Ready

sql/anomalies.sql

@@ -0,0 +1,30 @@
+-- Сигнал: день > 3 средних за предыдущие 7 календарных дней.
+-- Требуется 7 дней истории и ненулевая базовая сумма; пропущенные дни = 0.
+-- Целочисленное сравнение через quotient/remainder исключает overflow
+-- при умножении большой SQLite INTEGER суммы на 3 или 7.
+WITH RECURSIVE calendar(day) AS (
+    SELECT MIN(order_date) FROM facts HAVING COUNT(*) > 0
+    UNION ALL
+    SELECT date(day, '+1 day') FROM calendar
+    WHERE day < (SELECT MAX(order_date) FROM facts)
+), totals AS (
+    SELECT order_date, SUM(revenue_kopecks) AS revenue_kopecks
+    FROM daily_category GROUP BY order_date
+), filled AS (
+    SELECT day AS order_date, COALESCE(revenue_kopecks, 0) AS revenue_kopecks
+    FROM calendar LEFT JOIN totals ON totals.order_date = calendar.day
+), baseline AS (
+    SELECT *, SUM(revenue_kopecks) OVER w AS previous_7_sum_kopecks,
+              COUNT(*) OVER w AS history_days
+    FROM filled WINDOW w AS (ORDER BY order_date ROWS BETWEEN 7 PRECEDING AND 1 PRECEDING)
+), thresholds AS (
+    SELECT *, (previous_7_sum_kopecks / 7) * 3
+               + ((previous_7_sum_kopecks % 7) * 3) / 7 AS threshold_floor_kopecks
+    FROM baseline
+)
+SELECT order_date, revenue_kopecks, previous_7_sum_kopecks,
+       threshold_floor_kopecks
+FROM thresholds
+WHERE history_days = 7 AND previous_7_sum_kopecks > 0
+  AND revenue_kopecks > threshold_floor_kopecks
+ORDER BY order_date;

sql/daily_dynamics.sql

@@ -0,0 +1,18 @@
+-- Календарь от первого до последнего факта, отсутствующие дни дают 0.
+-- На пустой БД результат пуст; изменение первого дня NULL (нет предшественника).
+WITH RECURSIVE calendar(day) AS (
+    SELECT MIN(order_date) FROM facts HAVING COUNT(*) > 0
+    UNION ALL
+    SELECT date(day, '+1 day') FROM calendar
+    WHERE day < (SELECT MAX(order_date) FROM facts)
+), totals AS (
+    SELECT order_date, SUM(revenue_kopecks) AS revenue_kopecks
+    FROM daily_category GROUP BY order_date
+), filled AS (
+    SELECT day AS order_date, COALESCE(revenue_kopecks, 0) AS revenue_kopecks
+    FROM calendar LEFT JOIN totals ON totals.order_date = calendar.day
+)
+SELECT order_date, revenue_kopecks,
+       revenue_kopecks - LAG(revenue_kopecks) OVER (ORDER BY order_date)
+           AS change_kopecks
+FROM filled ORDER BY order_date;

sql/run_queries.py

@@ -0,0 +1,35 @@
+"""Execute bundled analytical SQL against SQLite read-only, write real CSVs."""
+import argparse
+import csv
+from pathlib import Path
+import sqlite3
+
+QUERY_NAMES = ('top_categories', 'daily_dynamics', 'anomalies')
+
+
+def run_queries(db, output):
+    output = Path(output)
+    output.mkdir(parents=True, exist_ok=True)
+    uri = Path(db).resolve().as_uri() + '?mode=ro'
+    counts = {}
+    with sqlite3.connect(uri, uri=True) as connection:
+        # One read snapshot keeps reports mutually consistent if a writer runs.
+        connection.execute('BEGIN')
+        for name in QUERY_NAMES:
+            sql = (Path(__file__).parent / f'{name}.sql').read_text(encoding='utf-8')
+            cursor = connection.execute(sql)
+            rows = cursor.fetchall()
+            with (output / f'{name}.csv').open('w', encoding='utf-8', newline='') as stream:
+                writer = csv.writer(stream)
+                writer.writerow([column[0] for column in cursor.description])
+                writer.writerows(rows)
+            counts[name] = len(rows)
+    return counts
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--db', type=Path, required=True)
+    parser.add_argument('--output', type=Path, required=True)
+    args = parser.parse_args()
+    print(run_queries(args.db, args.output))

sql/top_categories.sql

@@ -0,0 +1,5 @@
+-- Все категории по выручке RUB kopecks, при равенстве сортировка по имени.
+SELECT category, SUM(revenue_kopecks) AS revenue_kopecks
+FROM daily_category
+GROUP BY category
+ORDER BY revenue_kopecks DESC, category;