297 lines
13 KiB
Java
297 lines
13 KiB
Java
package com.nanri.aiimage.testutil;
|
||
|
||
import java.io.IOException;
|
||
import java.nio.charset.StandardCharsets;
|
||
import java.nio.file.Files;
|
||
import java.nio.file.Path;
|
||
import java.util.List;
|
||
import java.util.regex.Matcher;
|
||
import java.util.regex.Pattern;
|
||
import java.util.stream.Collectors;
|
||
import java.util.stream.Stream;
|
||
|
||
/** 源码扫描式护栏的取文助手。<b>路径解析照抄 {@code common/security/InternalTokenSingleOutletTest} 的
|
||
* moduleRoot()/mainSources() 两支</b>——surefire 的 cwd 是模块目录,从仓库根或 IDE 跑时要退到 backend-java/,
|
||
* 定位不到必须 fail 而不是静默跳过(静默跳过 = 整支护栏恒绿)。 */
|
||
public final class SourceScan {
|
||
|
||
private SourceScan() {
|
||
}
|
||
|
||
/** 一处 {@code sourceText("<裸名>")} 字面量点名。 */
|
||
public record CallSite(String bareName, String relativeFile, int line) {
|
||
}
|
||
|
||
private static final Pattern CALL_SITE = Pattern.compile("sourceText\\(\\s*\"([^\"]+)\"\\s*\\)");
|
||
|
||
/** {@code src/test/java} 的绝对根,与 {@link #mainSources()} 同一条定位分支。 */
|
||
public static Path testSourcesRoot() {
|
||
Path root = moduleRoot().resolve("src/test/java");
|
||
if (!Files.isDirectory(root)) {
|
||
throw new IllegalStateException("语料根不存在: " + root + "(cwd=" + Path.of("").toAbsolutePath() + ")");
|
||
}
|
||
return root;
|
||
}
|
||
|
||
/** 点名普查:列出 {@code src/test/java} 里每一处 {@code sourceText("<裸名>")} 字面量调用。 */
|
||
public static List<CallSite> censusSourceTextCallSites() {
|
||
return censusSourceTextCallSites(testSourcesRoot());
|
||
}
|
||
|
||
/**
|
||
* 点名普查(可指定语料根)。三条设计决定都是被坑逼出来的:
|
||
* <ol>
|
||
* <li><b>数的是代码不是文字</b>:判据打在 {@link #withoutComments} 上,注释里提一句
|
||
* {@code sourceText("Fake.java")} 不计入——否则台账的 15 会被 javadoc 喂成任意大的数。</li>
|
||
* <li><b>只配字面量</b>:正则要求紧跟一对引号,所以 {@code SourceScan.java:22} 那行
|
||
* 被调方自己的声明(参数是 {@code String fileName})天然不算调用点。plan 原文的「9 支文件」
|
||
* 正是 {@code grep -rl 'sourceText('} 把声明也算了进去;真实调用点是 15 处/8 支文件。</li>
|
||
* <li><b>语料根不存在必须抛</b>,不许返回空表:空表会被读成「已经没人用裸名口径了」。
|
||
* 目录存在且确实没有 {@code .java} 才返回空——那是"语料真的没了",与"尺子架错地方"是两回事。</li>
|
||
* </ol>
|
||
*/
|
||
public static List<CallSite> censusSourceTextCallSites(Path corpusRoot) {
|
||
Path root = corpusRoot.toAbsolutePath();
|
||
if (!Files.isDirectory(root)) {
|
||
throw new IllegalStateException("语料根不存在: " + root + "(cwd=" + Path.of("").toAbsolutePath() + ")");
|
||
}
|
||
Path base = moduleRoot();
|
||
List<CallSite> sites = new java.util.ArrayList<>();
|
||
try (Stream<Path> walk = Files.walk(root)) {
|
||
List<Path> 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())));
|
||
}
|
||
}
|
||
} catch (IOException e) {
|
||
throw new IllegalStateException("遍历 " + root + " 失败: " + e.getMessage(), e);
|
||
}
|
||
return List.copyOf(sites);
|
||
}
|
||
|
||
private static String relativizeForCensus(Path file, Path corpusRoot, Path moduleRoot) {
|
||
Path absolute = file.toAbsolutePath();
|
||
if (absolute.startsWith(moduleRoot)) {
|
||
return moduleRoot.relativize(absolute).toString();
|
||
}
|
||
return corpusRoot.relativize(absolute).toString();
|
||
}
|
||
|
||
/** 偏移不变的 {@link #withoutComments} 保证:剥完再数换行,行号仍等于原文件行号。 */
|
||
private static int lineAt(String code, int offset) {
|
||
int line = 1;
|
||
for (int i = 0; i < offset; i++) {
|
||
if (code.charAt(i) == '\n') {
|
||
line++;
|
||
}
|
||
}
|
||
return line;
|
||
}
|
||
|
||
/** 按文件名(不含目录)取 {@code src/main/java} 下那份源码的<b>全文字符串</b>。
|
||
* ⚠️ 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。 */
|
||
public static String sourceText(String fileName) {
|
||
return mainSources().stream()
|
||
.filter(p -> p.getFileName().toString().equals(fileName))
|
||
.findFirst()
|
||
.map(SourceScan::read)
|
||
.orElseThrow(() -> new IllegalStateException(
|
||
"src/main/java 下找不到 " + fileName + " —— 文件改名/挪目录了,护栏要红在这里"));
|
||
}
|
||
|
||
/**
|
||
* 剥掉注释,<b>偏移不变</b>:注释字符换成空格、换行原样保留,所以 {@link #methodBody} 这类
|
||
* 按位置切片的助手在剥前剥后切出同一段。
|
||
*
|
||
* <p>为什么护栏必须打在它上面:源码守卫的两种断言都会被<b>文字</b>喂到——
|
||
* 「必须不存在」那一形会被注释里提一句顶红(假红,还能忍),
|
||
* 「必须存在/次数等于 N」那一形会被 javadoc 里提一句<b>喂饱</b>(假绿,等于护栏没写)。
|
||
* 2026-10-07 实测:{@code AdminTaskCenterController} 的类注释里有 1 处 {@code readScope(},
|
||
* 整份原文 {@code readScope(} 数=1、剥完=0——文件级判据吃原文就会读出这种数。</p>
|
||
*
|
||
* <p>口径:{@code //} 到行尾、{@code /* … *}{@code /} 整段、字符串与字符字面量里的斜杠不算注释、
|
||
* 三个双引号的文本块按字面量整段保留。</p>
|
||
*/
|
||
public static String withoutComments(String java) {
|
||
StringBuilder out = new StringBuilder(java.length());
|
||
int i = 0;
|
||
int n = java.length();
|
||
while (i < n) {
|
||
char c = java.charAt(i);
|
||
char next = i + 1 < n ? java.charAt(i + 1) : '\0';
|
||
if (c == '/' && next == '/') {
|
||
while (i < n && java.charAt(i) != '\n') {
|
||
out.append(' ');
|
||
i++;
|
||
}
|
||
} else if (c == '/' && next == '*') {
|
||
out.append(" ");
|
||
i += 2;
|
||
while (i < n) {
|
||
if (java.charAt(i) == '*' && i + 1 < n && java.charAt(i + 1) == '/') {
|
||
out.append(" ");
|
||
i += 2;
|
||
break;
|
||
}
|
||
out.append(java.charAt(i) == '\n' ? '\n' : ' ');
|
||
i++;
|
||
}
|
||
} else if (java.startsWith("\"\"\"", i)) {
|
||
out.append("\"\"\"");
|
||
i += 3;
|
||
while (i < n && !java.startsWith("\"\"\"", i)) {
|
||
out.append(java.charAt(i));
|
||
i++;
|
||
}
|
||
out.append("\"\"\"");
|
||
i += 3;
|
||
} else if (c == '"' || c == '\'') {
|
||
out.append(c);
|
||
i++;
|
||
while (i < n) {
|
||
if (java.charAt(i) == '\\' && i + 1 < n) {
|
||
out.append(java.charAt(i)).append(java.charAt(i + 1));
|
||
i += 2;
|
||
continue;
|
||
}
|
||
if (java.charAt(i) == '\n') {
|
||
break;
|
||
}
|
||
out.append(java.charAt(i));
|
||
boolean closing = java.charAt(i) == c;
|
||
i++;
|
||
if (closing) {
|
||
break;
|
||
}
|
||
}
|
||
} else {
|
||
out.append(c);
|
||
i++;
|
||
}
|
||
}
|
||
return out.toString();
|
||
}
|
||
|
||
/** 大括号配平切出「某个方法」的源码:<b>从它的签名行往前一直含到注解</b>,到方法收尾的 `}` 为止。
|
||
*
|
||
* <p>两条它必须自己防住的坑,都写在用例里过:</p>
|
||
* <ul>
|
||
* <li><b>名字必须是完整词</b>:{@code delete} 是 {@code deleteGroup} 的前缀。用 {@code indexOf(name)}
|
||
* 会先撞上那个更长的方法 ⇒ "切出来的是隔壁",而断言仍然绿。所以下面用 {@code (?<![A-Za-z0-9_])} 卡左边界,
|
||
* 并且要求紧跟 {@code (}(构造器/局部变量名同前缀也不算方法)。</li>
|
||
* <li><b>找不到必须抛</b>,不许返回空串:返回 "" 的话 {@code doesNotContain("company_id")} 恒真,
|
||
* 反向钉会因为方法被改名而悄悄失效(memory「字符串断言必须排他、解析断言必须可匹配」)。
|
||
* 配套要求:每个用它的反向钉都要带一条"该做的确实做了"的取样自检。</li>
|
||
* </ul>
|
||
*/
|
||
public static String methodBody(String source, String methodName) {
|
||
var matcher = java.util.regex.Pattern
|
||
.compile("(?<![A-Za-z0-9_])" + java.util.regex.Pattern.quote(methodName) + "\\s*\\(")
|
||
.matcher(source);
|
||
while (matcher.find()) {
|
||
int open = matcher.end() - 1;
|
||
int signatureStart = findSignatureStart(source, matcher.start());
|
||
int bodyEnd = matchBraces(source, nextBrace(source, open));
|
||
if (bodyEnd < 0) {
|
||
continue;
|
||
}
|
||
return source.substring(signatureStart, bodyEnd + 1);
|
||
}
|
||
throw new AssertionError("源码里没有方法「" + methodName + "」——它被改名或删掉了。"
|
||
+ "反向钉宁可红在这里,也不许因为切不到方法而恒绿。");
|
||
}
|
||
|
||
/** 从签名的左括号往前退到上一条 `{`/`}`/`;` 之后,再退回到第一行注解的起始(注解属于这个方法,必须含进来)。 */
|
||
private static int findSignatureStart(String source, int nameStart) {
|
||
int boundary = 0;
|
||
for (int i = nameStart; i-- > 0; ) {
|
||
char c = source.charAt(i);
|
||
if (c == '{' || c == '}' || c == ';') {
|
||
boundary = i + 1;
|
||
break;
|
||
}
|
||
}
|
||
String head = source.substring(boundary, nameStart);
|
||
int firstMeaningful = head.length();
|
||
for (String raw : head.split("\n", -1)) {
|
||
if (!raw.isBlank()) {
|
||
firstMeaningful = head.indexOf(raw.stripLeading());
|
||
break;
|
||
}
|
||
}
|
||
return boundary + firstMeaningful;
|
||
}
|
||
|
||
private static int nextBrace(String source, int from) {
|
||
for (int i = from; i < source.length(); i++) {
|
||
if (source.charAt(i) == '{') {
|
||
return i;
|
||
}
|
||
if (source.charAt(i) == ';') {
|
||
return -1; // 抽象方法/接口声明,不是我们要切的东西
|
||
}
|
||
}
|
||
return -1;
|
||
}
|
||
|
||
/** 朴素配平:源码里的 `{` 与 `}` 若出现在字符串字面量里会算错。本仓库 controller 没有这种写法,
|
||
* 但这条限制必须写在脸上——将来有人往方法体里塞一段带花括号的字符串,这个助手会悄悄切短。 */
|
||
private static int matchBraces(String source, int open) {
|
||
if (open < 0) {
|
||
return -1;
|
||
}
|
||
int depth = 0;
|
||
for (int i = open; i < source.length(); i++) {
|
||
char c = source.charAt(i);
|
||
if (c == '{') {
|
||
depth++;
|
||
} else if (c == '}') {
|
||
if (--depth == 0) {
|
||
return i;
|
||
}
|
||
}
|
||
}
|
||
return -1;
|
||
}
|
||
|
||
private static List<Path> mainSources() {
|
||
Path root = moduleRoot().resolve("src/main/java");
|
||
try (Stream<Path> walk = Files.walk(root)) {
|
||
return walk.filter(Files::isRegularFile)
|
||
.filter(p -> p.getFileName().toString().endsWith(".java"))
|
||
.collect(Collectors.toList());
|
||
} catch (IOException e) {
|
||
// 定位不到必须抛,不许静默返回空表:返回空 ⇒ sourceText 那支护栏扫不到任何文件而恒绿。
|
||
throw new IllegalStateException("遍历 " + root + " 失败: " + e.getMessage(), e);
|
||
}
|
||
}
|
||
|
||
private static Path moduleRoot() {
|
||
Path here = Path.of("").toAbsolutePath();
|
||
if (Files.isDirectory(here.resolve("src/main/java"))) {
|
||
return here;
|
||
}
|
||
Path nested = here.resolve("backend-java");
|
||
if (Files.isDirectory(nested.resolve("src/main/java"))) {
|
||
return nested;
|
||
}
|
||
throw new IllegalStateException("无法定位 backend-java 模块目录(cwd=" + here + ")");
|
||
}
|
||
|
||
private static String read(Path file) {
|
||
try {
|
||
return Files.readString(file, StandardCharsets.UTF_8);
|
||
} catch (IOException e) {
|
||
throw new IllegalStateException("读取失败: " + file, e);
|
||
}
|
||
}
|
||
}
|