diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/AppearancePatentTaskService.java b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/AppearancePatentTaskService.java index 319439bc9..fd277d754 100644 --- a/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/AppearancePatentTaskService.java +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/AppearancePatentTaskService.java @@ -31,6 +31,7 @@ import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentParse import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentTaskBatchVo; import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentTaskDetailVo; import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentTaskItemVo; +import com.nanri.aiimage.modules.appearancepatent.service.support.AppearancePatentResultPackager; import com.nanri.aiimage.modules.appearancepatent.service.support.AppearancePatentRowOutcome; import com.nanri.aiimage.modules.appearancepatent.service.support.AppearancePatentSheetBuilder; import com.nanri.aiimage.modules.appearancepatent.service.support.AppearancePatentStaleTaskScanner; @@ -2066,7 +2067,7 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline parsedRows = parsed.getAllItems() == null ? List.of() : parsed.getAllItems(); - List sourceRows = splitRowsBySourceFile(parsed, parsedRows, result.getSourceFilename()); + List sourceRows = splitRowsBySourceFile(parsed, parsedRows, result.getSourceFilename()); if (sourceRows.isEmpty()) { throw new BusinessException("外观专利检测结果为空,请稍后重试生成结果文件"); } @@ -2079,10 +2080,10 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline workbooks = new ArrayList<>(); + List workbooks = new ArrayList<>(); File zip = null; try { - for (SourceRows item : sourceRows) { + for (AppearancePatentResultPackager.SourceRows item : sourceRows) { Map resultMap = loadPersistedResultRowsForRowsWithRetry(task.getId(), item.rows()); // 覆盖率先于「本文件无结果就跳过」判定:整源文件的 LLM 结果全丢(分片被跳过、 @@ -2118,7 +2119,7 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline 0) { log.info("[appearance-patent] group propagate conclusion taskId={} updatedRows={}", task.getId(), conclusionPropagated); @@ -2136,7 +2137,7 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline splitRowsBySourceFile(AppearancePatentParsedPayloadDto parsed, - List receivedRows, - String fallbackFilename) { - String defaultFilename = firstNonBlank(fallbackFilename, "appearance-patent"); - Map builders = new LinkedHashMap<>(); - if (parsed != null && parsed.getSourceFiles() != null) { - for (AppearancePatentSourceFileDto sourceFile : parsed.getSourceFiles()) { - if (sourceFile == null) { - continue; - } - String sourceFileKey = normalize(sourceFile.getFileKey()); - if (sourceFileKey.isBlank()) { - continue; - } - builders.putIfAbsent(sourceFileKey, - new SourceRowsBuilder(sourceFileKey, firstNonBlank(sourceFile.getOriginalFilename(), defaultFilename))); - } - } - - for (AppearancePatentParsedRowVo row : receivedRows == null ? List.of() : receivedRows) { - if (row == null) { - continue; - } - String sourceFileKey = normalize(row.getSourceFileKey()); - if (sourceFileKey.isBlank()) { - sourceFileKey = normalize(row.getSourceFilename()); - } - if (sourceFileKey.isBlank()) { - sourceFileKey = "__default__"; - } - SourceRowsBuilder builder = builders.computeIfAbsent(sourceFileKey, - key -> new SourceRowsBuilder(key, firstNonBlank(row.getSourceFilename(), defaultFilename))); - if (builder.sourceFilename().isBlank() || Objects.equals(builder.sourceFilename(), defaultFilename)) { - builder.setSourceFilename(firstNonBlank(row.getSourceFilename(), defaultFilename)); - } - builder.rows().add(row); - } - - return builders.values().stream() - .filter(builder -> !builder.rows().isEmpty()) - .map(builder -> new SourceRows(builder.sourceFileKey(), firstNonBlank(builder.sourceFilename(), defaultFilename), builder.rows())) - .toList(); + /** + * C 聚类·打包与命名住在 {@link AppearancePatentResultPackager}(3A 拆薄第四笔)。 + * 无状态、按调用现取,与 {@code historyAssembler()} 同一先例,所以本笔不动构造器、不动装配。 + * 三枚薄委托不是为了保住反射接缝(AP 的反射点名在 {@code writeResultWorkbook} 那一侧, + * 本笔搬的四枚方法没有被任何测试按名反射过),而是为了不让六个调用点跟着改形状。 + */ + private AppearancePatentResultPackager resultPackager() { + return new AppearancePatentResultPackager(); } - private void packageResultWorkbooksAsZip(File zip, List workbooks) { - Set entryNames = new LinkedHashSet<>(); - try (ZipOutputStream zos = new ZipOutputStream(new FileOutputStream(zip))) { - byte[] buffer = new byte[8192]; - for (SourceResultWorkbook workbook : workbooks) { - writeResultZipEntry(zos, buffer, workbook.file(), uniqueZipEntryName(workbook.filename(), entryNames)); - } - } catch (IOException ex) { - throw new BusinessException("生成外观专利检测结果压缩包失败"); - } + private List splitRowsBySourceFile( + AppearancePatentParsedPayloadDto parsed, + List receivedRows, + String fallbackFilename) { + return resultPackager().splitRowsBySourceFile(parsed, receivedRows, fallbackFilename); } - private void writeResultZipEntry(ZipOutputStream zos, - byte[] buffer, - File file, - String entryName) throws IOException { - zos.putNextEntry(new ZipEntry(entryName)); - try (FileInputStream inputStream = new FileInputStream(file)) { - int len; - while ((len = inputStream.read(buffer)) > 0) { - zos.write(buffer, 0, len); - } - } - zos.closeEntry(); + private void packageResultWorkbooksAsZip(File zip, + List workbooks) { + resultPackager().packageResultWorkbooksAsZip(zip, workbooks); } - private String uniqueZipEntryName(String filename, Set entryNames) { - String safeFilename = normalizeZipFilename(filename); - if (entryNames.add(safeFilename)) { - return safeFilename; - } - int dotIndex = safeFilename.lastIndexOf('.'); - String stem = dotIndex > 0 ? safeFilename.substring(0, dotIndex) : safeFilename; - String extension = dotIndex > 0 ? safeFilename.substring(dotIndex) : ""; - int index = 2; - while (true) { - String candidate = stem + "(" + index + ")" + extension; - if (entryNames.add(candidate)) { - return candidate; - } - index++; - } - } - - private String normalizeZipFilename(String filename) { - String normalized = firstNonBlank(filename, "appearance-patent-result.xlsx").replace('\\', '/'); - while (normalized.startsWith("/")) { - normalized = normalized.substring(1); - } - return normalized.isBlank() ? "appearance-patent-result.xlsx" : normalized; + private String safeFileStem(String filename) { + return resultPackager().safeFileStem(filename); } private List filterReceivedParsedRows(List parsedRows, @@ -3015,16 +2946,6 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline 0) { - name = name.substring(0, idx); - } - String safe = name.replaceAll("[\\\\/:*?\"<>|]", "_").trim(); - return safe.isBlank() ? "appearance-patent" : safe; - } - /** 只读收集任务范围/分片载荷指针,供事务提交后做远端删除。 */ private List collectTransientTaskPayloads(Long taskId) { if (taskId == null || taskId <= 0) { @@ -3119,43 +3040,6 @@ public class AppearancePatentTaskService extends AbstractFileTaskPipeline rows = new ArrayList<>(); - - private SourceRowsBuilder(String sourceFileKey, String sourceFilename) { - this.sourceFileKey = sourceFileKey; - this.sourceFilename = sourceFilename; - } - - private String sourceFileKey() { - return sourceFileKey; - } - - private String sourceFilename() { - return sourceFilename; - } - - private void setSourceFilename(String sourceFilename) { - this.sourceFilename = sourceFilename; - } - - private List rows() { - return rows; - } - } - - private record SourceRows(String sourceFileKey, - String sourceFilename, - List rows) { - } - - private record SourceResultWorkbook(File file, - String filename, - int rowCount) { - } - private record ParsedWorkbook(int totalRows, int droppedRows, List headers, List allRows) { } diff --git a/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentResultPackager.java b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentResultPackager.java new file mode 100644 index 000000000..f57a39aa2 --- /dev/null +++ b/backend-java/src/main/java/com/nanri/aiimage/modules/appearancepatent/service/support/AppearancePatentResultPackager.java @@ -0,0 +1,205 @@ +package com.nanri.aiimage.modules.appearancepatent.service.support; + +import com.nanri.aiimage.common.exception.BusinessException; +import com.nanri.aiimage.modules.appearancepatent.model.dto.AppearancePatentParsedPayloadDto; +import com.nanri.aiimage.modules.appearancepatent.model.dto.AppearancePatentSourceFileDto; +import com.nanri.aiimage.modules.appearancepatent.model.vo.AppearancePatentParsedRowVo; + +import java.io.File; +import java.io.FileInputStream; +import java.io.FileOutputStream; +import java.io.IOException; +import java.util.ArrayList; +import java.util.LinkedHashMap; +import java.util.LinkedHashSet; +import java.util.List; +import java.util.Map; +import java.util.Objects; +import java.util.Set; +import java.util.zip.ZipEntry; +import java.util.zip.ZipOutputStream; + +/** + * 外观专利结果侧的「按源文件分组 + 成品文件打包与命名」,从 + * {@code AppearancePatentTaskService} 原样搬移(3A 拆薄第四笔)。 + * + *

搬的是纯形状:分组键的取法({@code sourceFileKey} → 回落 {@code sourceFilename} → + * 再回落 {@code __default__})、zip 条目名的去重后缀 {@code (2)(3)…}、文件名里 Windows + * 非法字符换成下划线、空名兜底成 {@code appearance-patent}。判据、兜底值、异常文案一字未改。 + * + *

为什么这一簇先抽。它是 C 聚类里唯一「零反射接缝、零库表写入」的一段—— + * {@code assembleResultWorkbook} 留在服务侧继续编排(要读分片、要落结果行、要挂 OSS), + * 这里只把「给它算出该打包什么、叫什么名、怎么压成 zip」这件事收成一枚无状态件。 + * 反面判据也查过:AP 的测试没有按名字反射过这四枚私有方法(反射点名清单里是 + * {@code writeResultWorkbook}/{@code validateCompleteLlmCoverage} 那一批),所以本笔 + * 不需要为保住接缝而留委托壳——服务侧那三枚薄委托只是为了不让六个调用点改形状。 + * + *

文本归一化不再抄第三份:分组里用的 {@code normalize}/{@code firstNonBlank} 直接引用 + * {@link AppearancePatentRowNormalizer}(该 util 的 javadoc 自陈与服务侧那一版「逐字节一致」, + * 本笔逐条对过,body 相同)。⚠ 顺带登记一条既有事实:服务至今仍有自己那两枚私有同名件, + * Task 93 抽出 util 之后没有将服务侧切过去——那是重复的第二遍,不是本笔引入的。 + */ +public class AppearancePatentResultPackager { + + private static final String DEFAULT_STEM = "appearance-patent"; + private static final String DEFAULT_RESULT_FILENAME = "appearance-patent-result.xlsx"; + private static final int ZIP_BUFFER_SIZE = 8192; + + /** 一个源文件对应一批待写行的分组(顺序即分组首次出现的顺序)。 */ + public record SourceRows(String sourceFileKey, + String sourceFilename, + List rows) { + } + + /** 一份已写好的成品:临时文件、对外文件名、行数(zip 与单文件两条路共用这个形状)。 */ + public record SourceResultWorkbook(File file, + String filename, + int rowCount) { + } + + /** + * 按源文件把收到的行分组:先按解析载荷里声明的文件建骨架(保持用户上传顺序), + * 再按行自身的 {@code sourceFileKey} 归位——取不到 key 时回落 {@code sourceFilename}, + * 两者都空才落 {@code __default__}。空组不输出。 + */ + public List splitRowsBySourceFile(AppearancePatentParsedPayloadDto parsed, + List receivedRows, + String fallbackFilename) { + String defaultFilename = AppearancePatentRowNormalizer.firstNonBlank(fallbackFilename, DEFAULT_STEM); + Map builders = new LinkedHashMap<>(); + if (parsed != null && parsed.getSourceFiles() != null) { + for (AppearancePatentSourceFileDto sourceFile : parsed.getSourceFiles()) { + if (sourceFile == null) { + continue; + } + String sourceFileKey = AppearancePatentRowNormalizer.normalize(sourceFile.getFileKey()); + if (sourceFileKey.isBlank()) { + continue; + } + builders.putIfAbsent(sourceFileKey, new SourceRowsBuilder(sourceFileKey, + AppearancePatentRowNormalizer.firstNonBlank(sourceFile.getOriginalFilename(), defaultFilename))); + } + } + + for (AppearancePatentParsedRowVo row : receivedRows == null + ? List.of() : receivedRows) { + if (row == null) { + continue; + } + String sourceFileKey = AppearancePatentRowNormalizer.normalize(row.getSourceFileKey()); + if (sourceFileKey.isBlank()) { + sourceFileKey = AppearancePatentRowNormalizer.normalize(row.getSourceFilename()); + } + if (sourceFileKey.isBlank()) { + sourceFileKey = "__default__"; + } + SourceRowsBuilder builder = builders.computeIfAbsent(sourceFileKey, + key -> new SourceRowsBuilder(key, AppearancePatentRowNormalizer.firstNonBlank( + row.getSourceFilename(), defaultFilename))); + if (builder.sourceFilename().isBlank() || Objects.equals(builder.sourceFilename(), defaultFilename)) { + builder.setSourceFilename(AppearancePatentRowNormalizer.firstNonBlank( + row.getSourceFilename(), defaultFilename)); + } + builder.rows().add(row); + } + + return builders.values().stream() + .filter(builder -> !builder.rows().isEmpty()) + .map(builder -> new SourceRows(builder.sourceFileKey(), + AppearancePatentRowNormalizer.firstNonBlank(builder.sourceFilename(), defaultFilename), + builder.rows())) + .toList(); + } + + /** 多份成品压一个 zip;条目名去重,读流失败即整包拒收(文案与原实现同句)。 */ + public void packageResultWorkbooksAsZip(File zip, List workbooks) { + Set entryNames = new LinkedHashSet<>(); + try (ZipOutputStream zos = new ZipOutputStream(new FileOutputStream(zip))) { + byte[] buffer = new byte[ZIP_BUFFER_SIZE]; + for (SourceResultWorkbook workbook : workbooks) { + writeResultZipEntry(zos, buffer, workbook.file(), uniqueZipEntryName(workbook.filename(), entryNames)); + } + } catch (IOException ex) { + throw new BusinessException("生成外观专利检测结果压缩包失败"); + } + } + + /** 文件名去掉扩展名、把 Windows 非法字符换成下划线;空结果兜底 {@code appearance-patent}。 */ + public String safeFileStem(String filename) { + String name = filename == null || filename.isBlank() ? DEFAULT_STEM : filename; + int idx = name.lastIndexOf('.'); + if (idx > 0) { + name = name.substring(0, idx); + } + String safe = name.replaceAll("[\\\\/:*?\"<>|]", "_").trim(); + return safe.isBlank() ? DEFAULT_STEM : safe; + } + + private void writeResultZipEntry(ZipOutputStream zos, + byte[] buffer, + File file, + String entryName) throws IOException { + zos.putNextEntry(new ZipEntry(entryName)); + try (FileInputStream inputStream = new FileInputStream(file)) { + int len; + while ((len = inputStream.read(buffer)) > 0) { + zos.write(buffer, 0, len); + } + } + zos.closeEntry(); + } + + private String uniqueZipEntryName(String filename, Set entryNames) { + String safeFilename = normalizeZipFilename(filename); + if (entryNames.add(safeFilename)) { + return safeFilename; + } + int dotIndex = safeFilename.lastIndexOf('.'); + String stem = dotIndex > 0 ? safeFilename.substring(0, dotIndex) : safeFilename; + String extension = dotIndex > 0 ? safeFilename.substring(dotIndex) : ""; + int index = 2; + while (true) { + String candidate = stem + "(" + index + ")" + extension; + if (entryNames.add(candidate)) { + return candidate; + } + index++; + } + } + + private String normalizeZipFilename(String filename) { + String normalized = AppearancePatentRowNormalizer.firstNonBlank(filename, DEFAULT_RESULT_FILENAME) + .replace('\\', '/'); + while (normalized.startsWith("/")) { + normalized = normalized.substring(1); + } + return normalized.isBlank() ? DEFAULT_RESULT_FILENAME : normalized; + } + + private static final class SourceRowsBuilder { + private final String sourceFileKey; + private String sourceFilename; + private final List rows = new ArrayList<>(); + + private SourceRowsBuilder(String sourceFileKey, String sourceFilename) { + this.sourceFileKey = sourceFileKey; + this.sourceFilename = sourceFilename; + } + + private String sourceFileKey() { + return sourceFileKey; + } + + private String sourceFilename() { + return sourceFilename; + } + + private void setSourceFilename(String sourceFilename) { + this.sourceFilename = sourceFilename; + } + + private List rows() { + return rows; + } + } +} diff --git a/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScanTest.java b/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScanTest.java index 56b72eedc..5209401a7 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScanTest.java +++ b/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScanTest.java @@ -266,13 +266,13 @@ class SourceScanTest { // ---------------------------------------------------------------- N6 moduleSources @Test - void moduleSources_appearancepatent返回30支含拆分后的support件() { + void moduleSources_appearancepatent返回31支含拆分后的support件() { List files = SourceScan.moduleSources("modules/appearancepatent"); assertThat(files) - .as("真实集成:对真仓目录求值。30 支=1 controller+1 client+1 service 主件+3 spi/cache+" - + "7 service/support+16 model(dto7/vo9),与 find 实测同号") - .hasSize(30); + .as("真实集成:对真仓目录求值。31 支=1 controller+1 client+1 service 主件+3 spi/cache+" + + "8 service/support+16 model(dto7/vo9),与 find 实测同号") + .hasSize(31); assertThat(files.stream().map(p -> p.toString().replace('\\', '/')).toList()) .as("support 目录里的既有件必须在面里——聚合入口存在的理由就是拆分后的新文件要自动被看到") .anyMatch(rel -> rel.endsWith("service/support/AppearancePatentRowNormalizer.java")); @@ -366,16 +366,22 @@ class SourceScanTest { } @Test - void recursive_包含service_support下的既有7件() { + void recursive_包含service_support下的既有8件() { List support = SourceScan.moduleSources("modules/appearancepatent").stream() .map(p -> p.toString().replace('\\', '/')) .filter(rel -> rel.contains("/service/support/")) .toList(); assertThat(support) - .as("真实集成:对真仓 modules/appearancepatent/service/support 求值,实测 7 支(3A 拆薄把归属谓词与陈旧扫描抽成两枚 support 件,5→7)。" + .as("真实集成:对真仓 modules/appearancepatent/service/support 求值,实测 8 支(3A 拆薄把归属谓词、陈旧扫描、结果打包抽成三枚 support 件,5→8)。" + "这一格就是聚合入口存在的理由——拆分出去的新文件必须自动进面") - .hasSize(7); + .hasSize(8); + assertThat(support) + .as("3A 拆出去的三枚新件必须被逐枚点名扫到:只把分母数字改大,下一回拆件又只剩「数不对」" + + "一记红,证不了「扫得到」这件事") + .anyMatch(rel -> rel.endsWith("service/support/AppearancePatentTaskOwnership.java")) + .anyMatch(rel -> rel.endsWith("service/support/AppearancePatentStaleTaskScanner.java")) + .anyMatch(rel -> rel.endsWith("service/support/AppearancePatentResultPackager.java")); } @Test