-- Replace the unfinished weekly-bucket page-visit tracker with a daily-bucket
-- design. Daily granularity supports every trending window we expose
-- (week / month / 3-month / semester / year) with exact day boundaries; the
-- previous weekly bucket would have produced ±7-day error on any window that
-- isn't a whole number of weeks (i.e. month/year).
--
-- Also drops the never-read `locations.total_visited` integer column. The
-- lifetime total is recoverable from SUM(visit_count) on this table.

DROP INDEX IF EXISTS idx_location_page_visits_week_key_visit_count;
DROP TABLE IF EXISTS location_page_visits;

CREATE TABLE location_page_visits (
  id          SERIAL PRIMARY KEY,
  location_id INT  NOT NULL REFERENCES locations(id),
  day         DATE NOT NULL,
  visit_count BIGINT NOT NULL DEFAULT 0,
  updated_at  TIMESTAMP NOT NULL DEFAULT now(),
  UNIQUE(location_id, day)
);

-- Trending: WHERE day >= now()::date - N ORDER BY visit_count DESC
CREATE INDEX idx_location_page_visits_day_count
  ON location_page_visits(day, visit_count DESC);

-- Per-location lookup (lifetime sum, "visits over time" charts, etc.)
CREATE INDEX idx_location_page_visits_location_day
  ON location_page_visits(location_id, day);

ALTER TABLE locations DROP COLUMN IF EXISTS total_visited;