Crussell/backend/handlers/user/account.go

package user

import (
	"context"
	"crypto/sha256"
	"database/sql"
	"encoding/hex"
	"errors"
	"fmt"
	"log"
	"log/slog"
	"net/http"
	"net/url"
	"os"
	"time"

	"crussell/db"
	"crussell/handlers/payments"
	"crussell/internal/dav"
	"crussell/internal/s3"
	"crussell/internal/square"
	"crussell/mw"

	"github.com/jackc/pgx/v5"
)

// --- Square erasure retry + alerting + durable outbox (GDPR H3 / A1) ---
//
// The deletion goroutine in DeleteAccountHandler retries transient Square
// failures and, on final failure, logs at ERROR and raises a critical admin
// notification (reason 'critical_payment_log'). A failed Square-side erasure
// must never be silently dropped: by the time the goroutine runs, the local
// anonymization has already committed and NULLed square_card_id/
// square_customer_id, so the external Square references would otherwise be
// orphaned — permanently, with no way to retrace them.
//
// Fault A1 (crash window): the erasure targets are ALSO persisted as a durable
// outbox INSIDE the anonymization transaction (persistSquareErasureOutbox),
// before it commits. The just-scrubbed, soft-deleted user_saved_cards rows keep
// their square_card_id / square_customer_id (and lose their user_id, matching
// delete_guest_user) until the Square-side erasure completes. If the process
// dies between the local commit and the async goroutine finishing, those rows
// are the only remaining record of the external Square PII; the
// retry-square-erasures job (internal/jobs/cleanup.go) finds them and retries
// the deletion. The goroutine drains the outbox (NULLing the refs on success);
// the job is the crash safety net. Either the local tx rolls back (nothing
// NULLed) or the durable queue guarantees the Square deletion is eventually
// attempted and alerted on failure.

const (
	// squareDeleteMaxAttempts is the number of times a Square erasure call is
	// attempted before it is treated as failed (1 initial + 2 retries).
	squareDeleteMaxAttempts = 3
	// squareDeleteAttemptTimeout bounds a single Square erasure attempt.
	squareDeleteAttemptTimeout = 30 * time.Second
	// squareDeleteBackoffBase is the delay before the first retry (doubled per
	// subsequent retry).
	squareDeleteBackoffBase = 500 * time.Millisecond
)

// squareErasureTarget records one saved-card row's external Square references
// (row id + square_card_id + square_customer_id) captured BEFORE anonymization
// NULLs them, so the durable outbox and the async cleanup goroutine still have
// the external refs they need to complete Square-side erasure (Fault A1).
type squareErasureTarget struct {
	rowID      string
	cardID     string // empty when the row had no Square card on file
	customerID string // empty when the row had no Square customer profile
}

// squareDeletionRetryable reports whether a Square erasure error is transient
// and worth retrying: transport errors (wrapped *url.Error) and Square 5xx
// responses. Definitive 4xx rejections and not-found no-ops are NOT retried.
func squareDeletionRetryable(err error) bool {
	if square.IsNotFound(err) {
		return false
	}
	if sc := square.ErrorStatusCode(err); sc != 0 {
		return sc >= http.StatusInternalServerError
	}
	var urlErr *url.Error
	return errors.As(err, &urlErr)
}

// RetrySquareDeletion runs fn (a DeleteCardOnFile/DeleteCustomer call) up to
// squareDeleteMaxAttempts times, backing off between retries, retrying only
// transient failures (see squareDeletionRetryable). Returns the final error,
// nil on success. Exported for the retry-square-erasures job
// (internal/jobs/cleanup.go), which reuses the same retry budget as the
// account-deletion goroutine.
func RetrySquareDeletion(parent context.Context, fn func(context.Context) error) error {
	var lastErr error
	backoff := squareDeleteBackoffBase
	for attempt := 1; attempt <= squareDeleteMaxAttempts; attempt++ {
		if attempt > 1 {
			select {
			case <-time.After(backoff):
			case <-parent.Done():
				return lastErr
			}
			backoff *= 2
		}
		attemptCtx, cancel := context.WithTimeout(parent, squareDeleteAttemptTimeout)
		err := fn(attemptCtx)
		cancel()
		if err == nil {
			return nil
		}
		lastErr = err
		if !squareDeletionRetryable(err) {
			return err
		}
	}
	return lastErr
}

// squareErasureNotificationID derives a deterministic admin_notifications id
// for a failed Square-side erasure tied to key ("S" + 11 hex chars of a
// SHA-256 digest, mirroring the webhooks dispute-notification id scheme). For
// the account-deletion path key is the affected user id; for the
// retry-square-erasures job key is the user id when the outbox row still
// carries one, otherwise the outbox row id. One notification per key;
// re-delivery of the same failure is a no-op.
func squareErasureNotificationID(key string) string {
	sum := sha256.Sum256([]byte("square-erasure-failure:" + key))
	return "S" + hex.EncodeToString(sum[:])[:11]
}

// InsertSquareErasureCriticalNotification surfaces a failed Square-side
// erasure in the admin notification centre (reason 'critical_payment_log' —
// the DB-backed stand-in for CRITICAL logs that the payments sweep uses).
// user_id is deliberately NULL: by the time the async cleanup runs, the account
// may already be deleted (guests) or anonymized, and the deterministic id keeps
// exactly one notification per affected key. Exported for the
// retry-square-erasures job, which reuses the account-deletion alert scheme.
func InsertSquareErasureCriticalNotification(ctx context.Context, key string) {
	// Fresh bounded context so a near-expiry deletion context cannot suppress
	// the admin alert.
	actx, cancel := context.WithTimeout(ctx, 10*time.Second)
	defer cancel()
	tag, err := db.Conn.Exec(actx, `
		INSERT INTO admin_notifications (id, reason, booking_id, user_id, created_at)
		VALUES ($1, 'critical_payment_log'::admin_notification_reason, NULL, NULL, NOW())
		ON CONFLICT (id) DO NOTHING
	`, squareErasureNotificationID(key))
	if err != nil {
		slog.Error("failed to insert critical notification for failed Square erasure", "key", key, "err", err)
		return
	}
	if int(tag.RowsAffected()) > 0 {
		slog.Error("Square-side erasure FAILED after retries — critical notification raised", "key", key)
	}
}

// persistSquareErasureOutbox writes the captured Square erasure targets back
// onto the just-scrubbed, soft-deleted user_saved_cards rows INSIDE the
// anonymization transaction, BEFORE it commits (Fault A1). anonymize_user() /
// delete_guest_user() NULL square_card_id/square_customer_id for GDPR
// scrubbing; this UPDATE restores them on the soft-deleted rows so the erasure
// targets survive the commit as a durable queue entry. user_id is also NULLed
// (delete_guest_user already does this for guests): payments.EnsureSquareCustomer
// reads square_customer_id across ALL of a user's rows regardless of
// deleted_at, so an outbox row that kept its user_id would let the erased
// identity's deleted Square customer resurface on a later save-card flow. The
// rows remain in place for the 7-year financial retention; only the user link
// is dropped. If the process crashes before the async Square cleanup finishes,
// the retry-square-erasures job (internal/jobs/cleanup.go) finds these rows and
// completes the external deletion — the erasure is never permanently lost.
func persistSquareErasureOutbox(ctx context.Context, tx pgx.Tx, targets []squareErasureTarget) error {
	for _, t := range targets {
		var cardID, customerID any
		if t.cardID != "" {
			cardID = t.cardID
		}
		if t.customerID != "" {
			customerID = t.customerID
		}
		if _, err := tx.Exec(ctx, `
			UPDATE user_saved_cards
			SET square_card_id = $2, square_customer_id = $3, user_id = NULL
			WHERE id = $1 AND deleted_at IS NOT NULL
		`, t.rowID, cardID, customerID); err != nil {
			return err
		}
	}
	return nil
}

// clearSquareErasureOutboxRows NULLs the durable outbox marker on the given
// soft-deleted user_saved_cards rows after their Square-side erasure is
// complete (or deliberately abandoned as still-referenced). Best-effort: the
// retry-square-erasures job re-covers any row left behind by a failure.
func clearSquareErasureOutboxRows(ctx context.Context, rowIDs []string) {
	for _, rowID := range rowIDs {
		if _, err := db.Conn.Exec(ctx, `
			UPDATE user_saved_cards SET square_customer_id = NULL
			WHERE id = $1 AND deleted_at IS NOT NULL
		`, rowID); err != nil {
			slog.Error("failed to clear Square erasure outbox for customer", "row", rowID, "error", err)
		}
	}
}

// DELETE /api/user/account
func DeleteAccountHandler(w http.ResponseWriter, r *http.Request) {
	userID, ok := mw.GetUserID(r.Context())
	if !ok {
		http.Error(w, "unauthorized", http.StatusUnauthorized)
		return
	}

	var accountRole string
	var profilePicURL sql.NullString
	err := db.Conn.QueryRow(r.Context(), `SELECT account_role, profile_pic_url FROM users WHERE id = $1`, userID).
		Scan(&accountRole, &profilePicURL)
	if err != nil {
		if errors.Is(err, pgx.ErrNoRows) {
			http.Error(w, "user not found", http.StatusNotFound)
			return
		}
		log.Printf("Failed to fetch user for deletion: %v", err)
		http.Error(w, "server error", http.StatusInternalServerError)
		return
	}

	ctx := r.Context()

	// --- External system scrubbing (BEFORE SQL anonymize) ---

	// Delete profile picture from S3/R2
	if profilePicURL.Valid && profilePicURL.String != "" && s3.Client != nil {
		// #nosec G118 — intentional background goroutine for async profile pic cleanup
		go func(picURL string) {
			defer func() {
				if r := recover(); r != nil {
					log.Printf("Panic recovered in S3 profile picture deletion: %v", r)
				}
			}()
			ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
			defer cancel()
			bucket := os.Getenv("S3_PROFILE_PICS_BUCKET")
			if bucket == "" {
				bucket = "crussell-profile-pics"
			}
			// profiles/{userID}.jpg — matches UploadProfilePictureHandler key format
			key := fmt.Sprintf("profiles/%s.jpg", userID)
			if err := s3.Client.Delete(ctx, bucket, key); err != nil {
				log.Printf("Warning: Failed to delete profile picture for user %s: %v", userID, err)
			}
		}(profilePicURL.String)
	}

	// Snapshot the Square card AND customer references per saved-card row
	// synchronously BEFORE the SQL anonymization below NULLs
	// square_card_id/square_customer_id, so both the durable outbox and the
	// background cleanup still have the external Square references they need
	// (previously the goroutine read the rows itself, racing the anonymize step
	// which wiped them mid-flight). Distinct non-null customer IDs only: a
	// user's saved cards share one provisioned Square customer, so
	// DeleteCustomer runs once per customer. NULL customer IDs (users with no
	// saved cards) are skipped.
	var erasureTargets []squareErasureTarget
	needsSquareErasure := false
	// Capture the client synchronously so the async cleanup goroutine never
	// reads the global payments.SquareClient (which tests swap per-account).
	sqClient := payments.SquareClient
	if sqClient != nil {
		rows, err := db.Conn.Query(r.Context(),
			`SELECT id, square_card_id, square_customer_id FROM user_saved_cards WHERE user_id = $1 AND deleted_at IS NULL`, userID)
		if err != nil {
			// Fail-closed: anonymization must never proceed without the
			// external Square refs the cleanup needs — otherwise the card and
			// customer would be permanently orphaned at Square with the local
			// ids already NULLed (GDPR erasure completeness).
			log.Printf("Failed to query saved cards for user %s: %v", userID, err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}
		for rows.Next() {
			var t squareErasureTarget
			var cardID, customerID sql.NullString
			if err := rows.Scan(&t.rowID, &cardID, &customerID); err != nil {
				log.Printf("Warning: Failed to scan card ID for user %s: %v", userID, err)
				continue
			}
			if cardID.Valid && cardID.String != "" {
				t.cardID = cardID.String
			}
			if customerID.Valid && customerID.String != "" {
				t.customerID = customerID.String
			}
			if t.cardID != "" || t.customerID != "" {
				needsSquareErasure = true
			}
			erasureTargets = append(erasureTargets, t)
		}
		if err := rows.Err(); err != nil {
			log.Printf("Failed to iterate saved cards for user %s: %v", userID, err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}
		rows.Close()
	}

	// --- SQL-level anonymization/deletion ---

	if accountRole == "guest" {
		tx, err := db.Conn.Begin(ctx)
		if err != nil {
			log.Printf("Failed to begin transaction for guest user deletion: %v", err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}
		defer func() {
			if err := tx.Rollback(ctx); err != nil && !errors.Is(err, pgx.ErrTxClosed) {
				slog.Error("failed to rollback transaction", "err", err)
			}
		}()

		_, err = tx.Exec(ctx, `SELECT delete_guest_user($1)`, userID)
		if err != nil {
			log.Printf("Failed to delete guest user %s: %v", userID, err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}

		// A1 durable outbox: persist the Square erasure targets on the
		// just-scrubbed, soft-deleted rows BEFORE this tx commits — if the
		// process dies after the commit, the retry-square-erasures job still
		// finds them.
		if sqClient != nil && needsSquareErasure {
			if err := persistSquareErasureOutbox(ctx, tx, erasureTargets); err != nil {
				log.Printf("Failed to persist Square erasure outbox for guest %s: %v", userID, err)
				http.Error(w, "server error", http.StatusInternalServerError)
				return
			}
		}

		if err := tx.Commit(ctx); err != nil {
			log.Printf("Failed to commit transaction for guest user deletion: %v", err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}
	} else {
		tx, err := db.Conn.Begin(ctx)
		if err != nil {
			log.Printf("Failed to begin transaction for user anonymization: %v", err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}
		defer func() {
			if err := tx.Rollback(ctx); err != nil && !errors.Is(err, pgx.ErrTxClosed) {
				slog.Error("failed to rollback transaction", "err", err)
			}
		}()

		// anonymize_user() is the single source of truth for erasure: it NULLs
		// the 2FA columns and staff notes in the same statement that anonymizes
		// the rest of the row, so every call site (this handler AND the
		// idle-account batch cleanup in scheduling.CleanupIdleAccounts) is
		// GDPR-clean without a separate Go-side scrub.
		_, err = tx.Exec(ctx, `SELECT anonymize_user($1)`, userID)
		if err != nil {
			log.Printf("Failed to anonymize user %s: %v", userID, err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}

		// A1 durable outbox: persist the Square erasure targets on the
		// just-scrubbed, soft-deleted rows BEFORE this tx commits — if the
		// process dies after the commit, the retry-square-erasures job still
		// finds them.
		if sqClient != nil && needsSquareErasure {
			if err := persistSquareErasureOutbox(ctx, tx, erasureTargets); err != nil {
				log.Printf("Failed to persist Square erasure outbox for user %s: %v", userID, err)
				http.Error(w, "server error", http.StatusInternalServerError)
				return
			}
		}

		if err := tx.Commit(ctx); err != nil {
			log.Printf("Failed to commit transaction for user anonymization: %v", err)
			http.Error(w, "server error", http.StatusInternalServerError)
			return
		}

		// TODO: Create 'user_anonymized' notification for admin audit trail
	}

	// The local erasure committed: drop the user's cached Square customer id so
	// the erased identity cannot resurface from the process-local cache on a
	// later save-card flow (the Square customer is deleted below and the DB
	// columns are NULLed, but the cache is never touched by either).
	payments.InvalidateSquareCustomerCache(userID)

	// Build the context used for critical-notification inserts: route through
	// the request transaction when one is active (tests) so alerts roll back
	// with the fixture; otherwise fall back to the shared pool (production,
	// where the request context is cancelled once this handler returns).
	// Captured synchronously for the goroutine below.
	notifyCtx := context.Background()
	if activeTx := db.TxFromContext(r.Context()); activeTx != nil {
		notifyCtx = db.ContextWithTx(notifyCtx, activeTx)
	}

	// external Square cleanup fires only after local anonymization/deletion
	// commits, so a failed local tx leaves external state intact for retry. The
	// durable outbox persisted above guarantees the targets survive even a
	// process crash before this goroutine finishes — this goroutine is the
	// primary drain, and the retry-square-erasures job is the safety net.
	if sqClient != nil && needsSquareErasure {
		// #nosec G118 — intentional background goroutine for async account deletion
		go func(client square.SquareClient, targets []squareErasureTarget, notifyCtx context.Context) {
			defer func() {
				if r := recover(); r != nil {
					log.Printf("Panic recovered in Square cleanup: %v", r)
				}
			}()
			ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
			defer cancel()

			cardRow := map[string]string{}
			for _, t := range targets {
				if t.cardID != "" {
					cardRow[t.cardID] = t.rowID
				}
			}
			for cardID, rowID := range cardRow {
				if err := RetrySquareDeletion(ctx, func(actx context.Context) error {
					return client.DeleteCardOnFile(actx, cardID)
				}); err != nil {
					// TokenPrefix redacts the ccof: card token — the full ID must
					// never reach logs. A failed erasure is logged at ERROR and
					// raised as a critical admin notification — never silently
					// dropped after the local refs are NULLed.
					log.Printf("Error: Failed to delete Square card %s for user %s after %d attempts: %v", square.TokenPrefix(cardID), userID, squareDeleteMaxAttempts, err)
					slog.Error("square card deletion failed after retries", "user_id", userID, "card", square.TokenPrefix(cardID), "error", err)
					InsertSquareErasureCriticalNotification(notifyCtx, userID)
					continue
				}
				// Drain the durable outbox so the safety-net job has nothing to
				// retry. Best-effort: a failed clear leaves the entry in place
				// for the job, which treats NOT_FOUND as complete.
				if _, err := db.Conn.Exec(ctx, `
					UPDATE user_saved_cards SET square_card_id = NULL
					WHERE id = $1 AND deleted_at IS NOT NULL
				`, rowID); err != nil {
					slog.Error("failed to clear Square erasure outbox for card", "user_id", userID, "card", square.TokenPrefix(cardID), "error", err)
				}
			}

			customerRows := map[string][]string{}
			for _, t := range targets {
				if t.customerID != "" {
					customerRows[t.customerID] = append(customerRows[t.customerID], t.rowID)
				}
			}
			for customerID, rows := range customerRows {
				// Square customers are provisioned per-user from a deterministic
				// email-derived key, but the UNIQUE(email) index excludes guest
				// accounts — so a guest and a registered user sharing an email can
				// end up on the SAME Square customer profile (Square dedups within
				// its idempotency window). Deleting it would break the other
				// account's saved-card charges, so skip the deletion when any OTHER
				// non-deleted saved card still references the customer.
				var stillReferenced bool
				if err := db.Conn.QueryRow(ctx, `
					SELECT EXISTS(SELECT 1 FROM user_saved_cards WHERE square_customer_id = $1 AND user_id <> $2 AND deleted_at IS NULL)
				`, customerID, userID).Scan(&stillReferenced); err != nil {
					log.Printf("Error: Failed to check Square customer %s references before deletion: %v", square.TokenPrefix(customerID), err)
					slog.Error("failed to check Square customer references before deletion", "user_id", userID, "customer", square.TokenPrefix(customerID), "error", err)
					InsertSquareErasureCriticalNotification(notifyCtx, userID)
					continue
				}
				if stillReferenced {
					// PII-redacted customer id — the full id never reaches logs.
					log.Printf("Warning: skipping Square customer deletion — customer %s still referenced by another account", square.TokenPrefix(customerID))
					// The customer is deliberately kept (still referenced by
					// another account) — not a pending erasure, so drain the
					// outbox rows and let the job stop retrying it.
					clearSquareErasureOutboxRows(ctx, rows)
					continue
				}
				if err := RetrySquareDeletion(ctx, func(actx context.Context) error {
					return client.DeleteCustomer(actx, customerID)
				}); err != nil {
					log.Printf("Error: Failed to delete Square customer %s for user %s after %d attempts: %v", square.TokenPrefix(customerID), userID, squareDeleteMaxAttempts, err)
					slog.Error("square customer deletion failed after retries", "user_id", userID, "customer", square.TokenPrefix(customerID), "error", err)
					InsertSquareErasureCriticalNotification(notifyCtx, userID)
					continue
				}
				clearSquareErasureOutboxRows(ctx, rows)
			}
		}(sqClient, erasureTargets, notifyCtx)
	}

	// Delete CardDAV contact (non-blocking, best-effort)
	if dav.Service != nil {
		go func() {
			defer func() {
				if r := recover(); r != nil {
					log.Printf("Panic recovered in CardDAV contact deletion: %v", r)
				}
			}()
			uri := fmt.Sprintf("%s.vcf", userID)
			if err := dav.Service.DeleteContact(1, uri); err != nil {
				log.Printf("Warning: Failed to delete CardDAV contact for user %s: %v", userID, err)
			}
		}()
	}

	w.WriteHeader(http.StatusNoContent)
}