-- Migration 121: Sales Analytics Materialized Views
-- Pre-computed views for sales velocity, brand market share, and store performance

-- ============================================================
-- VIEW 1: Daily Sales Estimates (per product/store)
-- Calculates delta between consecutive snapshots
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_daily_sales_estimates AS
WITH qty_deltas AS (
  SELECT
    dispensary_id,
    product_id,
    brand_name,
    category,
    DATE(captured_at) AS sale_date,
    price_rec,
    quantity_available,
    LAG(quantity_available) OVER (
      PARTITION BY dispensary_id, product_id
      ORDER BY captured_at
    ) AS prev_quantity
  FROM inventory_snapshots
  WHERE quantity_available IS NOT NULL
    AND captured_at >= NOW() - INTERVAL '30 days'
)
SELECT
  dispensary_id,
  product_id,
  brand_name,
  category,
  sale_date,
  AVG(price_rec) AS avg_price,
  SUM(GREATEST(0, COALESCE(prev_quantity, 0) - quantity_available)) AS units_sold,
  SUM(GREATEST(0, quantity_available - COALESCE(prev_quantity, 0))) AS units_restocked,
  SUM(GREATEST(0, COALESCE(prev_quantity, 0) - quantity_available) * COALESCE(price_rec, 0)) AS revenue_estimate,
  COUNT(*) AS snapshot_count
FROM qty_deltas
WHERE prev_quantity IS NOT NULL
GROUP BY dispensary_id, product_id, brand_name, category, sale_date;

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_daily_sales_pk
  ON mv_daily_sales_estimates(dispensary_id, product_id, sale_date);
CREATE INDEX IF NOT EXISTS idx_mv_daily_sales_brand
  ON mv_daily_sales_estimates(brand_name, sale_date);
CREATE INDEX IF NOT EXISTS idx_mv_daily_sales_category
  ON mv_daily_sales_estimates(category, sale_date);
CREATE INDEX IF NOT EXISTS idx_mv_daily_sales_date
  ON mv_daily_sales_estimates(sale_date DESC);


-- ============================================================
-- VIEW 2: Brand Market Share by State
-- Weighted distribution across stores
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_brand_market_share AS
WITH brand_presence AS (
  SELECT
    sp.brand AS brand_name,
    d.state AS state_code,
    COUNT(DISTINCT sp.dispensary_id) AS stores_carrying,
    COUNT(*) AS sku_count,
    SUM(CASE WHEN sp.is_in_stock THEN 1 ELSE 0 END) AS in_stock_skus,
    AVG(sp.price_rec) AS avg_price
  FROM store_products sp
  JOIN dispensaries d ON d.id = sp.dispensary_id
  WHERE sp.brand IS NOT NULL
    AND d.state IS NOT NULL
  GROUP BY sp.brand, d.state
),
state_totals AS (
  SELECT
    d.state AS state_code,
    COUNT(DISTINCT d.id) FILTER (WHERE d.crawl_enabled) AS total_stores
  FROM dispensaries d
  WHERE d.state IS NOT NULL
  GROUP BY d.state
)
SELECT
  bp.brand_name,
  bp.state_code,
  bp.stores_carrying,
  st.total_stores,
  ROUND(bp.stores_carrying::NUMERIC * 100 / NULLIF(st.total_stores, 0), 2) AS penetration_pct,
  bp.sku_count,
  bp.in_stock_skus,
  bp.avg_price,
  NOW() AS calculated_at
FROM brand_presence bp
JOIN state_totals st ON st.state_code = bp.state_code;

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_brand_market_pk
  ON mv_brand_market_share(brand_name, state_code);
CREATE INDEX IF NOT EXISTS idx_mv_brand_market_state
  ON mv_brand_market_share(state_code);
CREATE INDEX IF NOT EXISTS idx_mv_brand_market_penetration
  ON mv_brand_market_share(penetration_pct DESC);


-- ============================================================
-- VIEW 3: SKU Velocity (30-day rolling)
-- Average daily units sold per SKU
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_sku_velocity AS
SELECT
  dse.product_id,
  dse.brand_name,
  dse.category,
  dse.dispensary_id,
  d.name AS dispensary_name,
  d.state AS state_code,
  SUM(dse.units_sold) AS total_units_30d,
  SUM(dse.revenue_estimate) AS total_revenue_30d,
  COUNT(DISTINCT dse.sale_date) AS days_with_sales,
  ROUND(SUM(dse.units_sold)::NUMERIC / NULLIF(COUNT(DISTINCT dse.sale_date), 0), 2) AS avg_daily_units,
  AVG(dse.avg_price) AS avg_price,
  CASE
    WHEN SUM(dse.units_sold)::NUMERIC / NULLIF(COUNT(DISTINCT dse.sale_date), 0) >= 5 THEN 'hot'
    WHEN SUM(dse.units_sold)::NUMERIC / NULLIF(COUNT(DISTINCT dse.sale_date), 0) >= 1 THEN 'steady'
    WHEN SUM(dse.units_sold)::NUMERIC / NULLIF(COUNT(DISTINCT dse.sale_date), 0) >= 0.1 THEN 'slow'
    ELSE 'stale'
  END AS velocity_tier,
  NOW() AS calculated_at
FROM mv_daily_sales_estimates dse
JOIN dispensaries d ON d.id = dse.dispensary_id
WHERE dse.sale_date >= CURRENT_DATE - INTERVAL '30 days'
GROUP BY dse.product_id, dse.brand_name, dse.category, dse.dispensary_id, d.name, d.state;

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_sku_velocity_pk
  ON mv_sku_velocity(dispensary_id, product_id);
CREATE INDEX IF NOT EXISTS idx_mv_sku_velocity_brand
  ON mv_sku_velocity(brand_name);
CREATE INDEX IF NOT EXISTS idx_mv_sku_velocity_tier
  ON mv_sku_velocity(velocity_tier);
CREATE INDEX IF NOT EXISTS idx_mv_sku_velocity_state
  ON mv_sku_velocity(state_code);
CREATE INDEX IF NOT EXISTS idx_mv_sku_velocity_units
  ON mv_sku_velocity(total_units_30d DESC);


-- ============================================================
-- VIEW 4: Store Performance Rankings
-- Revenue estimates and brand diversity per store
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_store_performance AS
SELECT
  d.id AS dispensary_id,
  d.name AS dispensary_name,
  d.city,
  d.state AS state_code,
  -- Revenue metrics from sales estimates
  COALESCE(sales.total_revenue_30d, 0) AS total_revenue_30d,
  COALESCE(sales.total_units_30d, 0) AS total_units_30d,
  -- Inventory metrics
  COUNT(DISTINCT sp.id) AS total_skus,
  COUNT(DISTINCT sp.id) FILTER (WHERE sp.is_in_stock) AS in_stock_skus,
  -- Brand diversity
  COUNT(DISTINCT sp.brand) AS unique_brands,
  COUNT(DISTINCT sp.category) AS unique_categories,
  -- Pricing
  AVG(sp.price_rec) AS avg_price,
  -- Activity
  MAX(sp.updated_at) AS last_updated,
  NOW() AS calculated_at
FROM dispensaries d
LEFT JOIN store_products sp ON sp.dispensary_id = d.id
LEFT JOIN (
  SELECT
    dispensary_id,
    SUM(revenue_estimate) AS total_revenue_30d,
    SUM(units_sold) AS total_units_30d
  FROM mv_daily_sales_estimates
  WHERE sale_date >= CURRENT_DATE - INTERVAL '30 days'
  GROUP BY dispensary_id
) sales ON sales.dispensary_id = d.id
WHERE d.crawl_enabled = TRUE
GROUP BY d.id, d.name, d.city, d.state, sales.total_revenue_30d, sales.total_units_30d;

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_store_perf_pk
  ON mv_store_performance(dispensary_id);
CREATE INDEX IF NOT EXISTS idx_mv_store_perf_state
  ON mv_store_performance(state_code);
CREATE INDEX IF NOT EXISTS idx_mv_store_perf_revenue
  ON mv_store_performance(total_revenue_30d DESC);


-- ============================================================
-- VIEW 5: Weekly Category Trends
-- Category performance over time
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_category_weekly_trends AS
SELECT
  dse.category,
  d.state AS state_code,
  DATE_TRUNC('week', dse.sale_date)::DATE AS week_start,
  COUNT(DISTINCT dse.product_id) AS sku_count,
  COUNT(DISTINCT dse.dispensary_id) AS store_count,
  SUM(dse.units_sold) AS total_units,
  SUM(dse.revenue_estimate) AS total_revenue,
  AVG(dse.avg_price) AS avg_price,
  NOW() AS calculated_at
FROM mv_daily_sales_estimates dse
JOIN dispensaries d ON d.id = dse.dispensary_id
WHERE dse.category IS NOT NULL
  AND dse.sale_date >= CURRENT_DATE - INTERVAL '90 days'
GROUP BY dse.category, d.state, DATE_TRUNC('week', dse.sale_date);

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_cat_weekly_pk
  ON mv_category_weekly_trends(category, state_code, week_start);
CREATE INDEX IF NOT EXISTS idx_mv_cat_weekly_state
  ON mv_category_weekly_trends(state_code, week_start);
CREATE INDEX IF NOT EXISTS idx_mv_cat_weekly_date
  ON mv_category_weekly_trends(week_start DESC);


-- ============================================================
-- VIEW 6: Product Intelligence (Hoodie-style per-product metrics)
-- Includes stock diff, days since OOS, days until stockout
-- ============================================================
CREATE MATERIALIZED VIEW IF NOT EXISTS mv_product_intelligence AS
WITH
-- Calculate stock diff over 120 days
stock_diff AS (
  SELECT
    dispensary_id,
    product_id,
    -- Get oldest and newest quantity in last 120 days
    FIRST_VALUE(quantity_available) OVER (
      PARTITION BY dispensary_id, product_id
      ORDER BY captured_at ASC
      ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING
    ) AS qty_120d_ago,
    LAST_VALUE(quantity_available) OVER (
      PARTITION BY dispensary_id, product_id
      ORDER BY captured_at ASC
      ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING
    ) AS qty_current
  FROM inventory_snapshots
  WHERE captured_at >= NOW() - INTERVAL '120 days'
),
stock_diff_calc AS (
  SELECT DISTINCT
    dispensary_id,
    product_id,
    qty_current - COALESCE(qty_120d_ago, qty_current) AS stock_diff_120
  FROM stock_diff
),
-- Get days since last OOS event
last_oos AS (
  SELECT
    dispensary_id,
    product_id,
    MAX(detected_at) AS last_oos_date
  FROM product_visibility_events
  WHERE event_type = 'oos'
  GROUP BY dispensary_id, product_id
),
-- Calculate avg daily units sold (from velocity view)
velocity AS (
  SELECT
    dispensary_id,
    product_id,
    avg_daily_units
  FROM mv_sku_velocity
)
SELECT
  sp.dispensary_id,
  d.name AS dispensary_name,
  d.state AS state_code,
  d.city,
  sp.provider_product_id AS sku,
  sp.name_raw AS product_name,
  sp.brand_name_raw AS brand,
  sp.category_raw AS category,
  sp.is_in_stock,
  sp.stock_status,
  sp.stock_quantity,
  sp.price_rec AS price,
  sp.first_seen_at AS first_seen,
  sp.last_seen_at AS last_seen,
  -- Calculated fields
  COALESCE(sd.stock_diff_120, 0) AS stock_diff_120,
  CASE
    WHEN lo.last_oos_date IS NOT NULL
    THEN EXTRACT(DAY FROM NOW() - lo.last_oos_date)::INT
    ELSE NULL
  END AS days_since_oos,
  -- Days until stockout = current stock / daily burn rate
  CASE
    WHEN v.avg_daily_units > 0 AND sp.stock_quantity > 0
    THEN ROUND(sp.stock_quantity::NUMERIC / v.avg_daily_units)::INT
    ELSE NULL
  END AS days_until_stock_out,
  v.avg_daily_units,
  NOW() AS calculated_at
FROM store_products sp
JOIN dispensaries d ON d.id = sp.dispensary_id
LEFT JOIN stock_diff_calc sd ON sd.dispensary_id = sp.dispensary_id
  AND sd.product_id = sp.provider_product_id
LEFT JOIN last_oos lo ON lo.dispensary_id = sp.dispensary_id
  AND lo.product_id = sp.provider_product_id
LEFT JOIN velocity v ON v.dispensary_id = sp.dispensary_id
  AND v.product_id = sp.provider_product_id
WHERE d.crawl_enabled = TRUE;

CREATE UNIQUE INDEX IF NOT EXISTS idx_mv_prod_intel_pk
  ON mv_product_intelligence(dispensary_id, sku);
CREATE INDEX IF NOT EXISTS idx_mv_prod_intel_brand
  ON mv_product_intelligence(brand);
CREATE INDEX IF NOT EXISTS idx_mv_prod_intel_state
  ON mv_product_intelligence(state_code);
CREATE INDEX IF NOT EXISTS idx_mv_prod_intel_stock_out
  ON mv_product_intelligence(days_until_stock_out ASC NULLS LAST);
CREATE INDEX IF NOT EXISTS idx_mv_prod_intel_oos
  ON mv_product_intelligence(days_since_oos DESC NULLS LAST);


-- ============================================================
-- REFRESH FUNCTION
-- ============================================================
CREATE OR REPLACE FUNCTION refresh_sales_analytics_views()
RETURNS TABLE(view_name TEXT, rows_affected BIGINT) AS $$
DECLARE
  row_count BIGINT;
BEGIN
  -- Must refresh in dependency order:
  -- 1. daily_sales (base view)
  -- 2. sku_velocity (depends on daily_sales)
  -- 3. product_intelligence (depends on sku_velocity)
  -- 4. others (independent)

  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_daily_sales_estimates;
  SELECT COUNT(*) INTO row_count FROM mv_daily_sales_estimates;
  view_name := 'mv_daily_sales_estimates';
  rows_affected := row_count;
  RETURN NEXT;

  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_brand_market_share;
  SELECT COUNT(*) INTO row_count FROM mv_brand_market_share;
  view_name := 'mv_brand_market_share';
  rows_affected := row_count;
  RETURN NEXT;

  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_sku_velocity;
  SELECT COUNT(*) INTO row_count FROM mv_sku_velocity;
  view_name := 'mv_sku_velocity';
  rows_affected := row_count;
  RETURN NEXT;

  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_store_performance;
  SELECT COUNT(*) INTO row_count FROM mv_store_performance;
  view_name := 'mv_store_performance';
  rows_affected := row_count;
  RETURN NEXT;

  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_category_weekly_trends;
  SELECT COUNT(*) INTO row_count FROM mv_category_weekly_trends;
  view_name := 'mv_category_weekly_trends';
  rows_affected := row_count;
  RETURN NEXT;

  -- Product intelligence depends on sku_velocity, so refresh last
  REFRESH MATERIALIZED VIEW CONCURRENTLY mv_product_intelligence;
  SELECT COUNT(*) INTO row_count FROM mv_product_intelligence;
  view_name := 'mv_product_intelligence';
  rows_affected := row_count;
  RETURN NEXT;
END;
$$ LANGUAGE plpgsql;

COMMENT ON FUNCTION refresh_sales_analytics_views IS
  'Refresh all sales analytics materialized views. Call hourly via scheduler.';


-- ============================================================
-- INITIAL REFRESH (populate views)
-- ============================================================
-- Note: Initial refresh must be non-concurrent (no unique index yet populated)
-- Run these manually after migration:
-- REFRESH MATERIALIZED VIEW mv_daily_sales_estimates;
-- REFRESH MATERIALIZED VIEW mv_brand_market_share;
-- REFRESH MATERIALIZED VIEW mv_sku_velocity;
-- REFRESH MATERIALIZED VIEW mv_store_performance;
-- REFRESH MATERIALIZED VIEW mv_category_weekly_trends;