779 lines
17 KiB
Markdown
779 lines
17 KiB
Markdown
下面是整合后的**最终完整方案**,包括:漫画解析、章节完整性、图片加载、调试能力、搜索性能,以及你刚刚提到的顶部搜索框 UI 问题。
|
||
Sakura 漫画模块完整改进方案
|
||
=================
|
||
|
||
一、改进目标
|
||
------
|
||
|
||
本轮改造解决 5 类问题:
|
||
1. 漫画章节目录不全
|
||
2. 漫画图片解析失败
|
||
3. 图片 URL 有数量但无法显示
|
||
4. 在线搜索速度慢
|
||
5. 漫画 / 图片页顶部搜索框布局异常,在线搜索入口位置不合理
|
||
|
||
* * *
|
||
|
||
二、整体优先级
|
||
=======
|
||
|
||
P0:顶部搜索框 UI 修复与在线搜索入口重构
|
||
P1:图片 URL 合法性验证
|
||
P2:PageImage + headers 拆分
|
||
P3:图片 HTTP 探测与调试能力
|
||
P4:TOC 分页抓全章节
|
||
P5:章节排序支持番外
|
||
P6:Script 图片变量提取
|
||
P7:漫画调试页
|
||
P8:搜索健康分与增量搜索
|
||
|
||
说明:
|
||
P0 先做,因为它是明显 UI 问题,改动小、收益快。
|
||
P1~P7 解决漫画“能不能看”。
|
||
P8 最后做,解决“搜得慢”。
|
||
|
||
* * *
|
||
|
||
P0:顶部搜索框 UI 修复与在线搜索入口重构
|
||
=======================
|
||
|
||
问题
|
||
--
|
||
|
||
截图中红框区域出现:
|
||
顶部搜索栏被状态栏 / 右上角按钮挤压
|
||
漫画页和图片页都出现同类问题
|
||
在线搜索入口放在右上角,不符合用户预期
|
||
目标
|
||
--
|
||
|
||
漫画页 / 图片页顶部布局稳定
|
||
搜索框不被状态栏遮挡
|
||
在线搜索入口集成到搜索框流程中
|
||
不再在右上角放在线搜索按钮
|
||
|
||
推荐交互
|
||
----
|
||
|
||
页面顶部只保留一个统一搜索框:
|
||
[ 搜索图标 ] 搜索标题、作者或标签 [筛选] [刷新]
|
||
|
||
点击搜索框进入统一搜索页:
|
||
SearchPage
|
||
├─ 搜索输入框
|
||
├─ 搜索范围切换:本地 / 在线 / 全部
|
||
└─ 结果分区:
|
||
├─ 本地漫画
|
||
├─ 在线漫画源
|
||
└─ 图片 / 插画
|
||
代码结构
|
||
----
|
||
|
||
新增统一组件:
|
||
class SakuraSearchBar extends StatelessWidget {
|
||
final String hintText;
|
||
final VoidCallback onTap;
|
||
final VoidCallback? onFilterTap;
|
||
final VoidCallback? onRefreshTap;
|
||
|
||
const SakuraSearchBar({
|
||
super.key,
|
||
required this.hintText,
|
||
required this.onTap,
|
||
this.onFilterTap,
|
||
this.onRefreshTap,
|
||
});
|
||
|
||
@override
|
||
Widget build(BuildContext context) {
|
||
return Material(
|
||
elevation: 6,
|
||
borderRadius: BorderRadius.circular(32),
|
||
color: Theme.of(context).colorScheme.surface,
|
||
child: InkWell(
|
||
borderRadius: BorderRadius.circular(32),
|
||
onTap: onTap,
|
||
child: SizedBox(
|
||
height: 56,
|
||
child: Row(
|
||
children: [
|
||
const SizedBox(width: 18),
|
||
const Icon(Icons.search),
|
||
const SizedBox(width: 12),
|
||
Expanded(
|
||
child: Text(
|
||
hintText,
|
||
maxLines: 1,
|
||
overflow: TextOverflow.ellipsis,
|
||
style: Theme.of(context).textTheme.titleMedium,
|
||
),
|
||
),
|
||
if (onFilterTap != null)
|
||
IconButton(
|
||
icon: const Icon(Icons.tune),
|
||
onPressed: onFilterTap,
|
||
),
|
||
if (onRefreshTap != null)
|
||
IconButton(
|
||
icon: const Icon(Icons.refresh),
|
||
onPressed: onRefreshTap,
|
||
),
|
||
const SizedBox(width: 8),
|
||
],
|
||
),
|
||
),
|
||
),
|
||
);
|
||
}
|
||
}
|
||
|
||
页面顶部必须用:
|
||
SafeArea(
|
||
bottom: false,
|
||
child: Padding(
|
||
padding: const EdgeInsets.fromLTRB(16, 8, 16, 8),
|
||
child: SakuraSearchBar(
|
||
hintText: '搜索标题、作者或标签',
|
||
onTap: () {
|
||
Navigator.push(
|
||
context,
|
||
MaterialPageRoute(
|
||
builder: (_) => UnifiedSearchPage(
|
||
initialMode: SearchMode.manga,
|
||
),
|
||
),
|
||
);
|
||
},
|
||
onFilterTap: _openFilter,
|
||
onRefreshTap: _refresh,
|
||
),
|
||
),
|
||
)
|
||
|
||
禁止再用:
|
||
Stack + Positioned 顶部硬定位
|
||
AppBar 右上角塞在线搜索入口
|
||
状态栏高度手动写死
|
||
|
||
* * *
|
||
|
||
P1:图片 URL 合法性验证
|
||
===============
|
||
|
||
目标
|
||
--
|
||
|
||
不要相信 `1/17` 这种页数。它可能是:
|
||
17 个真实图片
|
||
17 个空字符串
|
||
17 个 loading 图
|
||
17 个 placeholder
|
||
17 个错误 URL
|
||
|
||
新增模型:
|
||
class ImageCandidate {
|
||
final String raw;
|
||
final String? url;
|
||
final bool isValid;
|
||
final String? reason;
|
||
|
||
const ImageCandidate({
|
||
required this.raw,
|
||
required this.url,
|
||
required this.isValid,
|
||
this.reason,
|
||
});
|
||
}
|
||
|
||
验证逻辑:
|
||
ImageCandidate validateImageCandidate(String raw, String baseUrl) {
|
||
final value = raw.trim();
|
||
|
||
if (value.isEmpty) {
|
||
return ImageCandidate(
|
||
raw: raw,
|
||
url: null,
|
||
isValid: false,
|
||
reason: 'empty',
|
||
);
|
||
}
|
||
|
||
if (value.startsWith('data:')) {
|
||
return ImageCandidate(
|
||
raw: raw,
|
||
url: null,
|
||
isValid: false,
|
||
reason: 'data-uri',
|
||
);
|
||
}
|
||
|
||
final lower = value.toLowerCase();
|
||
if (lower.contains('loading') ||
|
||
lower.contains('placeholder') ||
|
||
lower.contains('logo') ||
|
||
lower.contains('icon')) {
|
||
return ImageCandidate(
|
||
raw: raw,
|
||
url: value,
|
||
isValid: false,
|
||
reason: 'placeholder',
|
||
);
|
||
}
|
||
|
||
final normalized = normalizeImageUrl(value, baseUrl);
|
||
final uri = Uri.tryParse(normalized);
|
||
|
||
if (uri == null || !uri.hasScheme || uri.host.isEmpty) {
|
||
return ImageCandidate(
|
||
raw: raw,
|
||
url: normalized,
|
||
isValid: false,
|
||
reason: 'invalid-url',
|
||
);
|
||
}
|
||
|
||
return ImageCandidate(
|
||
raw: raw,
|
||
url: normalized,
|
||
isValid: true,
|
||
);
|
||
}
|
||
|
||
URL 归一化:
|
||
String normalizeImageUrl(String raw, String baseUrl) {
|
||
var u = raw.trim()
|
||
.replaceAll(r'\/', '/')
|
||
.replaceAll(r'\u002F', '/');
|
||
|
||
if (u.startsWith('//')) {
|
||
u = 'https:$u';
|
||
}
|
||
|
||
return Uri.parse(baseUrl).resolve(u).toString();
|
||
}
|
||
|
||
* * *
|
||
|
||
P2:PageImage + headers 拆分
|
||
=========================
|
||
|
||
问题
|
||
--
|
||
|
||
Legado 漫画源经常返回:
|
||
https://xxx.jpg,{"headers":{"Referer":"https://site.com"}}
|
||
|
||
如果直接丢给:
|
||
Image.network(raw)
|
||
|
||
图片必炸。
|
||
新模型
|
||
---
|
||
|
||
class PageImage {
|
||
final int index;
|
||
final String raw;
|
||
final String url;
|
||
final Map<String, String> headers;
|
||
|
||
const PageImage({
|
||
required this.index,
|
||
required this.raw,
|
||
required this.url,
|
||
required this.headers,
|
||
});
|
||
}
|
||
|
||
拆分逻辑
|
||
----
|
||
|
||
class ParsedImageSpec {
|
||
final String url;
|
||
final Map<String, String> headers;
|
||
|
||
ParsedImageSpec({
|
||
required this.url,
|
||
required this.headers,
|
||
});
|
||
}
|
||
|
||
|
||
ParsedImageSpec splitLegadoImageSpec(String raw) {
|
||
final idx = raw.indexOf(',{');
|
||
if (idx < 0) {
|
||
return ParsedImageSpec(url: raw.trim(), headers: {});
|
||
}
|
||
|
||
final url = raw.substring(0, idx).trim();
|
||
final jsonText = raw.substring(idx + 1).trim();
|
||
|
||
final headers = <String, String>{};
|
||
|
||
try {
|
||
final decoded = jsonDecode(jsonText);
|
||
final h = decoded['headers'];
|
||
|
||
if (h is Map) {
|
||
h.forEach((k, v) {
|
||
headers[k.toString()] = v.toString();
|
||
});
|
||
} else if (h is String) {
|
||
final inner = jsonDecode(h);
|
||
if (inner is Map) {
|
||
inner.forEach((k, v) {
|
||
headers[k.toString()] = v.toString();
|
||
});
|
||
}
|
||
}
|
||
} catch (_) {}
|
||
|
||
return ParsedImageSpec(url: url, headers: headers);
|
||
}
|
||
|
||
生成 PageImage:
|
||
PageImage buildPageImage({
|
||
required String raw,
|
||
required String chapterUrl,
|
||
required LegadoSource source,
|
||
required int index,
|
||
}) {
|
||
final parsed = splitLegadoImageSpec(raw);
|
||
|
||
final headers = <String, String>{
|
||
...source.headers,
|
||
'User-Agent': source.userAgent ?? defaultUserAgent,
|
||
'Referer': chapterUrl,
|
||
...parsed.headers,
|
||
};
|
||
|
||
return PageImage(
|
||
index: index,
|
||
raw: raw,
|
||
url: normalizeImageUrl(parsed.url, chapterUrl),
|
||
headers: headers,
|
||
);
|
||
}
|
||
|
||
阅读器改成:
|
||
Image.network(
|
||
page.url,
|
||
headers: page.headers,
|
||
fit: BoxFit.contain,
|
||
errorBuilder: (_, error, stack) {
|
||
return const Icon(Icons.broken_image);
|
||
},
|
||
)
|
||
|
||
* * *
|
||
|
||
P3:图片 HTTP 探测
|
||
=============
|
||
|
||
目标
|
||
--
|
||
|
||
判断到底是:
|
||
URL 错
|
||
headers 错
|
||
反爬
|
||
图片损坏
|
||
Flutter 缓存问题
|
||
|
||
新增:
|
||
class ImageProbeResult {
|
||
final String url;
|
||
final int? statusCode;
|
||
final String? contentType;
|
||
final int contentLength;
|
||
final String first16BytesHex;
|
||
final bool looksLikeImage;
|
||
final String? error;
|
||
|
||
const ImageProbeResult({
|
||
required this.url,
|
||
this.statusCode,
|
||
this.contentType,
|
||
required this.contentLength,
|
||
required this.first16BytesHex,
|
||
required this.looksLikeImage,
|
||
this.error,
|
||
});
|
||
}
|
||
|
||
探测:
|
||
Future<ImageProbeResult> probeImage(Dio dio, PageImage image) async {
|
||
try {
|
||
final resp = await dio.get<List<int>>(
|
||
image.url,
|
||
options: Options(
|
||
responseType: ResponseType.bytes,
|
||
headers: image.headers,
|
||
validateStatus: (_) => true,
|
||
),
|
||
);
|
||
|
||
final bytes = resp.data ?? const <int>[];
|
||
final first16 = bytes
|
||
.take(16)
|
||
.map((b) => b.toRadixString(16).padLeft(2, '0'))
|
||
.join();
|
||
|
||
return ImageProbeResult(
|
||
url: image.url,
|
||
statusCode: resp.statusCode,
|
||
contentType: resp.headers.value('content-type'),
|
||
contentLength: bytes.length,
|
||
first16BytesHex: first16,
|
||
looksLikeImage: isImageMagic(bytes),
|
||
);
|
||
} catch (e) {
|
||
return ImageProbeResult(
|
||
url: image.url,
|
||
contentLength: 0,
|
||
first16BytesHex: '',
|
||
looksLikeImage: false,
|
||
error: e.toString(),
|
||
);
|
||
}
|
||
}
|
||
|
||
魔数判断:
|
||
bool isImageMagic(List<int> b) {
|
||
if (b.length < 4) return false;
|
||
|
||
// JPG
|
||
if (b[0] == 0xFF && b[1] == 0xD8 && b[2] == 0xFF) return true;
|
||
|
||
// PNG
|
||
if (b[0] == 0x89 && b[1] == 0x50 && b[2] == 0x4E && b[3] == 0x47) {
|
||
return true;
|
||
}
|
||
|
||
// WEBP
|
||
if (b.length >= 12 &&
|
||
b[0] == 0x52 &&
|
||
b[1] == 0x49 &&
|
||
b[2] == 0x46 &&
|
||
b[3] == 0x46 &&
|
||
b[8] == 0x57 &&
|
||
b[9] == 0x45 &&
|
||
b[10] == 0x42 &&
|
||
b[11] == 0x50) {
|
||
return true;
|
||
}
|
||
|
||
// GIF
|
||
if (b[0] == 0x47 && b[1] == 0x49 && b[2] == 0x46 && b[3] == 0x38) {
|
||
return true;
|
||
}
|
||
|
||
// BMP
|
||
if (b[0] == 0x42 && b[1] == 0x4D) return true;
|
||
|
||
return false;
|
||
}
|
||
|
||
诊断表:
|
||
403 / 444
|
||
→ Referer / User-Agent / Cookie 问题
|
||
|
||
404
|
||
→ 图片 URL 拼错或章节页已失效
|
||
|
||
200 + text/html
|
||
→ 反爬页 / 登录页 / 错误页
|
||
|
||
200 + image/* + contentLength < 1024
|
||
→ 可能是伪装错误图或破损图
|
||
|
||
200 + image/* + 魔数失败
|
||
→ 响应不是有效图片
|
||
|
||
200 + image/* + 魔数成功但 Image.network 不显示
|
||
→ headers 没传进 Image.network / 图片缓存污染 / Flutter 图片组件问题
|
||
|
||
* * *
|
||
|
||
P4:TOC 分页抓全章节
|
||
=============
|
||
|
||
问题
|
||
--
|
||
|
||
一人之下 600+ 话,但现在只显示几十章,说明目录分页没有抓全。
|
||
支持字段
|
||
----
|
||
|
||
tocNextPageRule
|
||
nextTocUrl
|
||
nextUrl
|
||
nextPage
|
||
tocNextPage
|
||
|
||
抓取流程
|
||
----
|
||
|
||
firstTocUrl
|
||
→ 请求当前页
|
||
→ 解析 chapterList
|
||
→ 解析 nextTocUrl
|
||
→ resolve 成绝对 URL
|
||
→ 合并章节并去重
|
||
→ 继续下一页
|
||
|
||
终止条件
|
||
----
|
||
|
||
nextUrl == null
|
||
seenPageUrls 已访问
|
||
连续 2 页新增章节为 0
|
||
pageCount >= 100
|
||
chapterCount >= 3000
|
||
|
||
建议参数:
|
||
const maxTocPages = 100;
|
||
const maxTocChapters = 3000;
|
||
const emptyPageDelay = Duration(milliseconds: 500);
|
||
|
||
* * *
|
||
|
||
P5:章节排序支持番外
|
||
===========
|
||
|
||
支持:
|
||
第660话
|
||
第660.5回
|
||
Ch.660
|
||
ch 660.5
|
||
660话
|
||
660.5 番外
|
||
|
||
章节号提取:
|
||
double? extractChapterNo(String title) {
|
||
var m = RegExp(r'第\s*(\d+(?:\.\d+)?)\s*[回话章节节]').firstMatch(title);
|
||
if (m != null) return double.tryParse(m.group(1)!);
|
||
|
||
m = RegExp(r'[Cc]h\.?\s*(\d+(?:\.\d+)?)').firstMatch(title);
|
||
if (m != null) return double.tryParse(m.group(1)!);
|
||
|
||
m = RegExp(r'^\s*(\d+(?:\.\d+)?)\s*(?:话|回|章|节)?').firstMatch(title);
|
||
if (m != null) return double.tryParse(m.group(1)!);
|
||
|
||
return null;
|
||
}
|
||
|
||
排序:
|
||
chapters.sort((a, b) {
|
||
final na = extractChapterNo(a.title);
|
||
final nb = extractChapterNo(b.title);
|
||
|
||
if (na != null && nb != null && na != nb) {
|
||
return na.compareTo(nb);
|
||
}
|
||
|
||
return a.index.compareTo(b.index);
|
||
});
|
||
|
||
* * *
|
||
|
||
P6:Script 图片变量提取
|
||
================
|
||
|
||
当 `ruleContent.content` 失败或候选图片全无效时,从 rawHtml 的 script 中提取图片。
|
||
|
||
支持变量:
|
||
newImgs
|
||
imageList
|
||
page_url
|
||
chapterImages
|
||
photosr
|
||
img_data
|
||
picTree
|
||
|
||
示例:
|
||
List<String> extractImagesFromScripts(String html) {
|
||
final out = <String>[];
|
||
|
||
final patterns = [
|
||
RegExp(r'newImgs\s*=\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'imageList\s*=\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'"page_url"\s*:\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'chapterImages\s*=\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'photosr\s*=\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'img_data\s*=\s*(\[[\s\S]*?\])'),
|
||
RegExp(r'picTree\s*=\s*(\[[\s\S]*?\])'),
|
||
];
|
||
|
||
for (final p in patterns) {
|
||
final m = p.firstMatch(html);
|
||
if (m == null) continue;
|
||
|
||
final raw = m.group(1)!;
|
||
final list = tryDecodeStringList(raw);
|
||
out.addAll(list.map(sanitizeScriptUrl));
|
||
}
|
||
|
||
return out;
|
||
}
|
||
|
||
清洗:
|
||
String sanitizeScriptUrl(String url) {
|
||
return url
|
||
.replaceAll(r'\/', '/')
|
||
.replaceAll(r'\u002F', '/');
|
||
}
|
||
|
||
* * *
|
||
|
||
P7:漫画调试页
|
||
========
|
||
|
||
入口
|
||
--
|
||
|
||
书籍详情页 → 长按章节 → 调试本章
|
||
阅读器右上角菜单 → 调试当前章
|
||
|
||
展示内容
|
||
----
|
||
|
||
{
|
||
"sourceName": "包子漫画5",
|
||
"bookTitle": "一人之下",
|
||
"chapterTitle": "第652回",
|
||
"chapterUrl": "...",
|
||
"ruleContent": "...",
|
||
"rawHtmlLength": 123456,
|
||
"contentOutputLength": 2048,
|
||
"rawCandidateCount": 17,
|
||
"validCandidateCount": 0,
|
||
"pageImageCount": 0,
|
||
"firstCandidates": [
|
||
{
|
||
"raw": "...",
|
||
"normalized": "...",
|
||
"reason": "placeholder"
|
||
}
|
||
],
|
||
"probe": [
|
||
{
|
||
"url": "...",
|
||
"status": 200,
|
||
"contentType": "image/jpeg",
|
||
"contentLength": 312,
|
||
"first16Bytes": "ffd8ff...",
|
||
"looksLikeImage": false
|
||
}
|
||
]
|
||
}
|
||
|
||
操作按钮
|
||
----
|
||
|
||
复制调试信息
|
||
重新检测前 5 张图
|
||
切换 Referer:章节页 / 详情页 / 源首页
|
||
清除图片缓存后重试
|
||
打开原始章节 URL
|
||
|
||
* * *
|
||
|
||
P8:搜索健康分与增量搜索
|
||
=============
|
||
|
||
问题
|
||
--
|
||
|
||
搜索慢主要来自:
|
||
坏源太多
|
||
每个源 timeout 太长
|
||
所有源一起等结果
|
||
模型
|
||
--
|
||
|
||
class SourceHealth {
|
||
final String sourceId;
|
||
|
||
int successCount = 0;
|
||
int failCount = 0;
|
||
int timeoutCount = 0;
|
||
int totalLatencyMs = 0;
|
||
int consecutiveFails = 0;
|
||
|
||
DateTime? lastSuccessAt;
|
||
DateTime? lastFailAt;
|
||
}
|
||
|
||
策略
|
||
--
|
||
|
||
健康源优先
|
||
坏源降权
|
||
连续失败 >= 5 临时跳过
|
||
搜索超时 5~8 秒
|
||
快速源先返回
|
||
慢源后台补充
|
||
|
||
搜索阶段
|
||
----
|
||
|
||
0~5s:快速源结果
|
||
5~8s:普通源结果
|
||
8s 后:慢源后台补
|
||
|
||
* * *
|
||
|
||
三、最终落地顺序
|
||
========
|
||
|
||
第 1 阶段:UI 先止血
|
||
-------------
|
||
|
||
P0:顶部搜索框 UI 修复
|
||
P0:在线搜索入口放进统一搜索页
|
||
|
||
结果:
|
||
漫画页 / 图片页顶部不再错位
|
||
在线搜索不再占右上角
|
||
搜索体验统一
|
||
第 2 阶段:图片链路可观测
|
||
--------------
|
||
|
||
P1:图片 URL 合法性验证
|
||
P2:PageImage + headers 拆分
|
||
P3:图片 HTTP 探测
|
||
P7:漫画调试页
|
||
|
||
结果:
|
||
可以准确判断图片失败原因
|
||
不再靠猜
|
||
第 3 阶段:章节抓全
|
||
-----------
|
||
|
||
P4:TOC 分页
|
||
P5:章节排序
|
||
|
||
结果:
|
||
一人之下这类 600+ 话漫画可以抓全
|
||
章节顺序稳定
|
||
第 4 阶段:提升漫画源覆盖率
|
||
---------------
|
||
|
||
P6:Script 图片变量提取
|
||
|
||
结果:
|
||
更多漫画源能解析图片
|
||
第 5 阶段:优化搜索速度
|
||
-------------
|
||
|
||
P8:搜索健康分与增量搜索
|
||
|
||
结果:
|
||
坏源不拖慢搜索
|
||
结果更快出现
|
||
|
||
* * *
|
||
|
||
|
||
=======
|