[FEAT] Add time window parameter selection to hashtags test

analyzers/hashtags/interface.py

@@ -2,14 +2,18 @@
     AnalyzerInput,
     AnalyzerInterface,
     AnalyzerOutput,
+    AnalyzerParam,
     InputColumn,
     OutputColumn,
 )
+from analyzer_interface.params import TimeBinningParam
 
 COL_AUTHOR_ID = "user_id"
 COL_TIME = "time"
 COL_POST = "text"
 
+PARAM_TIME_WINDOW = "time_window"
+
 OUTPUT_GINI = "hashtag_analysis"
 OUTPUT_COL_USERS = "users"
 OUTPUT_COL_TIMESPAN = "timewindow_start"
@@ -76,6 +80,29 @@
             ),
         ]
     ),
+    params=[
+        AnalyzerParam(
+            id=PARAM_TIME_WINDOW,
+            human_readable_name="Time window of analysis",
+            description="""
+The duration over which to compute the gini coefficient. After
+selecting a time window (e.g. 12 hours), the dataset will be
+chunked into equal-duration (e.g. 12-hour long) time windows. The gini 
+coefficient is computed over hashtags found in each time window chunk. 
+
+This value determines the time granularity of the analysis. The optimal
+value depends on the time span of the dataset. Generally, you want to 
+choose time window long enough that there is a meaningful number of hashtags
+in that time period (if there are none, gini will be 0 for those time windows).
+
+For example, if the full dataset spans multiple years, it makes sense
+to choose a time window that spans multiple days if not months. Conversely,
+if a dataset spans a few weeks, it makes sense to choose a time window of
+a few hours or one day etc.
+            """,
+            type=TimeBinningParam(),
+        )
+    ],
     outputs=[
         AnalyzerOutput(
             id=OUTPUT_GINI,

analyzers/hashtags/main.py

@@ -14,6 +14,7 @@
     OUTPUT_COL_TIMESPAN,
     OUTPUT_COL_USERS,
     OUTPUT_GINI,
+    PARAM_TIME_WINDOW,
 )
 
 
@@ -88,10 +89,12 @@ def main(context: PrimaryAnalyzerContext):
     input_reader = context.input()
     df_input = input_reader.preprocess(pl.read_parquet(input_reader.parquet_path))
 
+    time_window_param = context.params.get(PARAM_TIME_WINDOW)
+
     # window hard-coded to 1hr for now
     df_out = hashtag_analysis(
         data_frame=df_input,
-        every="12h",
+        every=time_window_param.to_polars_truncate_spec(),  # returns '12h', '5d' etc.
     )
 
     df_out.write_parquet(context.output(OUTPUT_GINI).parquet_path)

analyzers/hashtags/test_hashtags_analyzer.py

@@ -3,6 +3,7 @@
 import numpy as np
 import polars as pl
 
+from analyzer_interface.params import TimeBinningValue
 from preprocessing.series_semantic import datetime_string, identifier, text_catch_all
 from testing import CsvTestData, JsonTestData, test_primary_analyzer
 
@@ -88,6 +89,7 @@ def test_hashtag_analyzer():
                 COL_POST: text_catch_all,
             },
         ),
+        params={"time_window": TimeBinningValue(unit="hour", amount=12)},
         outputs={
             OUTPUT_GINI: JsonTestData(
                 os.path.join(test_data_dir, "hashtag_test_output.json")