From cdef922251112073fe59805db4da2a469c78df2b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=84=E8=87=AA=E8=BE=BE?= <980324341@qq.com> Date: Tue, 20 Oct 2026 22:35:25 +0800 Subject: [PATCH] =?UTF-8?q?fix(brand-label):=20=E5=93=81=E7=89=8C=E6=A0=87?= =?UTF-8?q?=E7=AD=BE=E6=B8=85=E6=B4=97=E6=8A=BD=E5=85=AC=E5=85=B1=E4=BB=B6?= =?UTF-8?q?=EF=BC=8C=E5=85=AD=E6=A8=A1=E5=9D=97=E8=AF=BB=E5=86=99=E4=B8=A4?= =?UTF-8?q?=E7=AB=AF=E7=BB=9F=E4=B8=80=E8=BF=87=E5=87=BA=E5=8F=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit PublishBrandLabel 从 publish 私有件搬入 common/util/BrandLabel(三条守卫尺子随迁: 唯一副本、实体写出不许绕过、解析端不许重复实现),品牌列的读入与出件两端都走这一处。 读入端接上:asincache 缓存落库、invalidasin 人工黑名单(先洗再判空——键对不上是静默失效, 比脏值出件更难发现)、appearancepatent resolveBrand(两条取值分支同一个串)、collectdata 入库、shopdatacrawl copyRow/sameRow/rowDedupKey(键与值分家会让同一行跨分片重复出件)、 撞款解析与 re-baseline、dedupe 按表头名只洗品牌列。 写出端接上:publish/similarasin×2/appearancepatent/collectdata/撞款 csv+xlsx/shopdatacrawl。 rows-snapshot golden 里品牌格去掉首尾空格:正本 strip 自 2026-10-08 就含 trim, Java/Python/MySQL 三处同形,不是这次新加的行为。 门禁:后端 mvn -o test 4594 例 0 红。BrandTaskService(含四张 sheet 出件与解析端匹配) 因同仓另一路在飞改动未动,见 docs/audit/BACKLOG.md B5。 --- .../util/BrandLabel.java} | 14 +++--- .../support/AppearancePatentSheetBuilder.java | 14 ++++-- .../service/AsinResultCacheService.java | 5 ++- .../CollectDataExcelAssemblyService.java | 5 ++- .../service/CollectDataService.java | 5 ++- .../dedupe/service/DedupeRunService.java | 8 +++- .../service/InvalidAsinDataService.java | 10 ++++- .../publish/service/PublishTaskService.java | 3 +- .../service/PublishWorkbookService.java | 5 ++- .../ShopDataCrawlItemStoreService.java | 5 ++- .../support/ShopDataCrawlRowNormalizer.java | 16 +++++-- .../support/ShopDataCrawlSheetBuilder.java | 6 ++- .../ShopDataDuplicateCheckImportService.java | 5 ++- .../ShopDataDuplicateCheckQueryService.java | 5 ++- .../ShopDataDuplicateCheckScanService.java | 5 ++- .../support/DuplicateCheckWorkbookParser.java | 4 +- .../SimilarAsinResultWorkbookAssembler.java | 4 +- .../support/SimilarAsinSheetBuilder.java | 4 +- .../util/BrandLabelTest.java} | 43 ++++++++++--------- .../AppearancePatentRowOutcomeTest.java | 33 ++++++++++++++ .../service/AsinResultCacheServiceTest.java | 17 ++++++++ .../CollectDataExcelAssemblyServiceTest.java | 27 ++++++++++++ .../service/CollectDataServiceTest.java | 21 +++++++++ .../DedupeRunServiceCleanRulesTest.java | 20 +++++++++ .../service/InvalidAsinDataServiceTest.java | 40 +++++++++++++++++ .../ShopDataCrawlRowNormalizerTest.java | 25 +++++++++++ .../ShopDataCrawlSheetBuilderTest.java | 23 ++++++++++ .../support/ShopDataCrawlSnapshotTest.java | 3 ++ ...opDataDuplicateCheckImportServiceTest.java | 32 ++++++++++++++ ...hopDataDuplicateCheckQueryServiceTest.java | 38 ++++++++++++++++ ...ShopDataDuplicateCheckScanServiceTest.java | 18 ++++++++ .../DuplicateCheckWorkbookParserTest.java | 25 +++++++++++ .../support/SimilarAsinSheetBuilderTest.java | 15 +++++++ .../shopdatacrawl/golden/rows-snapshot.txt | 2 +- 34 files changed, 456 insertions(+), 49 deletions(-) rename backend-java/src/main/java/com/nanri/aiimage/{modules/publish/service/PublishBrandLabel.java => common/util/BrandLabel.java} (62%) rename backend-java/src/test/java/com/nanri/aiimage/{modules/publish/service/PublishBrandLabelTest.java => common/util/BrandLabelTest.java} (80%) diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabel.java b/backend-java/src/main/java/com/nanri/aiimage/common/util/BrandLabel.java similarity index 62% rename from backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabel.java rename to backend-java/src/main/java/com/nanri/aiimage/common/util/BrandLabel.java index 1bbe135cd..602214af5 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabel.java +++ b/backend-java/src/main/java/com/nanri/aiimage/common/util/BrandLabel.java @@ -1,19 +1,23 @@ -package com.nanri.aiimage.modules.publish.service; +package com.nanri.aiimage.common.util; import java.util.regex.Pattern; /** - * 上架链路上品牌值的唯一清洗出口:剥掉 {@code Brand: } 这类"字段标签 + 冒号"前缀。 + * 品牌值的唯一清洗出口:剥掉 {@code Brand: } 这类"字段标签 + 冒号"前缀。全仓只此一份, + * 读入(解析上传的 Excel、接收客户端回传的行)与写出(出件写 cell)两端都走它。 * *

亚马逊详情页把卖家名渲染成 "Brand: XXX",用户从页面复制进 Excel 就会带上前缀。 * 该值进 WIPO 品牌库查询时会被当 Solr 字段限定符({@code undefined field Brand} → 上游 500), * 10-08 单日被顶着重跑 30 轮、连带把品牌全局锁持到 1h45m,压死 11 个别的用户的任务。 * - *

清洗必须落在这里而不是某个调用点:{@code biz_publish_item.brand} 有两个写入来源 + *

清洗必须落在这一处而不是某个调用点:{@code biz_publish_item.brand} 有两个写入来源 * (服务端 Excel 解析、客户端结果分片回传 {@code replaceRows}),历史上只堵了前者, * 后者在任务完成时整表删后重插,把脏值又写回去。 + * + *

读写两端都要过一遍,是因为干净值不等于干净文件:出件是用户的下一份上传物, + * 任何一个出件漏剥,用户下载再上传就把脏值送回链路里,清洗脚本的账也随之作废。 */ -public final class PublishBrandLabel { +public final class BrandLabel { /** * 只剥行首的一个标签,且要求冒号紧跟其后:{@code Brands: NIKE} 与 {@code Brand: NIKE: X} @@ -22,7 +26,7 @@ public final class PublishBrandLabel { private static final Pattern LABEL_PREFIX = Pattern.compile( "(?i)^\\s*(?:brand|品牌|marke|marca|marque|marka)\\s*[::]\\s*"); - private PublishBrandLabel() { + private BrandLabel() { } public static String strip(String value) { diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentSheetBuilder.java b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentSheetBuilder.java index e3caff32c..f91b98e1a 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentSheetBuilder.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentSheetBuilder.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.appearancepatent.service.support; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.appearancepatent.model.dto.AppearancePatentResultRowDto; import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentParsedRowVo; import org.apache.poi.ss.usermodel.Cell; @@ -231,10 +232,17 @@ public final class AppearancePatentSheetBuilder { return ""; } + /** + * 品牌列的唯一取值出口:回传行优先、退回用户上传表的原始列,两条来源都在这里过一遍 + * {@link BrandLabel}。清洗必须落在这处而不是 cell 那一行——亚马逊把卖家名渲染成 + * "Brand: XXX",用户复制进 Excel 或客户端原样回传就带着标签前缀,带着它去查品牌库会被 + * 当 Solr 字段限定符打成上游 500。回传值只剩标签("Brand:")时按"这一行没带品牌"处理, + * 继续回退到源表的品牌列,而不是把一个空标签洗出来的空串当成有效值写进格子。 + */ public static String resolveBrand(AppearancePatentResultRowDto resultRow, AppearancePatentParsedRowVo parsedRow) { - String pythonBrand = resultRow == null ? "" : resultRow.getBrand(); + String pythonBrand = resultRow == null ? "" : BrandLabel.strip(resultRow.getBrand()); if (pythonBrand != null && !pythonBrand.isBlank()) { - return pythonBrand.trim(); + return pythonBrand; } if (parsedRow == null || parsedRow.getValues() == null) { return ""; @@ -242,7 +250,7 @@ public final class AppearancePatentSheetBuilder { for (Map.Entry entry : parsedRow.getValues().entrySet()) { String header = entry.getKey() == null ? "" : entry.getKey().trim().toLowerCase(Locale.ROOT); if (header.contains("品牌") || header.contains("brand")) { - return entry.getValue() == null ? "" : entry.getValue(); + return firstNonBlank(BrandLabel.strip(entry.getValue()), ""); } } return ""; diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheService.java index 4b499ec5c..ad55dea32 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheService.java @@ -1,6 +1,7 @@ package com.nanri.aiimage.modules.asincache.service; import com.fasterxml.jackson.databind.ObjectMapper; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.asincache.mapper.AsinResultCacheMapper; import com.nanri.aiimage.modules.asincache.model.dto.AsinCacheWriteRowDto; import com.nanri.aiimage.modules.asincache.model.entity.AsinResultCacheEntity; @@ -260,7 +261,9 @@ public class AsinResultCacheService { entity.setAsin(asin); entity.setAvailability(availability); entity.setPrice(clip(row.getPrice(), MAX_PRICE_LENGTH)); - entity.setBrand(clip(row.getBrand(), MAX_BRAND_LENGTH)); + // 缓存行来自客户端回传的采集结果,品牌那格常常是详情页署名行原样("Brand: XXX"): + // 这一列洗不洗决定了后面多少个任务从缓存里读到脏值,落库前剥一次比在每个下游各剥一次可靠 + entity.setBrand(clip(BrandLabel.strip(row.getBrand()), MAX_BRAND_LENGTH)); entity.setDeliveryMethod(clip(row.getDeliveryMethod(), MAX_DELIVERY_LENGTH)); entity.setSellerName(clip(row.getSellerName(), MAX_SELLER_NAME_LENGTH)); entity.setSellerId(clip(row.getSellerId(), MAX_SELLER_ID_LENGTH)); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyService.java index fd811d821..38d3b94e2 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyService.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.collectdata.service; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.UserMessages; import com.nanri.aiimage.common.exception.BusinessException; import com.nanri.aiimage.common.util.SxssfWorkbookWriter; @@ -140,7 +141,9 @@ public class CollectDataExcelAssemblyService { continue; } Row row = sheet.createRow((int) state[0]++); - SxssfWorkbookWriter.setText(row, 0, item.getBrand()); + // 出件这一端也要过一遍:干净值不等于干净文件——这份 Excel 是用户的下一份上传物, + // 而改口径之前落库的历史明细 payload 里仍可能带着标签前缀。 + SxssfWorkbookWriter.setText(row, 0, BrandLabel.strip(item.getBrand())); SxssfWorkbookWriter.setText(row, 1, item.getAsin()); SxssfWorkbookWriter.setText(row, 2, item.getPrice()); SxssfWorkbookWriter.setText(row, 3, item.getSellerName()); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataService.java index bf648e3f3..fbb2eece2 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/service/CollectDataService.java @@ -9,6 +9,7 @@ import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.ObjectMapper; import com.nanri.aiimage.common.exception.BusinessCodes; import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.ExcelStreamReader; import com.nanri.aiimage.modules.asincache.service.AsinResultCacheService; import com.nanri.aiimage.modules.collectdata.mapper.CollectDataCountryPrefMapper; @@ -971,7 +972,9 @@ public class CollectDataService extends AbstractFileTaskPipeline EXPORT_COLUMN_PRIORITY = List.of("id", "ASIN", "国家", "价格", "品牌"); + /** 去重源文件的列集由用户勾选,品牌列只能按表头认出来,其余列一律原文透传。 */ + private static final String BRAND_COLUMN_HEADER = "品牌"; /** 同时处理的文件数上限,避免多文件并行时打满内存与 OSS 上传带宽 */ private static final int MAX_PARALLEL_FILES = 4; /** 单用户同时运行中的去重任务数上限 */ @@ -856,7 +859,10 @@ public class DedupeRunService { List selectedColumns, Integer asinColumnIndex) { List selectedValues = new ArrayList<>(selectedColumns.size()); for (String selectedColumn : selectedColumns) { - selectedValues.add(normalizeCellText(cellText(rowMap, headerIndex.get(selectedColumn)))); + String value = normalizeCellText(cellText(rowMap, headerIndex.get(selectedColumn))); + // 品牌列的值可能是用户从亚马逊详情页复制来的署名("Brand: XXX"):在这里剥一次, + // 下方出件写 cell 用的就是同一个字符串,比对与交付不会拿到两个版本 + selectedValues.add(BRAND_COLUMN_HEADER.equals(selectedColumn) ? BrandLabel.strip(value) : value); } if (asinColumnIndex == null) { asinColumnIndex = headerIndex.get("ASIN"); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataService.java index a5b89fe8d..0c3a1e367 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataService.java @@ -2,6 +2,7 @@ package com.nanri.aiimage.modules.invalidasin.service; import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper; import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.invalidasin.mapper.InvalidAsinDataMapper; import com.nanri.aiimage.modules.invalidasin.model.dto.InvalidAsinDataCreateRequest; import com.nanri.aiimage.modules.invalidasin.model.dto.InvalidAsinDataUpdateRequest; @@ -83,7 +84,11 @@ public class InvalidAsinDataService { public InvalidAsinDataItemVo create(InvalidAsinDataCreateRequest request, Long operatorId) { ShopManageGroupEntity group = requireGroup(request.getGroupId(), operatorId); String dataValue = normalizeRequired(request.getDataValue(), "ASIN 不能为空"); - String brand = normalizeRequired(request.getBrand(), "品牌不能为空").toLowerCase(Locale.ROOT); + // 品牌是这张表的匹配键(下游按它判"这个 ASIN 属于被屏蔽品牌"),键值一律先过 BrandLabel: + // 人工把详情页署名行整段粘进来会存成 "brand: nike",而采集侧现在给 "nike", + // 对不上就是屏蔽静默失效——比脏值出件更坏,因为它不报错。先洗再判空,纯标签值必须拒收。 + String brand = normalizeRequired(BrandLabel.strip(request.getBrand()), "品牌不能为空") + .toLowerCase(Locale.ROOT); ensureUnique(dataValue, null); InvalidAsinDataEntity entity = new InvalidAsinDataEntity(); entity.setDataValue(dataValue); @@ -101,7 +106,8 @@ public class InvalidAsinDataService { // 自动导入的行分组由采集侧决定,这一页不替它重归:沿用改造前口径,与角色无关。 ShopManageGroupEntity writeGroup = isManualRecord(entity) ? requireGroup(request.getGroupId(), operatorId) : null; String dataValue = normalizeRequired(request.getDataValue(), "ASIN 不能为空"); - String brand = normalizeRequired(request.getBrand(), "品牌不能为空").toLowerCase(Locale.ROOT); + String brand = normalizeRequired(BrandLabel.strip(request.getBrand()), "品牌不能为空") + .toLowerCase(Locale.ROOT); ensureUnique(dataValue, id); entity.setDataValue(dataValue); entity.setBrand(brand); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishTaskService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishTaskService.java index dc08757ea..799393e77 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishTaskService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishTaskService.java @@ -10,6 +10,7 @@ import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.ObjectMapper; import com.nanri.aiimage.common.exception.BusinessException; import com.nanri.aiimage.common.exception.TaskOwnerMismatchException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.config.InstanceMetadata; import com.nanri.aiimage.modules.file.service.LocalFileStorageService; import com.nanri.aiimage.modules.file.service.oss.OssStorageService; @@ -1830,7 +1831,7 @@ public class PublishTaskService extends AbstractFileTaskPipeline entity.setSourceId(row.getSourceId()); entity.setAsin(row.getAsin()); entity.setCountry(row.getCountry()); - entity.setBrand(PublishBrandLabel.strip(row.getBrand())); + entity.setBrand(BrandLabel.strip(row.getBrand())); entity.setPriceValue(row.getPrice()); entity.setStatusValue(row.getStatus()); entity.setSyncStatus(row.getSyncStatus()); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishWorkbookService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishWorkbookService.java index 15add2899..ddc833d2a 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishWorkbookService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/publish/service/PublishWorkbookService.java @@ -1,6 +1,7 @@ package com.nanri.aiimage.modules.publish.service; import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.ExcelStreamReader; import com.nanri.aiimage.common.util.SxssfWorkbookWriter; import com.nanri.aiimage.modules.publish.model.dto.PublishRowDto; @@ -169,7 +170,7 @@ public class PublishWorkbookService { row.setSourceId(values.get(0)); row.setAsin(values.get(1)); row.setCountry(values.get(2)); - row.setBrand(PublishBrandLabel.strip(values.get(3))); + row.setBrand(BrandLabel.strip(values.get(3))); row.setPrice(values.get(4)); return row; } @@ -194,7 +195,7 @@ public class PublishWorkbookService { setText(row, 0, value.getSourceId()); setText(row, 1, value.getAsin()); setText(row, 2, publishCountry); - setText(row, 3, value.getBrand()); + setText(row, 3, BrandLabel.strip(value.getBrand())); setPrice(row, 4, value.getPrice()); setText(row, 5, value.getStatus()); setText(row, 6, value.getSyncStatus()); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/ShopDataCrawlItemStoreService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/ShopDataCrawlItemStoreService.java index 711b7ede4..3b4479eac 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/ShopDataCrawlItemStoreService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/ShopDataCrawlItemStoreService.java @@ -2,6 +2,7 @@ package com.nanri.aiimage.modules.shopdatacrawl.service; import com.fasterxml.jackson.databind.ObjectMapper; import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.shopdatacrawl.mapper.ShopDataCrawlItemMapper; import com.nanri.aiimage.modules.shopdatacrawl.model.dto.ShopDataCrawlCountryResultDto; import com.nanri.aiimage.modules.shopdatacrawl.model.dto.ShopDataCrawlRowDto; @@ -104,7 +105,9 @@ public class ShopDataCrawlItemStoreService { entity.setBusinessDate(businessDate); entity.setCountry(country); entity.setAsin(item.getAsin().trim()); - entity.setBrand(trimToNull(item.getBrand())); + // 明细表这一列是撞款扫描与台账出件唯一的品牌来源,而存量迁移的快照 JSON 不经 copyRow: + // 标签前缀必须在进表这一端剥掉,否则脏值每扫一次就复活一次 + entity.setBrand(trimToNull(BrandLabel.strip(item.getBrand()))); entity.setPrice(trimToNull(item.getPrice())); entity.setItemDate(trimToNull(item.getDate())); entity.setUnitsSold(trimToNull(item.getUnitsSold())); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizer.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizer.java index c905b2559..0a8e161a5 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizer.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizer.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.shopdatacrawl.service.support; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.shopdatacrawl.model.dto.ShopDataCrawlRowDto; /** @@ -21,6 +22,15 @@ public final class ShopDataCrawlRowNormalizer { return value == null ? "" : value.trim(); } + /** + * 品牌列的唯一取值口径:客户端从亚马逊详情页的署名行原样回传,值里带着 "Brand: " 这类标签前缀。 + * 进模型、比相等、算去重键三处必须取同一个字符串 —— 只在写出那一端剥,会让清洗过的存量行与 + * 未清洗的回传行算出两个键,跨分片合并时同一行重复出件。 + */ + public static String brand(String value) { + return BrandLabel.strip(trim(value)); + } + public static boolean blank(String value) { return value == null || value.isBlank(); } @@ -56,7 +66,7 @@ public final class ShopDataCrawlRowNormalizer { ShopDataCrawlRowDto row = new ShopDataCrawlRowDto(); row.setDate(trim(source.getDate())); row.setAsin(trim(source.getAsin())); - row.setBrand(trim(source.getBrand())); + row.setBrand(brand(source.getBrand())); row.setCommodityImage(trim(source.getCommodityImage())); row.setInventorySales(trim(source.getInventorySales())); row.setSalesRank(trim(source.getSalesRank())); @@ -77,7 +87,7 @@ public final class ShopDataCrawlRowNormalizer { return left != null && right != null && java.util.Objects.equals(trim(left.getDate()), trim(right.getDate())) && java.util.Objects.equals(trim(left.getAsin()), trim(right.getAsin())) - && java.util.Objects.equals(trim(left.getBrand()), trim(right.getBrand())) + && java.util.Objects.equals(brand(left.getBrand()), brand(right.getBrand())) && java.util.Objects.equals(trim(left.getCommodityImage()), trim(right.getCommodityImage())) && java.util.Objects.equals(trim(left.getInventorySales()), trim(right.getInventorySales())) && java.util.Objects.equals(trim(left.getSalesRank()), trim(right.getSalesRank())) @@ -92,7 +102,7 @@ public final class ShopDataCrawlRowNormalizer { if (row == null) { return null; } - return trim(row.getDate()) + ROW_KEY_SEPARATOR + trim(row.getAsin()) + ROW_KEY_SEPARATOR + trim(row.getBrand()) + return trim(row.getDate()) + ROW_KEY_SEPARATOR + trim(row.getAsin()) + ROW_KEY_SEPARATOR + brand(row.getBrand()) + ROW_KEY_SEPARATOR + trim(row.getCommodityImage()) + ROW_KEY_SEPARATOR + trim(row.getInventorySales()) + ROW_KEY_SEPARATOR + trim(row.getSalesRank()) + ROW_KEY_SEPARATOR + trim(row.getPageViews()) + ROW_KEY_SEPARATOR + trim(row.getUnitsSold()) + ROW_KEY_SEPARATOR + trim(row.getPrice()) diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilder.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilder.java index 9ba742540..9a39f4b0a 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilder.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilder.java @@ -1,6 +1,7 @@ package com.nanri.aiimage.modules.shopdatacrawl.service.support; import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.shopdatacrawl.model.dto.ShopDataCrawlCountryResultDto; import com.nanri.aiimage.modules.shopdatacrawl.model.dto.ShopDataCrawlRowDto; import com.nanri.aiimage.modules.shopdatacrawl.model.vo.ShopDataCrawlResultItemVo; @@ -128,7 +129,10 @@ public final class ShopDataCrawlSheetBuilder { /** 数据行 10 列值布局 + 样式应用(图片嵌入由调用方负责)。 */ public static void writeDataRowValues(Row row, ShopDataCrawlRowDto value, CellStyle[] styles) { String[] values = {value.getDate(), value.getAsin(), "", value.getInventorySales(), value.getSalesRank(), - value.getPageViews(), value.getUnitsSold(), value.getPrice(), value.getRecommendedOffer(), value.getBrand()}; + // 出件是用户的下一份上传物:改口径之前落库的累计文件快照与 row_payload 仍带 "Brand:" 前缀, + // 这一格不剥,下载再上传就把脏值送回链路里 + value.getPageViews(), value.getUnitsSold(), value.getPrice(), value.getRecommendedOffer(), + BrandLabel.strip(value.getBrand())}; for (int column = 0; column < values.length; column++) { Cell cell = row.createCell(column); if (styles[column] != null) cell.setCellStyle(styles[column]); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportService.java index 36cde8414..c368b5e0b 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportService.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.shopduplicatecheck.service; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.UserMessages; import com.fasterxml.jackson.databind.ObjectMapper; import com.nanri.aiimage.common.exception.BusinessException; @@ -105,8 +106,10 @@ public class ShopDataDuplicateCheckImportService { } for (DuplicateItem item : base.items()) { for (DuplicateOccurrence occ : item.occurrences()) { + // 基线行读自上一份落库 payload(改口径之前可能是脏值),这里不剥就直接把脏值重插回扫描表 rowsByShop.computeIfAbsent(occ.shopName(), k -> new ArrayList<>()) - .add(new RawRow(occ.asin(), occ.date(), occ.price(), occ.brand(), occ.country())); + .add(new RawRow(occ.asin(), occ.date(), occ.price(), + BrandLabel.strip(occ.brand()), occ.country())); } } List baseParsed = new ArrayList<>(); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryService.java index e3c80a176..3b9f73ead 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryService.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.shopduplicatecheck.service; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.shopduplicatecheck.mapper.ShopDuplicateCheckSourceMapper; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateItem; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateOccurrence; @@ -433,8 +434,10 @@ public class ShopDataDuplicateCheckQueryService { for (DuplicateItem item : matched) { String shopCount = String.valueOf(item.shopCount()); for (DuplicateOccurrence occ : item.occurrences()) { + // 出件是用户的下一份上传物:库里的历史 payload 仍可能带标签前缀,最后一列(品牌)在此剥掉 flat.add(new Object[]{item.asin(), shopCount, occ.shopName(), occ.groupName(), - effectiveCountryCn(occ), safe(occ.date()), safe(occ.price()), safe(occ.brand())}); + effectiveCountryCn(occ), safe(occ.date()), safe(occ.price()), + BrandLabel.strip(safe(occ.brand()))}); } } flat.sort(flatExportComparator()); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanService.java index 2da17a8dc..007b00f92 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanService.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.shopduplicatecheck.service; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.UserMessages; import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.ObjectMapper; @@ -325,7 +326,9 @@ public class ShopDataDuplicateCheckScanService implements DuplicateCheckRefreshP row.asin().trim(), nvl(row.itemDate()), nvl(row.price()), - nvl(row.brand()), + // 明细表里的品牌是改口径之前写进去的,标签前缀在进聚合这一端剥, + // 落库 payload 与页面、出件拿到的才是同一个干净值 + BrandLabel.strip(nvl(row.brand())), country)); if (!country.isEmpty()) { countries.add(country); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParser.java b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParser.java index dff33ed67..7175e9b4e 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParser.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParser.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.shopduplicatecheck.service.support; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.ExcelStreamReader; import java.io.ByteArrayInputStream; @@ -105,7 +106,8 @@ public final class DuplicateCheckWorkbookParser { asin, cell(rowMap.get(dateCol)), cell(rowMap.get(priceCol)), - cell(rowMap.get(brandCol)), + // 上传文件的品牌列是亚马逊详情页复制来的署名("Brand: XXX"),标签前缀只能在这一处剥 + BrandLabel.strip(cell(rowMap.get(brandCol))), sheetCountry)); } } diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinResultWorkbookAssembler.java b/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinResultWorkbookAssembler.java index 990b1563e..e93d7e923 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinResultWorkbookAssembler.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinResultWorkbookAssembler.java @@ -4,6 +4,7 @@ import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper; import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.ObjectMapper; import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.common.util.GroupResultPropagator; import com.nanri.aiimage.config.SimilarAsinProperties; import com.nanri.aiimage.config.StorageProperties; @@ -901,7 +902,8 @@ public class SimilarAsinResultWorkbookAssembler { resultRow.getPrice(), firstNonBlank(parsedRow.getPrice(), readValueByHeader(parsedRow, "价格", "price")))); row.createCell(col++).setCellValue(readValueByHeader(parsedRow, "卖家名称", "卖家名", "卖家", "店铺名称", "店铺名", "seller name", "seller_name", "seller-name", "sellername", "store name", "shop name")); - row.createCell(col++).setCellValue(readValueByHeader(parsedRow, "品牌", "brand")); + // 与 SheetBuilder 同口径:结果文件的品牌列必须剥掉页面复制带上的标签前缀,否则二次上传把脏值带回链路 + row.createCell(col++).setCellValue(BrandLabel.strip(readValueByHeader(parsedRow, "品牌", "brand"))); String isStock = resultRow == null ? "" : SimilarAsinResultTextSupport.userFacingLlmCellValue(resultRow, resultRow.getIsStock()); String similarity = resultRow == null ? "" : SimilarAsinResultTextSupport.userFacingLlmCellValue(resultRow, resultRow.getSimilarity()); String isConform = resultRow == null ? "" : SimilarAsinResultTextSupport.userFacingLlmCellValue(resultRow, resultRow.getIsConform()); diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilder.java b/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilder.java index 2656b4573..0382a2151 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilder.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilder.java @@ -1,5 +1,6 @@ package com.nanri.aiimage.modules.similarasin.service.support; +import com.nanri.aiimage.common.util.BrandLabel; import com.nanri.aiimage.modules.similarasin.model.vo.SimilarAsinParsedRowVo; import com.nanri.aiimage.modules.similarasin.util.SimilarAsinImageEmbedder; import org.apache.poi.ss.usermodel.Cell; @@ -84,7 +85,8 @@ public final class SimilarAsinSheetBuilder { row.createCell(col++).setCellValue(readValueByHeader( parsedRow, "卖家名称", "卖家名", "卖家", "店铺名称", "店铺名", "seller name", "seller_name", "seller-name", "sellername", "store name", "shop name")); - row.createCell(col).setCellValue(readValueByHeader(parsedRow, "品牌", "brand")); + // 出件是用户的下一份上传物:品牌列带 "Brand:" 标签前缀不剥,下载再上传就把脏值送回链路(清洗出口唯 BrandLabel) + row.createCell(col).setCellValue(BrandLabel.strip(readValueByHeader(parsedRow, "品牌", "brand"))); row.setHeightInPoints(SimilarAsinImageEmbedder.IMAGE_ROW_HEIGHT_POINTS); rowIndex++; } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabelTest.java b/backend-java/src/test/java/com/nanri/aiimage/common/util/BrandLabelTest.java similarity index 80% rename from backend-java/src/test/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabelTest.java rename to backend-java/src/test/java/com/nanri/aiimage/common/util/BrandLabelTest.java index 1a3667982..c8a7cf147 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/publish/service/PublishBrandLabelTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/common/util/BrandLabelTest.java @@ -1,4 +1,4 @@ -package com.nanri.aiimage.modules.publish.service; +package com.nanri.aiimage.common.util; import static org.assertj.core.api.Assertions.assertThat; import static org.assertj.core.api.Assertions.fail; @@ -14,7 +14,7 @@ import java.util.stream.Stream; import org.junit.jupiter.api.Test; /** - * 上架品牌值的清洗出口。 + * 品牌标签前缀清洗出口。 * *

2026-10-08 的事故链是:Excel 里带 "Brand:" 标签前缀 → WIPO 把它当 Solr 字段限定符 * ({@code undefined field Brand},单日 19,247 次 500)→ 上游错误信封不是合法 Base64、被客户端 @@ -24,29 +24,32 @@ import org.junit.jupiter.api.Test; * {@code PublishTaskService.replaceRows} 会用客户端回传的分片整表删后重插, * 脏值照样落库(现网 7 个脏值来源任务里 5 个的行创建时间 == 任务完成时刻,即由这条路径写入)。 * 所以护栏从"注释里提醒调用方"改成"落在唯一写库出口 + 可实跑的用例"。 + * + *

本类原在 {@code modules/publish/service} 下,因品牌列的读写两端散在多个模块而出搬为公共件; + * 三条守卫尺子(唯一副本、出口不被绕过、解析端不重复实现)随迁。 */ -class PublishBrandLabelTest { +class BrandLabelTest { - private static final String OUTLET_CLASS = "PublishBrandLabel.java"; + private static final String OUTLET_CLASS = "BrandLabel.java"; /** 标签正则的特征片段:出现第二处就等于开了第二条会漂移的副本。 */ private static final String REGEX_SIGNATURE = "marque|marka"; @Test void stripsLabelPrefixButKeepsLookalikes() { - assertThat(PublishBrandLabel.strip("Brand: NIKE")).isEqualTo("NIKE"); - assertThat(PublishBrandLabel.strip(" brand : NIKE ")).isEqualTo("NIKE"); - assertThat(PublishBrandLabel.strip("品牌:耐克")).isEqualTo("耐克"); - assertThat(PublishBrandLabel.strip("MARCA: X")).isEqualTo("X"); + assertThat(BrandLabel.strip("Brand: NIKE")).isEqualTo("NIKE"); + assertThat(BrandLabel.strip(" brand : NIKE ")).isEqualTo("NIKE"); + assertThat(BrandLabel.strip("品牌:耐克")).isEqualTo("耐克"); + assertThat(BrandLabel.strip("MARCA: X")).isEqualTo("X"); // "Brands:" 不是标签;值中间的冒号属于品牌自身内容,只剥行首一次 - assertThat(PublishBrandLabel.strip("Brands: NIKE")).isEqualTo("Brands: NIKE"); - assertThat(PublishBrandLabel.strip("Brand: NIKE: X")).isEqualTo("NIKE: X"); - assertThat(PublishBrandLabel.strip("NIKE Brand")).isEqualTo("NIKE Brand"); + assertThat(BrandLabel.strip("Brands: NIKE")).isEqualTo("Brands: NIKE"); + assertThat(BrandLabel.strip("Brand: NIKE: X")).isEqualTo("NIKE: X"); + assertThat(BrandLabel.strip("NIKE Brand")).isEqualTo("NIKE Brand"); // 标签只有出现在行首才是"字段名";句中出现的要原样留着,否则把 ^ 锚摘掉(变成"逢 Brand: 就裁")这条就会红 - assertThat(PublishBrandLabel.strip("Super Brand: Nike")) + assertThat(BrandLabel.strip("Super Brand: Nike")) .isEqualTo("Super Brand: Nike"); - assertThat(PublishBrandLabel.strip("BrandNIKE")).isEqualTo("BrandNIKE"); - assertThat(PublishBrandLabel.strip("")).isEmpty(); - assertThat(PublishBrandLabel.strip(null)).isNull(); + assertThat(BrandLabel.strip("BrandNIKE")).isEqualTo("BrandNIKE"); + assertThat(BrandLabel.strip("")).isEmpty(); + assertThat(BrandLabel.strip(null)).isNull(); } @Test @@ -54,8 +57,8 @@ class PublishBrandLabelTest { // 现网 1,533 个脏值清洗后 0 个变空串(实测),但 "Brand:" 这种纯标签值确实会洗成空。 // 钉住这一条是因为 PublishTaskService.isBlankRow 要求八个字段同时为空才算空行, // 单靠 brand 变空不会让一行消失 —— 改了这条判据就会静默丢行。 - assertThat(PublishBrandLabel.strip("Brand:")).isEmpty(); - assertThat(PublishBrandLabel.strip("Brand: ")).isEmpty(); + assertThat(BrandLabel.strip("Brand:")).isEmpty(); + assertThat(BrandLabel.strip("Brand: ")).isEmpty(); } @Test @@ -74,7 +77,7 @@ class PublishBrandLabelTest { // 只有写进实体的那一行算出口;DTO 之间的搬运(copyRow/toRowDto)不是 if (line.matches("^entity\\.setBrand\\(.*\\);")) { written.add(file.getFileName() + ": " + line); - if (!line.contains("PublishBrandLabel.strip(")) { + if (!line.contains("BrandLabel.strip(")) { offenders.add(file.getFileName() + ": " + line); } } @@ -83,7 +86,7 @@ class PublishBrandLabelTest { // 取样集非空自检:扫描 0 命中时下面这条恒绿,等于没护栏 assertThat(written).as("publish 模块里写 PublishItemEntity.brand 的行应被扫到").isNotEmpty(); assertThat(offenders) - .as("绕过 PublishBrandLabel.strip 的 brand 写入点(客户端回传 replaceRows 会整表重插," + .as("绕过 BrandLabel.strip 的 brand 写入点(客户端回传 replaceRows 会整表重插," + "少堵一处就等于每完成一个任务把脏值写回去一次)") .isEmpty(); } @@ -104,7 +107,7 @@ class PublishBrandLabelTest { @Test void parseSideAlsoDelegatesInsteadOfReimplementing() throws IOException { String text = read(sourceOf("PublishWorkbookService.java")); - assertThat(text).contains("row.setBrand(PublishBrandLabel.strip(values.get(3)));"); + assertThat(text).contains("row.setBrand(BrandLabel.strip(values.get(3)));"); assertThat(text).doesNotContain("stripBrandLabel"); } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentRowOutcomeTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentRowOutcomeTest.java index 65722efb7..95098ad8f 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentRowOutcomeTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentRowOutcomeTest.java @@ -29,6 +29,7 @@ import static org.junit.jupiter.api.Assertions.assertTrue; */ class AppearancePatentRowOutcomeTest { + private static final int COL_BRAND = 4; private static final int COL_TITLE = 6; private static final int COL_URL = 7; private static final int COL_TITLE_RISK = 9; @@ -355,4 +356,36 @@ class AppearancePatentRowOutcomeTest { } throw new AssertionError("原因 sheet 里没有 ASIN=" + asin + " 的行"); } + + // ---------- 4. 品牌列只认洗过标签前缀的值 ---------- + + @Test + void brandCellNeverCarriesTheAmazonBylinePrefix() throws Exception { + // 品牌列有两个来源:用户上传表的「品牌」列(这里就是 "Brand: Acme")与客户端回传行的 Brand, + // 详情页整行复制来的署名前缀既脏出件、又会在品牌库查询里被当字段限定符打成上游 500。 + List all = List.of(sheetRowOnlyId("1", "B0A")); + AppearancePatentResultRowDto resultRow = judgedRow("1", "B0A", "T", "https://img/1.jpg"); + Map byRowKey = new LinkedHashMap<>(); + byRowKey.put(AppearancePatentSheetBuilder.rowKey(all.get(0)), resultRow); + + assertEquals("Acme", brandCell(all, byRowKey), "回传行没带品牌时,源表值也要剥掉标签"); + + resultRow.setBrand("Brand: Nike"); + assertEquals("Nike", brandCell(all, byRowKey)); + + // "Brands:" 是真实品牌写法、不是字段标签;值中间的冒号属于品牌本体,只能剥行首一次 + resultRow.setBrand("Brands: Nike"); + assertEquals("Brands: Nike", brandCell(all, byRowKey)); + + // 只剩标签的回传值等于这行没带品牌:回退源表列,而不是把洗出来的空串当成有效值写进格子 + resultRow.setBrand("Brand:"); + assertEquals("Acme", brandCell(all, byRowKey)); + } + + private static String brandCell(List rows, + Map byRowKey) throws Exception { + try (Workbook workbook = render(rows, byRowKey)) { + return cell(workbook.getSheet("外观专利检测结果"), 1, COL_BRAND); + } + } } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheServiceTest.java index 543ad9c41..456dcd1e2 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/asincache/service/AsinResultCacheServiceTest.java @@ -175,6 +175,23 @@ class AsinResultCacheServiceTest { assertEquals(255, captor.getValue().get(0).getBrand().length()); } + /** + * 缓存行来自客户端回传的采集结果,品牌那格常常是亚马逊详情页署名行原样("Brand: XXX")。 + * 这张缓存之后被 AsinCacheHitVo 喂给多个下游任务,入库前不剥一次,脏值就从这里向下游复播。 + */ + @Test + @SuppressWarnings({"rawtypes", "unchecked"}) + void writeStoresBrandWithoutTheBylineLabel() { + AsinCacheWriteRowDto dto = writeRow("OK", "9.90"); + dto.setBrand("Brand: NIKE"); + + service.write(List.of(dto)); + + ArgumentCaptor> captor = ArgumentCaptor.forClass(List.class); + verify(mapper).upsertBatch(captor.capture()); + assertEquals("NIKE", captor.getValue().get(0).getBrand()); + } + @Test void writeSplitsByBatchSizeAndSurvivesFailingBatch() { when(mapper.upsertBatch(anyList())).thenReturn(100).thenThrow(new RuntimeException("锁等待超时")); diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyServiceTest.java index d6c5e1df6..9b2796a42 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataExcelAssemblyServiceTest.java @@ -2,6 +2,7 @@ package com.nanri.aiimage.modules.collectdata.service; import com.fasterxml.jackson.databind.ObjectMapper; import com.nanri.aiimage.modules.collectdata.model.dto.CollectDataSummaryRowDto; +import com.nanri.aiimage.modules.collectdata.model.vo.CollectDataResultRowVo; import org.apache.poi.ss.usermodel.Sheet; import org.apache.poi.ss.usermodel.Workbook; import org.apache.poi.ss.usermodel.WorkbookFactory; @@ -34,4 +35,30 @@ class CollectDataExcelAssemblyServiceTest { assertThat(sheet.getRow(1).getCell(6).getNumericCellValue()).isEqualTo(3); } } + + /** + * 出件这一端的品牌列必须是没有标签前缀的:这份结果文件是用户的下一份上传物, + * 漏剥一次就把脏值送回链路里,清洗的账也随之作废;改口径之前落库的历史明细里也仍带着前缀。 + */ + @Test + void stripsAmazonBylinePrefixFromBrandColumn() throws Exception { + File output = tempDir.resolve("collect-data-brand-label.xlsx").toFile(); + + new CollectDataExcelAssemblyService().writeWorkbook(output, + List.of(row("Brand: Nike", "B000000001"), row("品牌:耐克", "B000000002")), + List.of(), List.of()); + + try (Workbook workbook = WorkbookFactory.create(output)) { + Sheet detail = workbook.getSheet("采集数据结果"); + assertThat(detail.getRow(1).getCell(0).getStringCellValue()).isEqualTo("Nike"); + assertThat(detail.getRow(2).getCell(0).getStringCellValue()).isEqualTo("耐克"); + } + } + + private static CollectDataResultRowVo row(String brand, String asin) { + CollectDataResultRowVo row = new CollectDataResultRowVo(); + row.setBrand(brand); + row.setAsin(asin); + return row; + } } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataServiceTest.java index fb31f1531..8f238be04 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/collectdata/service/CollectDataServiceTest.java @@ -5,6 +5,8 @@ import com.baomidou.mybatisplus.core.MybatisConfiguration; import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper; import com.baomidou.mybatisplus.core.metadata.TableInfoHelper; import org.apache.ibatis.builder.MapperBuilderAssistant; +import com.nanri.aiimage.modules.collectdata.model.dto.CollectDataSubmitRowDto; +import com.nanri.aiimage.modules.collectdata.model.vo.CollectDataResultRowVo; import com.nanri.aiimage.modules.collectdata.model.vo.CollectDataDashboardVo; import com.nanri.aiimage.modules.collectdata.model.vo.CollectDataTaskBatchVo; import com.nanri.aiimage.modules.task.mapper.FileResultMapper; @@ -219,6 +221,25 @@ class CollectDataServiceTest { verify(taskFileJobService, never()).enqueueAssembleResult(any(), any(), any(), any()); } + // 亚马逊详情页把卖家名渲染成 "Brand: XXX",用户复制、客户端就原样回传这种署名值。 + // 分片回传是脏值进服务端的唯一入口:品牌检查、结果明细落库、出件读的都是这里产出的 VO, + // 所以标签必须在进模型这一端就被剥掉(口径只在 BrandLabel 一处)。 + @Test + void submittedBrandLosesTheAmazonBylinePrefixAtIngest() throws Exception { + CollectDataSubmitRowDto byline = objectMapper.readValue( + "{\"品牌\":\"Brand: NIKE\"}", CollectDataSubmitRowDto.class); + CollectDataSubmitRowDto lookalike = objectMapper.readValue( + "{\"brand_name\":\"Brands: Nike\"}", CollectDataSubmitRowDto.class); + + List rows = ReflectionTestUtils.invokeMethod( + service, "normalizeSubmitRows", List.of(byline, lookalike)); + + assertThat(rows).hasSize(2); + assertThat(rows.get(0).getBrand()).isEqualTo("NIKE"); + // "Brands:" 是真实品牌写法,不是字段标签 + assertThat(rows.get(1).getBrand()).isEqualTo("Brands: Nike"); + } + private static FileTaskEntity staleTask(long taskId) { FileTaskEntity task = new FileTaskEntity(); task.setId(taskId); diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/dedupe/service/DedupeRunServiceCleanRulesTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/dedupe/service/DedupeRunServiceCleanRulesTest.java index 29dde606b..d6740b497 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/dedupe/service/DedupeRunServiceCleanRulesTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/dedupe/service/DedupeRunServiceCleanRulesTest.java @@ -93,6 +93,26 @@ class DedupeRunServiceCleanRulesTest { assertEquals(List.of("7", "B0FFFFFF01", "英国", "9.99", "品牌F", "在售"), readOutputRow(output, 1)); } + /** + * 去重出件的品牌列也必须剥掉亚马逊署名标签前缀:源文件是用户从详情页复制再上传的, + * 结果文件又是他的下一份上传物。源文件列集由用户勾选,所以只能按表头认出「品牌」这一列, + * 其余所选列一律原文透传(这里用同样带前缀的「状态」列钉住不越界)。 + * 摘掉 buildCandidateRow 里按表头认出的 strip,下面第一条断言读回来的就是 "Brand: NIKE"。 + */ + @Test + @DisplayName("品牌列出件剥掉行首标签前缀,其余所选列原文透传") + void stripsBrandLabelPrefixOnlyFromBrandColumn(@TempDir Path tempDir) throws Exception { + File output = runClean(tempDir, List.of( + new String[]{"1", "B0BRAND001", "英国", "10.00", "FBM", "Brand: NIKE", "Brand: 在售", "2026-09-01"}, + new String[]{"2", "B0BRAND002", "英国", "11.00", "FBM", "Brands: NIKE", "在售", "2026-09-01"}), + List.of("id", "ASIN", "国家", "价格", "品牌", "状态")); + + assertEquals(List.of("NIKE", "Brands: NIKE"), readOutputColumn(output, 4), + "品牌列只剥行首一个标签,「Brands:」是真实品牌写法"); + assertEquals(List.of("Brand: 在售", "在售"), readOutputColumn(output, 5), + "非品牌列不参与清洗"); + } + /** 纯规则单测:保留主链接开关打开时,整数行不再受子链接影响。 */ @Test @DisplayName("开启保留主链接时,整数行不受子链接影响") diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataServiceTest.java index 567797c2b..c9d294d92 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/invalidasin/service/InvalidAsinDataServiceTest.java @@ -298,6 +298,46 @@ class InvalidAsinDataServiceTest { return captor.getValue(); } + /** + * 品牌是这张表的匹配键(下游按它判"这个 ASIN 属于被屏蔽品牌")。人工把详情页署名行整段 + * 粘进来会存成 "brand: nike",而采集侧现在给 "nike" —— 对不上时屏蔽**静默失效**,不报错, + * 比脏值出件更难发现。摘掉 create 里的 BrandLabel.strip,这两条一条读到 "brand: nike"、 + * 一条根本不抛。 + */ + @Test + @SuppressWarnings({"rawtypes", "unchecked"}) + void createStripsTheBylineLabelFromTheMatchKey() { + InvalidAsinDataCreateRequest request = new InvalidAsinDataCreateRequest(); + request.setDataValue("B012345678"); + request.setBrand("Brand: Acme"); + request.setGroupId(9L); + when(shopManageGroupService.getAccessibleById(9L, 7L, true)).thenReturn(group(9L, "group-a")); + when(invalidAsinDataMapper.selectOne(any())).thenReturn(null); + when(invalidAsinDataMapper.insert(any(InvalidAsinDataEntity.class))).thenAnswer(invocation -> { + invocation.getArgument(0, InvalidAsinDataEntity.class).setId(91L); + return 1; + }); + when(invalidAsinDataMapper.selectById(91L)).thenReturn(data(91L, "MANUAL", 9L)); + + service.create(request, 7L); + + ArgumentCaptor captor = ArgumentCaptor.forClass(InvalidAsinDataEntity.class); + verify(invalidAsinDataMapper).insert(captor.capture()); + assertEquals("acme", captor.getValue().getBrand(), "标签前缀必须在落库前就剥掉,键值才是同一个串"); + } + + @Test + void createRejectsALabelOnlyBrand() { + // 先洗再判空:洗完全空=用户其实没填品牌,不能让它带着空串进唯一键 + InvalidAsinDataCreateRequest request = new InvalidAsinDataCreateRequest(); + request.setDataValue("B012345678"); + request.setBrand("Brand:"); + request.setGroupId(9L); + when(shopManageGroupService.getAccessibleById(9L, 7L, true)).thenReturn(group(9L, "group-a")); + + assertThrows(BusinessException.class, () -> service.create(request, 7L)); + } + private ShopManageGroupEntity group(Long id, String groupName) { ShopManageGroupEntity group = new ShopManageGroupEntity(); group.setId(id); diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizerTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizerTest.java index 0a964c0fd..9e726d066 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizerTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlRowNormalizerTest.java @@ -106,6 +106,31 @@ class ShopDataCrawlRowNormalizerTest { assertEquals(" B001 ", source.getAsin()); } + /** + * 客户端从亚马逊详情页署名行原样回传,品牌值带着 "Brand: " 标签前缀。 + * 读入(copyRow)与比对(sameRow / rowDedupKey)必须用同一个清洗后的字符串: + * 只洗其中一处,跨分片合并时已洗过的存量行与未洗的回传行会算出两个键,同一行重复出件。 + * 摘掉 ShopDataCrawlRowNormalizer.brand 里的 strip,下面三条断言全部回退。 + */ + @Test + void test_brand_label_prefix_stripped_on_ingest_and_in_dedup_key() { + ShopDataCrawlRowDto byline = row("2026-07-25", "B001", "Brand: NIKE", "img", + "10", "20", "30", "40", "50", "60"); + ShopDataCrawlRowDto plain = row("2026-07-25", "B001", "NIKE", "img", + "10", "20", "30", "40", "50", "60"); + + assertEquals("NIKE", ShopDataCrawlRowNormalizer.copyRow(byline).getBrand(), "回传行进模型即剥标签"); + assertTrue(ShopDataCrawlRowNormalizer.sameRow(byline, plain), "署名行与干净行是同一行"); + assertEquals(ShopDataCrawlRowNormalizer.rowDedupKey(plain), + ShopDataCrawlRowNormalizer.rowDedupKey(byline), "去重键按剥完标签的品牌计算"); + + // "Brands:" 是真实品牌写法,不是字段标签 + assertEquals("Brands: NIKE", + ShopDataCrawlRowNormalizer.copyRow(row("2026-07-25", "B001", "Brands: NIKE", "img", + "10", "20", "30", "40", "50", "60")).getBrand(), + "近似标签原样保留"); + } + // ---- rowEmpty ---- @Test diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilderTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilderTest.java index 99401a47e..a8a8312bc 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilderTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSheetBuilderTest.java @@ -128,6 +128,29 @@ class ShopDataCrawlSheetBuilderTest { } } + /** + * 出件的品牌末列必须剥掉亚马逊署名标签前缀:这份结果文件是用户的下一份上传物, + * 而改口径之前落库的累计文件快照与 row_payload 里仍带着 "Brand: " 前缀。 + * 摘掉 writeDataRowValues 里的 strip,下面第一条断言读回来的就是 "Brand: NIKE"。 + */ + @Test + void test_write_data_row_brand_label_prefix_stripped() throws Exception { + try (Workbook wb = new XSSFWorkbook()) { + Sheet sheet = ShopDataCrawlSheetBuilder.createStreamingSheet(wb, 0); + CellStyle[] none = new CellStyle[ShopDataCrawlSheetBuilder.HEADERS.size()]; + Row dirty = sheet.createRow(1); + ShopDataCrawlSheetBuilder.writeDataRowValues(dirty, + row("2026-07-25", "B012345678", "Brand: NIKE", "https://img/x.jpg"), none); + Row lookalike = sheet.createRow(2); + ShopDataCrawlSheetBuilder.writeDataRowValues(lookalike, + row("2026-07-25", "B012345679", "Brands: NIKE", "https://img/x.jpg"), none); + + assertEquals("NIKE", dirty.getCell(9).getStringCellValue(), "品牌末列剥掉行首标签"); + assertEquals("Brands: NIKE", lookalike.getCell(9).getStringCellValue(), "近似标签是真实品牌写法,不剥"); + assertEquals("2026-07-25", lookalike.getCell(0).getStringCellValue(), "其余列不参与品牌清洗"); + } + } + @Test void test_write_data_row_null_values_blank_cells() throws Exception { try (Workbook wb = new XSSFWorkbook()) { diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSnapshotTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSnapshotTest.java index 59494e987..23d88266e 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSnapshotTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopdatacrawl/service/support/ShopDataCrawlSnapshotTest.java @@ -82,6 +82,9 @@ class ShopDataCrawlSnapshotTest { private static ShopDataCrawlRowDto sparseRow() { ShopDataCrawlRowDto row = new ShopDataCrawlRowDto(); row.setAsin("B099999999"); + // 首尾空格刻意留着:品牌列出件要过 BrandLabel.strip,它除剥 "Brand:" 标签外还 trim + // (口径自 2026-10-08 就是这两手,Java/Python 正本与 MySQL 清洗脚本同形)。 + // golden 里这格是无空格的 "Brand With Spaces" —— 快照改动是这条,不是有人手抖改了基线。 row.setBrand(" Brand With Spaces "); return row; } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportServiceTest.java index afeb2ba89..99f453490 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckImportServiceTest.java @@ -12,6 +12,7 @@ import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateShop; import com.nanri.aiimage.modules.shopduplicatecheck.service.ShopDataDuplicateCheckScanService.DuplicateScanView; import org.junit.jupiter.api.BeforeEach; import org.junit.jupiter.api.Test; +import org.mockito.ArgumentCaptor; import java.util.List; @@ -87,6 +88,37 @@ class ShopDataDuplicateCheckImportServiceTest { verify(scanMapper).insert(any(ShopDataDuplicateScanEntity.class)); } + /** 基线 payload:郭悦晴那行的品牌带着改口径之前的标签前缀(这一店不在本次导入的店铺里,会被原样重插)。 */ + private DuplicateScanView legacyBrandBaselineView() { + DuplicateOccurrence kept = new DuplicateOccurrence("B09PNTZM4L", "2026-08-19 05:48", "10.1", + "Brand: HANYTON", "郭悦晴", "G2", List.of("DE"), "DE"); + DuplicateShop shop = new DuplicateShop("郭悦晴", "G2", List.of("DE"), 1, 1); + return new DuplicateScanView("2026-09-04 00:00:00", null, new DuplicateScanPayload( + List.of(shop), List.of(new DuplicateItem("B09PNTZM4L", 1, 1, List.of(kept))))); + } + + /** + * 导入的基线行读自上一份落库 payload,改口径之前它带着 "Brand: " 前缀;原样重插就是替用户把 + * 脏值写回扫描表一次(下载再上传同构的问题)。摘掉 applyToLatest 里的 strip,下面这条会红。 + */ + @Test + void addImportDoesNotReseedLegacyBrandLabelIntoNewScanRow() throws Exception { + when(scanService.loadLatest()).thenReturn(legacyBrandBaselineView()); + when(scanMapper.insert(any(ShopDataDuplicateScanEntity.class))).thenAnswer(invocation -> { + invocation.getArgument(0).setId(99L); + return 1; + }); + + service.importExcel("小曾", "UK", workbook(), false); + + ArgumentCaptor inserted = + ArgumentCaptor.forClass(ShopDataDuplicateScanEntity.class); + verify(scanMapper).insert(inserted.capture()); + assertThat(inserted.getValue().getPayloadJson()) + .doesNotContain("Brand:") + .contains("HANYTON"); + } + @Test void importWithoutBaselineFails() { when(scanService.loadLatest()).thenReturn(null); diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryServiceTest.java index 02f489ab7..5efba1124 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckQueryServiceTest.java @@ -4,13 +4,18 @@ import com.nanri.aiimage.modules.shopduplicatecheck.TestFixtures; import com.nanri.aiimage.modules.shopduplicatecheck.mapper.ShopDuplicateCheckSourceMapper; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateItem; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateOccurrence; +import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateScanPayload; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateScanSummary; import com.nanri.aiimage.modules.shopduplicatecheck.model.payload.DuplicateShop; +import org.apache.poi.ss.usermodel.Sheet; +import org.apache.poi.xssf.usermodel.XSSFWorkbook; import org.junit.jupiter.api.BeforeEach; import org.junit.jupiter.api.Test; +import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.nio.charset.StandardCharsets; +import java.util.ArrayList; import java.util.List; import java.util.Map; import java.util.Set; @@ -265,6 +270,39 @@ class ShopDataDuplicateCheckQueryServiceTest { assertThat(lines[1]).doesNotContain(",UK,"); } + /** + * 导出文件的品牌列必须是剥过标签前缀的:扫描行是改口径之前落库的,payload 里的品牌仍带 + * "Brand: "(亚马逊详情页复制来的署名),而出件是用户的下一份上传物——这一格不洗, + * 下载再上传就把脏值送回链路里。摘掉 buildExportRows 里的 strip,这一格读回来就是 "Brand: NIKE"。 + */ + @Test + void exportBrandColumnLosesTheAmazonBylinePrefix() throws Exception { + ShopDataDuplicateCheckScanService.DuplicateScanView legacyPayload = + new ShopDataDuplicateCheckScanService.DuplicateScanView( + "2026-09-04 03:10:00", null, new DuplicateScanPayload(List.of(), List.of( + new DuplicateItem("B0BRAND01", 2, 2, List.of( + legacyOccurrence("ShopA", "Brand: NIKE"), + legacyOccurrence("ShopB", "Brands: NIKE")))))); + + ByteArrayOutputStream out = new ByteArrayOutputStream(); + service.writeExportXlsx(legacyPayload, null, "all", noFilter(), out); + + List brands = new ArrayList<>(); + try (XSSFWorkbook workbook = new XSSFWorkbook(new ByteArrayInputStream(out.toByteArray()))) { + Sheet sheet = workbook.getSheetAt(0); + for (int r = 1; r <= sheet.getLastRowNum(); r++) { + brands.add(sheet.getRow(r).getCell(7).getStringCellValue()); + } + } + // "Brands:" 是真实品牌写法,只剥行首的那个标签 + assertThat(brands).containsExactlyInAnyOrder("NIKE", "Brands: NIKE"); + } + + private static DuplicateOccurrence legacyOccurrence(String shopName, String brand) { + return new DuplicateOccurrence("B0BRAND01", "2026-08-30", "GBP 9.99", brand, + shopName, "GroupA", List.of("UK"), "UK"); + } + @Test void groupFilterNarrowsToGroupMembers() { // 分组筛选:GroupA 只保留 ShopA/ShopB 键。 diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanServiceTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanServiceTest.java index d4bf545f7..233ac72b9 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanServiceTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/ShopDataDuplicateCheckScanServiceTest.java @@ -49,4 +49,22 @@ class ShopDataDuplicateCheckScanServiceTest { void buildParsedShops_emptyInputReturnsEmpty() { assertTrue(ShopDataDuplicateCheckScanService.buildParsedShops(List.of(), Map.of()).isEmpty()); } + + /** + * 明细表的品牌是改口径之前落库的(采集端原样写入亚马逊署名行),扫描直接查这张表: + * 标签前缀要在进聚合这一端剥掉,落库 payload、台账页面与出件拿到的才是同一个干净值。 + * 摘掉 buildParsedShops 里的 strip,下面第一条断言读回来的就是 "Brand: NIKE"。 + */ + @Test + void buildParsedShops_stripsAmazonBylinePrefixFromDbBrandColumn() { + List parsed = ShopDataDuplicateCheckScanService.buildParsedShops( + List.of(new ShopLatestItemRowDto("店A", "UK", "B0001", "2026-09-01 10:00", "9.90", "Brand: NIKE"), + new ShopLatestItemRowDto("店A", "UK", "B0002", "2026-09-01 10:00", "9.90", "品牌:耐克"), + new ShopLatestItemRowDto("店A", "UK", "B0003", "2026-09-01 10:00", "9.90", "Brands: NIKE")), + Map.of()); + + assertEquals(List.of("NIKE", "耐克", "Brands: NIKE"), + parsed.get(0).rows().stream().map(RawRow::brand).toList(), + "行首标签剥掉,「Brands:」这类真实品牌写法原样保留"); + } } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParserTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParserTest.java index dac016c71..72ce0ce04 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParserTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/shopduplicatecheck/service/support/DuplicateCheckWorkbookParserTest.java @@ -75,4 +75,29 @@ class DuplicateCheckWorkbookParserTest { new String[]{"2026-08-15", "9.99"}); assertEquals(0, DuplicateCheckWorkbookParser.parse(bytes).size(), "无 ASIN 列的 sheet 跳过"); } + + /** + * 品牌列读入即剥标签前缀:用户上传的结果文件是亚马逊详情页复制来的,"Brand: XXX" 带着标签 + * 进聚合就一路带到 payload、台账与出件。摘掉解析这一端的 strip,这里读到的就是原始署名值。 + */ + @Test + void stripsAmazonBylinePrefixFromBrandColumn() throws Exception { + byte[] bytes = workbookWith(false, "英国", + new String[]{"ASIN", "日期", "价格", "品牌"}, + new String[]{"B0BRAND01", "2026-08-15", "£34.46", "Brand: NIKE"}); + assertEquals("NIKE", DuplicateCheckWorkbookParser.parse(bytes).get(0).brand(), + "行首标签在读入这一端剥掉"); + } + + /** "Brands:" 是真实品牌写法,不是字段标签;同一份文件里的价格/日期列不受品牌清洗影响。 */ + @Test + void keepsLookalikeBrandAndOtherColumnsVerbatim() throws Exception { + byte[] bytes = workbookWith(false, "英国", + new String[]{"ASIN", "日期", "价格", "品牌"}, + new String[]{"B0BRAND02", "Brand: 2026-08-15", "£Brand: 34.46", "Brands: NIKE"}); + var row = DuplicateCheckWorkbookParser.parse(bytes).get(0); + assertEquals("Brands: NIKE", row.brand(), "近似标签原样保留"); + assertEquals("Brand: 2026-08-15", row.date(), "非品牌列不参与清洗"); + assertEquals("£Brand: 34.46", row.price(), "非品牌列不参与清洗"); + } } diff --git a/backend-java/src/test/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilderTest.java b/backend-java/src/test/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilderTest.java index 7457f6c90..7f930245c 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilderTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/modules/similarasin/service/support/SimilarAsinSheetBuilderTest.java @@ -209,6 +209,21 @@ class SimilarAsinSheetBuilderTest { assertEquals("品牌Y", cellString(workbook, 1, 5), "品牌按包含匹配"); } + @Test + void test_build_brand_label_prefix_stripped() { + // 出件是用户的下一份上传物:亚马逊页面复制带上的 "Brand: " 前缀不在此剥掉,二次上传就把脏值送回链路 + Workbook workbook = build(List.of(row("1", "B01A", "US", "19.9", "卖家1", "Brand: NIKE"))); + + assertEquals("NIKE", cellString(workbook, 1, 5), "品牌列剥掉行首标签前缀"); + } + + @Test + void test_build_brand_lookalike_kept() { + Workbook workbook = build(List.of(row("1", "B01A", "US", "19.9", "卖家1", "Brands: NIKE"))); + + assertEquals("Brands: NIKE", cellString(workbook, 1, 5), "近似标签是真实品牌写法,不随行首清洗被误剥"); + } + @Test void test_build_null_input() { Workbook workbook = build(null); diff --git a/backend-java/src/test/resources/shopdatacrawl/golden/rows-snapshot.txt b/backend-java/src/test/resources/shopdatacrawl/golden/rows-snapshot.txt index 1956e3622..5d5d4e497 100644 --- a/backend-java/src/test/resources/shopdatacrawl/golden/rows-snapshot.txt +++ b/backend-java/src/test/resources/shopdatacrawl/golden/rows-snapshot.txt @@ -1,3 +1,3 @@ rows=2 日期=2026-07-25;ASIN=B012345678;商品图片=;库存销量=128;销售排名=#1,245;页面浏览量=3560;售出件数=42;价格=GBP 19.99;推荐报价=GBP 18.99;品牌=Example Brand; - 日期=;ASIN=B099999999;商品图片=;库存销量=;销售排名=;页面浏览量=;售出件数=;价格=;推荐报价=;品牌= Brand With Spaces ; + 日期=;ASIN=B099999999;商品图片=;库存销量=;销售排名=;页面浏览量=;售出件数=;价格=;推荐报价=;品牌=Brand With Spaces;