toom1996/backend_v2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(publish): 入库实体键读错误 fail-closed，复用加 rejected 守卫，去重排除软删行

Browse Source
This commit is contained in:
toom1996
2026-09-23 10:41:16 +08:00
parent dba615d636
commit 8f8b06def7
3 changed files with 117 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
internal/repository/ingest_repository.go
View File

@ -3,6 +3,7 @@ package repository
import (
"context"
"errors"
"fmt"
"math"
"time"
@ -290,7 +291,11 @@ func (r *ingestRepository) CreateRunwayWithImages(ctx context.Context, rw *model
func (r *ingestRepository) ReuseRejectedRunway(ctx context.Context, id uint32, rw *model.BrandRunway, imgs []model.BrandRunwayImage) error {
now := uint32(time.Now().Unix())
return r.db.WithContext(ctx).Transaction(func(tx *gorm.DB) error {
if uErr := tx.Model(&model.BrandRunway{}).Where("id = ?", id).Updates(map[string]any{
// WHERE 带 status=rejected 守卫：状态是读阶段拿到的，而写库发生在整批图片下载/上传之后
// （可能数十秒窗口）。若无条件覆盖，期间已被改成 published 的行会被置回 pending，
// 等于把已发布内容从公开视图上撤下。
upd := tx.Model(&model.BrandRunway{}).Where("id = ? AND status = ?", id, model.StatusRejected)
if uErr := upd.Updates(map[string]any{
"title_en": rw.TitleEn,
"title_cn": rw.TitleCn,
"description_en": rw.DescriptionEn,
@ -309,6 +314,9 @@ func (r *ingestRepository) ReuseRejectedRunway(ctx context.Context, id uint32, r
}).Error; uErr != nil {
return uErr
}
if upd.RowsAffected == 0 {
return fmt.Errorf("runway %d 已不是 rejected 状态，复用前置条件失效", id)
}
if dErr := tx.Model(&model.BrandRunwayImage{}).
Where("runway_id = ? AND is_deleted = 0", id).
Updates(map[string]any{"is_deleted": 1, "updated_at": now}).Error; dErr != nil {
@ -373,7 +381,9 @@ func (r *ingestRepository) CreateStreetSnapWithImages(ctx context.Context, snap
func (r *ingestRepository) ReuseRejectedStreetSnap(ctx context.Context, id uint32, snap *model.StreetSnap, imgs []model.StreetSnapImage) error {
now := uint32(time.Now().Unix())
return r.db.WithContext(ctx).Transaction(func(tx *gorm.DB) error {
if uErr := tx.Model(&model.StreetSnap{}).Where("id = ?", id).Updates(map[string]any{
// 同 runway：带 status=rejected 守卫 + 行数校验，避免把期间已发布的记录撤下公开视图。
upd := tx.Model(&model.StreetSnap{}).Where("id = ? AND status = ?", id, model.StatusRejected)
if uErr := upd.Updates(map[string]any{
"title": snap.Title,
"year": snap.Year,
"city": snap.City,
@ -387,6 +397,9 @@ func (r *ingestRepository) ReuseRejectedStreetSnap(ctx context.Context, id uint3
}).Error; uErr != nil {
return uErr
}
if upd.RowsAffected == 0 {
return fmt.Errorf("street snap %d 已不是 rejected 状态，复用前置条件失效", id)
}
if dErr := tx.Model(&model.StreetSnapImage{}).
Where("snap_id = ? AND is_deleted = 0", id).
Updates(map[string]any{"is_deleted": 1, "updated_at": now}).Error; dErr != nil {
@ -448,6 +461,10 @@ func (r *ingestRepository) RetryJob(ctx context.Context, id uint32) error {
// FindNearDuplicateImage 在给定图片表中按 dHash 汉明距离检索近重复，取距离 ≤ threshold 的最近一条。
// 算子用 pgvector 的 L2（<->）；因 phash 是 {0,1}^64 向量，L2² == 汉明距离，故 L2 阈值 = sqrt(threshold)。
// phashBits 为 vector(64) 二进制向量串；NULL 的 phash 不参与比较。
//
// 只比对未软删的行（is_deleted = 0）：复用驳回行时去重先于软删旧图执行，
// 否则重爬到的同一张图会命中「即将被软删的旧行」，留痕 dup_of 指向一条公开不可见的记录。
// 本方法被 runway / street 两条入库路径共用，过滤对两者语义一致。
func (r *ingestRepository) FindNearDuplicateImage(ctx context.Context, tables []string, phashBits string, threshold int) (uint32, bool, error) {
// 汉明阈值转 L2 阈值：phash 为 {0,1}^64 向量，L2² == 汉明距离，故 L2 阈值 = sqrt(汉明阈值)。
l2Limit := math.Sqrt(float64(threshold))
@ -459,6 +476,7 @@ func (r *ingestRepository) FindNearDuplicateImage(ctx context.Context, tables []
err := r.db.WithContext(ctx).Table(t).
Select("id, (phash <-> ?::vector) AS dist", phashBits).
Where("(phash <-> ?::vector) <= ?", phashBits, l2Limit).
Where("is_deleted = 0").
Order("dist ASC").
Limit(1).
Scan(&row).Error