package storage

import (
	"context"
	"fmt"
	"time"

	"git.tornberg.me/go-otel/internal/uaclass"
)

// EndpointStats reads raw trace spans because the production entries table is
// the source of truth and may not have the optional normalized_events table.
// The query intentionally returns both span and trace counts: a reverse proxy
// can legitimately produce nested spans for one request, as with nginx
// ingress -> upstream instrumentation.
func (s *ClickHouseStore) EndpointStats(ctx context.Context, host, service string, hours, limit int) ([]EndpointStat, error) {
	if hours <= 0 {
		hours = 72
	}
	if limit <= 0 || limit > 500 {
		limit = 100
	}

	where := "host != ''"
	rawWhere := "type = 'traces' AND timestamp >= ?"
	args := []interface{}{time.Now().UTC().Add(-time.Duration(hours) * time.Hour)}
	if host != "" {
		where += " AND host = ?"
		args = append(args, host)
		rawWhere += " AND positionCaseInsensitive(data, ?) > 0"
		args = append(args, host)
	}
	if service != "" {
		where += " AND service = ?"
		args = append(args, service)
	}

	query := fmt.Sprintf(`
		SELECT service, host, method, path, if(route_raw = '/' AND path != '/', '', route_raw) AS route, status_code, user_agent,
			origin_host, referer_host, fetch_site, fetch_mode,
			count() AS span_count, uniqExact(trace_id) AS trace_count,
			min(timestamp) AS first_seen, max(timestamp) AS last_seen
		FROM (
			WITH JSONExtractArrayRaw(span, 'attributes') AS attrs
			SELECT
				timestamp,
				JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'service.name', JSONExtractArrayRaw(JSONExtractRaw(resource_span, 'resource'), 'attributes')), 'value', 'stringValue') AS service,
				JSONExtractString(span, 'traceId') AS trace_id,
				upper(coalesce(nullIf(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.request.method', 'http.method'), attrs), 'value', 'stringValue'), ''), '')) AS method,
				replaceRegexpOne(coalesce(nullIf(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'url.path', attrs), 'value', 'stringValue'), ''), JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'http.target', attrs), 'value', 'stringValue')), '[?#].*$', '') AS path,
				JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'http.route', attrs), 'value', 'stringValue') AS route_raw,
				lower(replaceRegexpOne(coalesce(nullIf(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.host', 'server.address', 'net.host.name'), attrs), 'value', 'stringValue'), ''), ''), ':\\d+$', '')) AS host,
				toUInt16OrZero(coalesce(nullIf(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.response.status_code', 'http.status_code'), attrs), 'value', 'intValue'), ''), JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.response.status_code', 'http.status_code'), attrs), 'value', 'stringValue'))) AS status_code,
				JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.user_agent', 'user_agent.original'), attrs), 'value', 'stringValue') AS user_agent,
				replaceRegexpOne(replaceRegexpOne(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.request.header.origin', 'http.origin', 'origin'), attrs), 'value', 'stringValue'), '^https?://', ''), '/.*$', '') AS origin_host,
				replaceRegexpOne(replaceRegexpOne(JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') IN ('http.request.header.referer', 'http.referer'), attrs), 'value', 'stringValue'), '^https?://', ''), '/.*$', '') AS referer_host,
				JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'http.request.header.sec_fetch_site', attrs), 'value', 'stringValue') AS fetch_site,
				JSONExtractString(arrayFirst(x -> JSONExtractString(x, 'key') = 'http.request.header.sec_fetch_mode', attrs), 'value', 'stringValue') AS fetch_mode
			FROM (
				SELECT timestamp, span
				FROM %s.entries
				ARRAY JOIN JSONExtractArrayRaw(data, 'resourceSpans') AS resource_span
				ARRAY JOIN JSONExtractArrayRaw(resource_span, 'scopeSpans') AS scope_span
				ARRAY JOIN JSONExtractArrayRaw(scope_span, 'spans') AS span
				WHERE %s
			)
		) AS spans
		WHERE %s
		GROUP BY service, host, method, path, route_raw, status_code, user_agent, origin_host, referer_host, fetch_site, fetch_mode
		ORDER BY trace_count DESC, span_count DESC, last_seen DESC
		LIMIT ?`, s.db, rawWhere, where)
	args = append(args, limit)

	rows, err := s.conn.Query(ctx, query, args...)
	if err != nil {
		return nil, fmt.Errorf("clickhouse endpoint stats: %w", err)
	}
	defer rows.Close()

	var out []EndpointStat
	for rows.Next() {
		var stat EndpointStat
		if err := rows.Scan(
			&stat.Service, &stat.Host, &stat.Method, &stat.Path, &stat.Route, &stat.StatusCode,
			&stat.UserAgent, &stat.OriginHost, &stat.RefererHost, &stat.FetchSite,
			&stat.FetchMode, &stat.SpanCount, &stat.TraceCount, &stat.FirstSeen, &stat.LastSeen,
		); err != nil {
			return nil, err
		}
		if stat.UserAgent != "" {
			stat.UAClass = uaclass.Of(stat.UserAgent)
		}
		if stat.Route == "/" && stat.Path != "/" {
			stat.Route = ""
		}
		out = append(out, stat)
	}
	return out, rows.Err()
}