tm_back/internal/dashboard/statistics_repository.go

package dashboard

import (
	"context"
	"fmt"
	"strings"
	"time"

	"tm/internal/notice"
	"tm/pkg/ai_summarizer"
	orm "tm/pkg/mongo"

	"go.mongodb.org/mongo-driver/v2/bson"
	mongodriver "go.mongodb.org/mongo-driver/v2/mongo"
	"golang.org/x/sync/errgroup"
)

const noticesCollectionName = "notices"

// scrapedScopeCacheTTL avoids listing every MinIO object on each statistics request while
// keeping counts aligned with the tender panel (MinIO is the source of truth for scraped docs).
const scrapedScopeCacheTTL = 5 * time.Minute

const (
	scrapedScopeResolveTimeout = 2 * time.Minute
	statisticsQueryTimeout     = 2 * time.Minute
	maxScrapedFolderInClause   = 5000
)

// scrapedTendersScope is the resolved Mongo filter for tenders with scraped documents.
// zero is set when MinIO reports no procedure folders; callers must return zero without querying.
type scrapedTendersScope struct {
	match       bson.M
	zero        bool
	fromMinIO   bool
	dailyCounts map[string]int64 // non-nil when scope was resolved from MinIO
}

func (r *repository) Statistics(ctx context.Context, days int, startUnix int64) (*StatisticsReportResponse, error) {
	queryCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), statisticsQueryTimeout)
	defer cancel()

	now := time.Now().UTC()
	endDay := time.Date(now.Year(), now.Month(), now.Day(), 0, 0, 0, 0, time.UTC)
	startDay := endDay.AddDate(0, 0, -(days - 1))

	// Resolve scraped-document scope from MinIO (cached); falls back to Mongo flags on error.
	scrapedScope, err := r.cachedScrapedTendersScope(queryCtx)
	if err != nil {
		return nil, err
	}

	var (
		scrapedTED        map[string]int64
		scrapedDocuments  map[string]int64
		translatedNotices map[string]int64
		totalDocuments    int64
		totalTranslated   int64
	)

	g, gctx := errgroup.WithContext(queryCtx)

	g.Go(func() error {
		// Use created_at (first ingest time). processing_metadata.scraped_at is reset on
		// notice refresh merges, which would collapse historical scrapes onto the latest day.
		counts, err := r.dailyCountByTimestamp(gctx, noticesCollectionName, bson.M{
			"source":     notice.TenderSourceTEDScraper,
			"created_at": bson.M{"$gte": startUnix, "$gt": 0},
		}, "created_at")
		if err != nil {
			return fmt.Errorf("scraped ted per day: %w", err)
		}
		scrapedTED = counts
		return nil
	})

	g.Go(func() error {
		counts, err := r.scrapedDocumentsPerDay(gctx, startUnix, scrapedScope)
		if err != nil {
			return fmt.Errorf("scraped documents per day: %w", err)
		}
		scrapedDocuments = counts
		return nil
	})

	g.Go(func() error {
		counts, err := r.translatedNoticesPerDay(gctx, startDay, endDay)
		if err != nil {
			return fmt.Errorf("translated notices per day: %w", err)
		}
		translatedNotices = counts
		return nil
	})

	g.Go(func() error {
		total, err := r.totalScrapedTenders(gctx, scrapedScope)
		if err != nil {
			return fmt.Errorf("total scraped tenders: %w", err)
		}
		totalDocuments = total
		return nil
	})

	g.Go(func() error {
		total, err := r.metricsCounter.Get(gctx, orm.AITranslationSuccessCounterKey)
		if err != nil {
			return fmt.Errorf("total translated tenders: %w", err)
		}
		totalTranslated = total
		return nil
	})

	if err := g.Wait(); err != nil {
		return nil, err
	}

	return &StatisticsReportResponse{
		Days:        days,
		GeneratedAt: now.Unix(),
		Daily: StatisticsDailySeries{
			ScrapedTED:        fillTrendSeries(days, scrapedTED),
			ScrapedDocuments:  fillTrendSeries(days, scrapedDocuments),
			TranslatedNotices: fillTrendSeries(days, translatedNotices),
		},
		Totals: StatisticsLifetimeTotals{
			ScrapedDocuments:  totalDocuments,
			TranslatedTenders: totalTranslated,
		},
	}, nil
}

func (r *repository) dailyCountByTimestamp(ctx context.Context, collection string, match bson.M, timestampField string) (map[string]int64, error) {
	pipeline := mongodriver.Pipeline{
		{{Key: "$match", Value: match}},
		{{Key: "$addFields", Value: bson.M{
			"metric_ts": normalizeTimestampExpr(timestampField),
		}}},
		{{Key: "$group", Value: bson.M{
			"_id": bson.M{
				"$dateToString": bson.M{
					"format":   "%Y-%m-%d",
					"date":     bson.M{"$toDate": bson.M{"$multiply": bson.A{"$metric_ts", 1000}}},
					"timezone": "UTC",
				},
			},
			"count": bson.M{"$sum": 1},
		}}},
	}

	cursor, err := r.mongoManager.GetCollection(collection).Aggregate(ctx, pipeline)
	if err != nil {
		return nil, err
	}
	defer cursor.Close(ctx)

	return decodeDailyCounts(ctx, cursor)
}

func (r *repository) scrapedDocumentsPerDay(ctx context.Context, startUnix int64, scope scrapedTendersScope) (map[string]int64, error) {
	if scope.zero {
		return map[string]int64{}, nil
	}

	if scope.dailyCounts != nil {
		return filterDailyCountsSince(scope.dailyCounts, startUnix), nil
	}

	return r.scrapedDocumentsPerDayFromMongo(ctx, startUnix, scope.match)
}

func filterDailyCountsSince(counts map[string]int64, startUnix int64) map[string]int64 {
	if len(counts) == 0 {
		return map[string]int64{}
	}
	startDay := time.Unix(startUnix, 0).UTC().Format("2006-01-02")
	filtered := make(map[string]int64)
	for date, count := range counts {
		if date >= startDay {
			filtered[date] = count
		}
	}
	return filtered
}

func (r *repository) scrapedDocumentsPerDayFromMongo(ctx context.Context, startUnix int64, match bson.M) (map[string]int64, error) {
	pipeline := mongodriver.Pipeline{
		{{Key: "$match", Value: match}},
		{{Key: "$unwind", Value: "$scraped_documents"}},
		{{Key: "$addFields", Value: bson.M{
			"metric_ts": bson.M{
				"$cond": bson.A{
					bson.M{"$gt": bson.A{"$scraped_documents.scraped_at", 0}},
					"$scraped_documents.scraped_at",
					"$processing_metadata.documents_scraped_at",
				},
			},
		}}},
		{{Key: "$addFields", Value: bson.M{
			"metric_ts": normalizeTimestampExpr("metric_ts"),
		}}},
		{{Key: "$match", Value: bson.M{
			"metric_ts": bson.M{"$gte": startUnix, "$gt": 0},
		}}},
		{{Key: "$group", Value: bson.M{
			"_id": bson.M{
				"$dateToString": bson.M{
					"format":   "%Y-%m-%d",
					"date":     bson.M{"$toDate": bson.M{"$multiply": bson.A{"$metric_ts", 1000}}},
					"timezone": "UTC",
				},
			},
			"count": bson.M{"$sum": 1},
		}}},
	}

	cursor, err := r.mongoManager.GetCollection(tenderCollectionName()).Aggregate(ctx, pipeline)
	if err != nil {
		return nil, err
	}
	defer cursor.Close(ctx)

	return decodeDailyCounts(ctx, cursor)
}

func (r *repository) translatedNoticesPerDay(ctx context.Context, startDay, endDay time.Time) (map[string]int64, error) {
	return r.metricsCounter.GetDailyCounts(ctx, startDay, endDay)
}

func (r *repository) totalScrapedTenders(ctx context.Context, scope scrapedTendersScope) (int64, error) {
	if scope.zero {
		return 0, nil
	}

	return r.ormRepo.Count(ctx, scope.match)
}

func mongoScrapedTendersScope() scrapedTendersScope {
	return scrapedTendersScope{
		match: bson.M{
			"processing_metadata.documents_scraped": true,
			"scraped_documents":                   bson.M{"$exists": true, "$ne": bson.A{}},
		},
	}
}

func (r *repository) cachedScrapedTendersScope(ctx context.Context) (scrapedTendersScope, error) {
	now := time.Now()

	r.scrapedScopeMu.Lock()
	if r.scrapedScopeCached && now.Before(r.scrapedScopeExpiry) {
		scope := r.scrapedScope
		r.scrapedScopeMu.Unlock()
		return scope, nil
	}
	r.scrapedScopeMu.Unlock()

	v, err, _ := r.scrapedScopeGroup.Do("scraped-scope", func() (interface{}, error) {
		r.scrapedScopeMu.Lock()
		if r.scrapedScopeCached && time.Now().Before(r.scrapedScopeExpiry) {
			scope := r.scrapedScope
			r.scrapedScopeMu.Unlock()
			return scope, nil
		}
		r.scrapedScopeMu.Unlock()

		resolveCtx, cancel := context.WithTimeout(context.Background(), scrapedScopeResolveTimeout)
		defer cancel()

		scope, resolveErr := r.resolveScrapedTendersScope(resolveCtx)
		if resolveErr != nil {
			r.logger.Warn("MinIO scraped scope unavailable, using Mongo fallback for dashboard statistics", map[string]interface{}{
				"error": resolveErr.Error(),
			})
			return mongoScrapedTendersScope(), nil
		}

		r.scrapedScopeMu.Lock()
		r.scrapedScope = scope
		r.scrapedScopeExpiry = time.Now().Add(scrapedScopeCacheTTL)
		r.scrapedScopeCached = true
		r.scrapedScopeMu.Unlock()

		return scope, nil
	})
	if err != nil {
		return scrapedTendersScope{}, err
	}

	return v.(scrapedTendersScope), nil
}

func (r *repository) resolveScrapedTendersScope(ctx context.Context) (scrapedTendersScope, error) {
	procedures, dailyCounts, err := r.listScrapedProcedures(ctx)
	if err != nil {
		return scrapedTendersScope{}, err
	}

	folderIDs := folderIDsFromProcedures(procedures)
	if len(folderIDs) > 0 {
		if len(folderIDs) > maxScrapedFolderInClause {
			r.logger.Warn("Too many MinIO procedure folders for dashboard statistics, using Mongo fallback", map[string]interface{}{
				"folder_count":  len(folderIDs),
				"max_in_clause": maxScrapedFolderInClause,
			})
			return mongoScrapedTendersScope(), nil
		}
		return scrapedTendersScope{
			match:       bson.M{"contract_folder_id": bson.M{"$in": folderIDs}},
			fromMinIO:   true,
			dailyCounts: dailyCounts,
		}, nil
	}
	return mongoScrapedTendersScope(), nil
}

func (r *repository) listScrapedProcedures(ctx context.Context) ([]ai_summarizer.ProcedureDocumentsSummary, map[string]int64, error) {
	if r.procedureLister == nil {
		return nil, nil, nil
	}

	if scanner, ok := r.procedureLister.(scrapedDocumentsScanner); ok {
		return scanner.ScanScrapedDocuments(ctx)
	}

	procedures, err := r.procedureLister.ListProceduresWithDocuments(ctx)
	if err != nil {
		return nil, nil, err
	}
	return procedures, nil, nil
}

func folderIDsFromProcedures(procedures []ai_summarizer.ProcedureDocumentsSummary) []string {
	folderIDs := make([]string, 0, len(procedures))
	seen := make(map[string]struct{}, len(procedures))
	for _, proc := range procedures {
		if proc.DocumentCount <= 0 {
			continue
		}
		id := strings.TrimSpace(proc.ContractFolderID)
		if id == "" {
			continue
		}
		if _, ok := seen[id]; ok {
			continue
		}
		seen[id] = struct{}{}
		folderIDs = append(folderIDs, id)
	}
	return folderIDs
}

func (r *repository) scrapedDocumentsFolderIDs(ctx context.Context) ([]string, error) {
	if r.procedureLister == nil {
		return nil, nil
	}

	procedures, _, err := r.listScrapedProcedures(ctx)
	if err != nil {
		return nil, err
	}
	return folderIDsFromProcedures(procedures), nil
}

func decodeDailyCounts(ctx context.Context, cursor *mongodriver.Cursor) (map[string]int64, error) {
	counts := make(map[string]int64)
	for cursor.Next(ctx) {
		var row bson.M
		if err := cursor.Decode(&row); err != nil {
			return nil, err
		}
		date, ok := row["_id"].(string)
		if !ok || date == "" {
			continue
		}
		counts[date] = aggregateCount(row, "count")
	}
	if err := cursor.Err(); err != nil {
		return nil, err
	}
	return counts, nil
}