized Series cal.plot_row()")
+
+    plt.show()
+
+

tests/test_extensions.py

@@ -6,7 +6,13 @@
 import matplotlib.pyplot as plt
 
 import latent_calendar  # noqa
-from latent_calendar.const import TIME_SLOTS, FULL_VOCAB, DAYS_IN_WEEK, HOURS_IN_DAY
+from latent_calendar.const import (
+    TIME_SLOTS,
+    FULL_VOCAB,
+    DAYS_IN_WEEK,
+    HOURS_IN_DAY,
+    create_full_vocab,
+)
 
 
 @pytest.fixture
@@ -270,3 +276,36 @@ def test_dataframe_conditional_probabilities(
         answer, index=df_wide_subset.index, columns=df_wide_subset.columns
     )
     pd.testing.assert_frame_equal(result, expected)
+
+
+def test_series_aggregate_events() -> None:
+    name = "events"
+    ser = pd.Series(
+        pd.to_datetime(
+            [
+                # 2021-01-01 is a Friday
+                "2021-01-01 00:00:00",
+                "2021-01-01 00:30:00",
+                "2021-01-01 01:00:00",
+                "2021-01-01 01:30:00",
+                "2021-01-01 01:45:00",
+                "2021-01-01 01:50:00",
+            ]
+        ),
+        name=name,
+    )
+
+    result = ser.cal.aggregate_events(minutes=30)
+
+    expected = pd.Series(
+        0,
+        index=create_full_vocab(days_in_week=7, minutes=30),
+        name=name,
+        dtype=result.dtype,
+    )
+    expected.loc[(4, 0)] = 1
+    expected.loc[(4, 0.5)] = 1
+    expected.loc[(4, 1)] = 1
+    expected.loc[(4, 1.5)] = 3
+
+    pd.testing.assert_series_equal(result, expected)