diff --git a/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScan.java b/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScan.java index 204048bc6..d603d4b76 100644 --- a/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScan.java +++ b/backend-java/src/test/java/com/nanri/aiimage/testutil/SourceScan.java @@ -2,8 +2,12 @@ package com.nanri.aiimage.testutil; import java.io.IOException; import java.nio.charset.StandardCharsets; +import java.nio.file.FileVisitOption; +import java.nio.file.FileVisitResult; import java.nio.file.Files; import java.nio.file.Path; +import java.nio.file.SimpleFileVisitor; +import java.nio.file.attribute.BasicFileAttributes; import java.util.List; import java.util.regex.Matcher; import java.util.regex.Pattern; @@ -60,21 +64,14 @@ public final class SourceScan { } Path base = moduleRoot(); List sites = new java.util.ArrayList<>(); - try (Stream walk = Files.walk(root)) { - List files = walk.filter(Files::isRegularFile) - .filter(p -> p.getFileName().toString().endsWith(".java")) - .sorted(java.util.Comparator.comparing(p -> relativizeForCensus(p, root, base).replace('\\', '/'))) - .toList(); - for (Path file : files) { - String code = withoutComments(read(file)); - String relative = relativizeForCensus(file, root, base).replace('\\', '/'); - Matcher matcher = CALL_SITE.matcher(code); - while (matcher.find()) { - sites.add(new CallSite(matcher.group(1), relative, lineAt(code, matcher.start()))); - } + // N7 之后普查与聚合共用同一条遍历(含隐藏目录/重解析点排除),免得两套口径读出不一样多的文件 + for (Path file : collectJavaFiles(root)) { + String code = withoutComments(read(file)); + String relative = relativizeForCensus(file, root, base).replace('\\', '/'); + Matcher matcher = CALL_SITE.matcher(code); + while (matcher.find()) { + sites.add(new CallSite(matcher.group(1), relative, lineAt(code, matcher.start()))); } - } catch (IOException e) { - throw new IllegalStateException("遍历 " + root + " 失败: " + e.getMessage(), e); } return List.copyOf(sites); } @@ -188,17 +185,69 @@ public final class SourceScan { return List.copyOf(collectJavaFiles(module)); } - private static List collectJavaFiles(Path root) { + /** + * 收集 {@code root} 下任意深度的 {@code .java}(N7 定递归语义)。用 {@code walkFileTree} 而不是 + * {@code Files.walk},因为三种失败模式必须在进入之前决定,事后 filter 来不及: + *
    + *
  1. 隐藏目录不进面:{@code .git/}、{@code .idea/} 里的 {@code .java} 是工具生成的副本, + * 进面等于同一份代码被数两遍,精确计数格(publish 恰 1)当场失真。
  2. + *
  3. 重解析点不进面:本机实测 {@code Files.walk} 会钻进 Windows 的 junction + * (用例读出 {@code [One.java, One.java]}),"默认不 FOLLOW_LINKS 所以安全"这条假设已被证伪。 + * {@code dos:ReparsePoint} 属性是公开 API 里唯一的识别手段({@code DosFileAttributes} 接口上没有 isReparsePoint())。
  4. + *
  5. 遍历失败必须抛:{@code visitFileFailed} 与"根本不存在"两条道都包成 + * {@code IllegalStateException} 并点名路径——吞成空表就是「读不到 = 没违规」。
  6. + *
+ */ + static List collectJavaFiles(Path root) { + Path start = root.toAbsolutePath(); + java.util.Map deduped = new java.util.LinkedHashMap<>(); + try { + Files.walkFileTree(start, java.util.EnumSet.noneOf(FileVisitOption.class), Integer.MAX_VALUE, + new SimpleFileVisitor() { + @Override + public FileVisitResult preVisitDirectory(Path dir, BasicFileAttributes attrs) { + return !dir.equals(start) && isExcludedDirectory(dir) + ? FileVisitResult.SKIP_SUBTREE + : FileVisitResult.CONTINUE; + } - try (Stream walk = Files.walk(root)) { - return walk.filter(Files::isRegularFile) - .filter(p -> p.getFileName().toString().endsWith(".java")) - .sorted(java.util.Comparator.comparing( - p -> root.relativize(p).toString().replace('\\', '/'))) - .toList(); + @Override + public FileVisitResult visitFile(Path file, BasicFileAttributes attrs) { + String name = file.getFileName().toString(); + if (name.endsWith(".java") && !name.startsWith(".")) { + deduped.putIfAbsent(realPathOf(file), file); + } + return FileVisitResult.CONTINUE; + } + + @Override + public FileVisitResult visitFileFailed(Path file, IOException exc) { + throw new IllegalStateException("遍历 " + start + " 时在 " + file + " 失败: " + + exc.getMessage(), exc); + } + }); } catch (IOException e) { - throw new IllegalStateException("遍历 " + root + " 失败: " + e.getMessage(), e); + throw new IllegalStateException("遍历 " + start + " 失败: " + e.getMessage(), e); } + List found = new java.util.ArrayList<>(deduped.values()); + found.sort(java.util.Comparator.comparing(p -> start.relativize(p).toString().replace('\\', '/'))); + return List.copyOf(found); + } + + /** 链接折叠用的键;拿不到实路径(并发被删等)就退回绝对路径,宁可少折一次也不吞异常。 */ + private static Path realPathOf(Path file) { + try { + return file.toRealPath(); + } catch (IOException e) { + return file.toAbsolutePath().normalize(); + } + } + + private static boolean isExcludedDirectory(Path dir) { + if (dir.getFileName().toString().startsWith(".") || Files.isSymbolicLink(dir)) { + return true; + } + return false; } /** 按文件名(不含目录)取 {@code src/main/java} 下那份源码的全文字符串。 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 4454ce7ec..f553d08ca 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 @@ -354,4 +354,137 @@ class SourceScanTest { .hasMessageContaining("cwd=") .hasMessageContaining(Path.of("").toAbsolutePath().toString()); } + + // ---------------------------------------------------------------- N7 递归语义 + + @Test + void recursive_包含service下的直接件() { + assertThat(SourceScan.moduleSources("modules/appearancepatent")) + .as("模块根下第一层的 service/ 直接件必须在面里") + .anyMatch(p -> p.toString().replace('\\', '/').endsWith( + "service/AppearancePatentTaskService.java")); + } + + @Test + void recursive_包含service_support下的既有5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 5 支。" + + "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(5); + } + + @Test + void recursive_包含model下两层dto与vo() { + List rel = SourceScan.moduleSources("modules/appearancepatent").stream() + .map(p -> p.toString().replace('\\', '/')) + .toList(); + + assertThat(rel.stream().filter(p -> p.contains("/model/dto/")).count()) + .as("model/dto 一层嵌套") + .isEqualTo(7); + assertThat(rel.stream().filter(p -> p.contains("/model/vo/")).count()) + .as("model/vo 一层嵌套") + .isEqualTo(9); + } + + @Test + void recursive_深度为1的目录不出现在面里() throws IOException { + Path module = Files.createDirectories(tmp.resolve("N7深度树/src/main/java/modules/deep")); + Path deep = Files.createDirectories(module.resolve("a/b/c")); + Files.writeString(deep.resolve("D.java"), "class D {}", StandardCharsets.UTF_8); + Files.writeString(module.resolve("Root.java"), "class Root {}", StandardCharsets.UTF_8); + + List files = SourceScan.moduleSources(tmp.resolve("N7深度树/src/main/java"), "modules/deep"); + + assertThat(files) + .as("三层嵌套里的 a/b/c/D.java 必须被计入:只爬一层的实现会静默漏掉它," + + "而 support/ 恰恰就是这一层以下的位置") + .extracting(p -> p.getFileName().toString()) + .containsExactlyInAnyOrder("D.java", "Root.java"); + assertThat(files) + .as("面里只许有文件——目录本身不是元素,否则聚合文本会多出无内容条目") + .allMatch(Files::isRegularFile); + } + + @Test + void recursive_隐藏目录件被排除() throws IOException { + Path module = Files.createDirectories(tmp.resolve("N7隐藏树/src/main/java/modules/vis")); + Path hidden = Files.createDirectories(module.resolve(".git")); + Files.writeString(hidden.resolve("Ghost.java"), "class Ghost {}", StandardCharsets.UTF_8); + Files.writeString(module.resolve("Real.java"), "class Real {}", StandardCharsets.UTF_8); + + assertThat(SourceScan.moduleSources(tmp.resolve("N7隐藏树/src/main/java"), "modules/vis")) + .as(".git / .idea 这类目录里的 .java(IDE 与工具生成的副本)不许进面——" + + "否则同一份代码会被算两遍,精确计数格(publish 恰 1)当场失真") + .extracting(p -> p.getFileName().toString()) + .containsExactly("Real.java"); + } + + @Test + void recursive_符号链接目录不被重复计入() throws IOException, InterruptedException { + Path javaRoot = Files.createDirectories(tmp.resolve("N7链树/src/main/java")); + Path real = Files.createDirectories(javaRoot.resolve("modules/linkreal")); + Files.writeString(real.resolve("One.java"), "class One {}", StandardCharsets.UTF_8); + // Windows 上建符号链接要特权,junction 不用:mklink /J 是本机的可复现路径。 + // ⚠ 这里刻意不用 Assumptions 跳过——跳过会新增一格 Skipped,直接违反 N99「一个新 skip 都不许有」, + // 那是拿跳过换绿。建不出来就判红并点名环境缺口(plan N3 第 9 例同一条口径)。 + Process mk = new ProcessBuilder("cmd.exe", "/c", "mklink", "/J", + javaRoot.resolve("modules").resolve("linkalias").toString(), real.toString()) + .redirectErrorStream(true).start(); + String out; + try (var in = mk.getInputStream()) { + out = new String(in.readAllBytes(), StandardCharsets.UTF_8); + } + int mklink = mk.waitFor(); + assertThat(mklink) + .as("junction 建不出来 ⇒ 本机环境不支持,这格判红而不是跳过(跳过会新增 Skipped," + + "违反 N99「一个新 skip 都不许有」)。cmd 输出:%s", out) + .isZero(); + + List files = SourceScan.moduleSources(javaRoot, "modules"); + + assertThat(files) + .as("目录联接指回同一棵树:本机实测 Files.walk 会照钻 junction(改之前读出 [One.java, One.java])," + + "所以「默认不 FOLLOW_LINKS 就安全」这条假设已被证伪。实现按 toRealPath() 折叠同一份文件" + + "(公开 NIO 读不到重解析点标志);重复计入会让精确计数格(publish 恰 1)假绿") + .extracting(p -> p.getFileName().toString()) + .containsExactly("One.java"); + } + + @Test + void recursive_目录内全是非java时命中为空() throws IOException { + Path module = Files.createDirectories(tmp.resolve("N7非java树/src/main/java/modules/docs")); + Files.writeString(module.resolve("README.md"), "# 迁移说明", StandardCharsets.UTF_8); + Files.createDirectories(module.resolve("sub")); + Files.writeString(module.resolve("sub").resolve("notes.txt"), "x", StandardCharsets.UTF_8); + + assertThat(SourceScan.moduleSources(tmp.resolve("N7非java树/src/main/java"), "modules/docs")) + .as("整棵树没有 .java ⇒ 空表(N11 会把这条改成抛;本格先钉「非 java 不进面」)") + .isEmpty(); + } + + @Test + void recursive_遍历失败抛IllegalStateException不吞() throws IOException { + Path vanished = tmp.resolve("N7消失树"); // 刻意不创建:walk 必须在第一步就抛 IOException + + // 正对照:先证明这里真的会抛 IOException —— 否则「没吞」这条断言是空的 + try (var raw = java.nio.file.Files.walk(vanished)) { + raw.toList(); + org.junit.jupiter.api.Assertions.fail("JDK 的 Files.walk 在这条路径上没抛,正对照失效"); + } catch (java.io.IOException expected) { + assertThat(expected).isInstanceOf(java.nio.file.NoSuchFileException.class); + } + + assertThatThrownBy(() -> SourceScan.collectJavaFiles(vanished)) + .as("遍历抛 IOException 时必须包成 IllegalStateException 并点名路径,不许吞成空表——" + + "「扫不到」被读成「没违规」是本族护栏的标准死法(10-15 §12 第一条)") + .isInstanceOf(IllegalStateException.class) + .hasMessageContaining("遍历") + .hasMessageContaining(vanished.getFileName().toString()); + } }