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 d00ce6beb..8697083ed 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 @@ -95,6 +95,57 @@ public final class SourceScan { return line; } + /** {@code src/main/java} 的绝对根。定位不到必须抛——理由与 {@link #mainSources()} 同一条。 */ + public static Path mainSourcesRoot() { + Path root = moduleRoot().resolve("src/main/java"); + if (!Files.isDirectory(root)) { + throw new IllegalStateException("源码根不存在: " + root + "(cwd=" + Path.of("").toAbsolutePath() + ")"); + } + return root; + } + + /** 主源码面全部 {@code .java},按相对路径字典序(与 {@link #censusSourceTextCallSites} 同一条排序键)。 */ + public static List mainSourceFiles() { + return collectJavaFiles(mainSourcesRoot()); + } + + /** 同名面普查:{@code src/main/java} 下裸文件名出现 >1 次的那些。 */ + public static java.util.Map> duplicateMainSourceNames() { + return duplicateSourceNames(mainSourcesRoot()); + } + + /** + * 分组键是裸文件名——因为 {@link #sourceText(String)} 就是按裸名找文件的, + * 两个包下放同名件,那个助手就会命中 >1(N20 要把它改成歧义即抛)。 + * 只算 {@code .java}:{@code Foo.md} 与 {@code Foo.java} 不同名,把文档也算进来会让聚合面多吃一支。 + */ + public static java.util.Map> duplicateSourceNames(Path treeRoot) { + Path root = treeRoot.toAbsolutePath(); + if (!Files.isDirectory(root)) { + throw new IllegalStateException("同名面普查的树根不是目录: " + root + "(cwd=" + + Path.of("").toAbsolutePath() + ")"); + } + java.util.Map> grouped = new java.util.TreeMap<>(); + for (Path file : collectJavaFiles(root)) { + grouped.computeIfAbsent(file.getFileName().toString(), k -> new java.util.ArrayList<>()) + .add(root.relativize(file).toString().replace('\\', '/')); + } + grouped.entrySet().removeIf(entry -> entry.getValue().size() < 2); + return java.util.Map.copyOf(grouped); + } + + private static List collectJavaFiles(Path root) { + 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(); + } catch (IOException e) { + throw new IllegalStateException("遍历 " + root + " 失败: " + e.getMessage(), e); + } + } + /** 按文件名(不含目录)取 {@code src/main/java} 下那份源码的全文字符串。 * ⚠️ 返回 String,与 InternalTokenSingleOutletTest 里那个返回 Path 的私有 {@code sourceOf} 不是一回事, * 别指望复用它的——那个是 private。 */ 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 f2dda8203..9b4e7090c 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 @@ -175,4 +175,91 @@ class SourceScanTest { .hasMessageContaining("cwd=") .hasMessageContaining(wrong.toAbsolutePath().toString()); } + + // ---------------------------------------------------------------- N5 同名面普查 + + @Test + void duplicateNames_实测三组且不多不少() { + assertThat(SourceScan.duplicateMainSourceNames().keySet()) + .as("src/main/java 下同名 >1 的裸文件名——实测就这三组,多一组说明有新落点,少一组说明有人搬走了") + .containsExactlyInAnyOrder( + "BoundedImageCache.java", + "GroupDeletionGuard.java", + "SkipPriceAsinPageVo.java"); + } + + @Test + void duplicateNames_与点名单交集为空() { + Set duplicates = SourceScan.duplicateMainSourceNames().keySet(); + + assertThat(duplicates) + .as("这是 N20 把 sourceText 收紧成「歧义即抛」的前提证明:今天被点名的 10 个裸名" + + "全部单命中,所以当场不会打红。若未来第 11 个点名撞上这三组之一," + + "N20 会当场红——那是正确行为,不是回归") + .doesNotContainAnyElementsOf(REGISTERED_BARE_NAMES); + } + + @Test + void duplicateNames_全树java数不少于900() { + assertThat(SourceScan.mainSourceFiles()) + .as("分母下限:扫到 0 支必须被读成尺子坏了,而不是「主源码空了」") + .hasSizeGreaterThanOrEqualTo(900); + } + + @Test + void duplicateNames_唯一名数小于总文件数() { + List all = SourceScan.mainSourceFiles(); + long uniqueNames = all.stream() + .map(p -> p.getFileName().toString()) + .distinct() + .count(); + + assertThat(uniqueNames) + .as("973 支文件 / 970 个唯一名 ⇒ 确有 3 组同名。相等就说明同名面被搬空了,N5 的台账要同步改") + .isLessThan((long) all.size()); + } + + @Test + void duplicateNames_空目录返回空集合() throws IOException { + Path empty = Files.createDirectories(tmp.resolve("空树")); + + assertThat(SourceScan.duplicateSourceNames(empty)) + .as("目录存在且没有 .java ⇒ 空 Map(与 N4 的「语料真空」同一条口径)") + .isEmpty(); + } + + @Test + void duplicateNames_同名恰好三支时计数为3() throws IOException { + for (String dir : new String[] {"a", "b/c", "d/e/f"}) { + Path sub = Files.createDirectories(tmp.resolve("三支").resolve(dir)); + Files.writeString(sub.resolve("Trio.java"), "class Trio {}", StandardCharsets.UTF_8); + } + + assertThat(SourceScan.duplicateSourceNames(tmp.resolve("三支"))) + .as("三支同名必须聚成一组、值为 3 条路径;分组写成「每对相邻比较」就会数出 3 组") + .containsEntry("Trio.java", List.of( + "a/Trio.java", "b/c/Trio.java", "d/e/f/Trio.java")); + } + + @Test + void duplicateNames_忽略非java文件同名() throws IOException { + Path tree = Files.createDirectories(tmp.resolve("混放树")); + Files.createDirectories(tree.resolve("docs")); + Files.writeString(tree.resolve("Foo.java"), "class Foo {}", StandardCharsets.UTF_8); + Files.writeString(tree.resolve("docs").resolve("Foo.md"), "# Foo", StandardCharsets.UTF_8); + + assertThat(SourceScan.duplicateSourceNames(tree)) + .as("`Foo.md` 与 `Foo.java` 不同名——只按裸名分组会把文档也算成同名,聚合面就多吃一支") + .isEmpty(); + } + + @Test + void duplicateNames_路径不可读时抛不静默跳过() throws IOException { + Path notADirectory = Files.writeString(tmp.resolve("我是文件不是目录"), "x", StandardCharsets.UTF_8); + + assertThatThrownBy(() -> SourceScan.duplicateSourceNames(notADirectory)) + .as("传进来的是文件而非目录 ⇒ 必须抛。静默返回空 Map 会被读成「主源码没有同名件」") + .isInstanceOf(IllegalStateException.class) + .hasMessageContaining("cwd="); + } }