From dda7cc3b8027b759d28035ccc28f5ffcaaa083d8 Mon Sep 17 00:00:00 2001
From: Peter Stockings <pstockings@jadeworld.com>
Date: Sat, 13 Jan 2024 22:33:16 +1100
Subject: [PATCH] Optimise count_prs_over_time function to result in 14.2X
 improvement in speed which will hopefully remove lag on page requests

---
 utils.py | 61 ++++++++++++++++++++++++++++++--------------------------
 1 file changed, 33 insertions(+), 28 deletions(-)

diff --git a/utils.py b/utils.py
index 43b7b3b..9a3c3a4 100644
--- a/utils.py
+++ b/utils.py
@@ -344,47 +344,52 @@ def get_workout_counts(workouts, period='week'):
     return workout_counts
 
 def count_prs_over_time(workouts, period='week'):
-    # Convert to DataFrame
     df = pd.DataFrame(workouts)
     
     # Convert 'StartDate' to datetime
     df['StartDate'] = pd.to_datetime(df['StartDate'])
 
-    # Determine the range of periods to cover
-    min_date = df['StartDate'].min()
-    max_date = pd.Timestamp(datetime.now())
+    # Set period as week or month
+    df['Period'] = df['StartDate'].dt.to_period('W' if period == 'week' else 'M')
 
-    # Generate a complete range of periods
-    period_range = pd.date_range(start=min_date, end=max_date, freq='W-MON' if period == 'week' else 'MS')
+    # Group by Person, Exercise, and Period to find max Estimated1RM in each period
+    period_max = df.groupby(['PersonId', 'ExerciseId', 'Period'])['Estimated1RM'].max().reset_index()
 
-    # Initialize a dictionary to store PR counts and names
-    pr_counts = {
-        person_id: {
-            "PersonName": person_name,
-            "PRCounts": {p: 0 for p in period_range}
-        } for person_id, person_name in df[['PersonId', 'PersonName']].drop_duplicates().values
-    }
+    # Determine all-time max Estimated1RM up to the start of each period
+    period_max['AllTimeMax'] = period_max.groupby(['PersonId', 'ExerciseId'])['Estimated1RM'].cummax().shift(1)
 
-    # Process the workouts
-    for person_id, person_data in pr_counts.items():
-        person_df = df[df['PersonId'] == person_id]
+    # Identify PRs as entries where the period's max Estimated1RM exceeds the all-time max
+    period_max['IsPR'] = period_max['Estimated1RM'] > period_max['AllTimeMax']
 
-        for period_start in person_data["PRCounts"]:
-            period_end = period_start + pd.DateOffset(weeks=1) if period == 'week' else period_start + pd.DateOffset(months=1)
-            period_workouts = person_df[(person_df['StartDate'] >= period_start) & (person_df['StartDate'] < period_end)]
+    # Count PRs in each period for each person
+    pr_counts = period_max.groupby(['PersonId', 'Period'])['IsPR'].sum().reset_index()
 
-            for exercise_id in period_workouts['ExerciseId'].unique():
-                exercise_max = period_workouts[period_workouts['ExerciseId'] == exercise_id]['Estimated1RM'].max()
+    # Convert 'Period' to timestamp using the start date of the period
+    pr_counts['Period'] = pr_counts['Period'].apply(lambda x: x.start_time)
 
-                # Check if this is a PR
-                previous_max = person_df[(person_df['StartDate'] < period_start) & 
-                                         (person_df['ExerciseId'] == exercise_id)]['Estimated1RM'].max()
-                
-                if pd.isna(previous_max) or exercise_max > previous_max:
-                    person_data["PRCounts"][period_start] += 1
+    # Pivot table to get the desired output format
+    output = pr_counts.pivot(index='PersonId', columns='Period', values='IsPR').fillna(0)
 
-    return pr_counts
+    # Convert only the PR count columns to integers
+    for col in output.columns:
+        output[col] = output[col].astype(int)
 
+    # Merge with names and convert to desired format
+    names = df[['PersonId', 'PersonName']].drop_duplicates().set_index('PersonId')
+    output = names.join(output, how='left').fillna(0)
+
+    # Reset the index to bring 'PersonId' back as a column
+    output.reset_index(inplace=True)
+
+    # Convert to the final dictionary format with PRCounts nested
+    result = {}
+    for index, row in output.iterrows():
+        person_id = row['PersonId']
+        person_name = row['PersonName']
+        pr_counts = row.drop(['PersonId', 'PersonName']).to_dict()
+        result[person_id] = {"PersonName": person_name, "PRCounts": pr_counts}
+
+    return result
 
 def get_weekly_pr_graph_model(title, weekly_pr_data):
     # Assuming weekly_pr_data is in the format {1: {"PersonName": "Alice", "PRCounts": {Timestamp('2022-01-01', freq='W-MON'): 0, ...}}, 2: {...}, ...}