/// <summary>
/// 取图片分页
/// </summary>
public static IEnumerable <PageModel> GetImagePage(SexSpider sex, string url)
{
string html = sex.ImgType != null && sex.ImgType.Contains("ajax") ? GetJSContent(url, sex.PageEncode) : GetHtmlContent(url, sex.PageEncode, sex.Domain);
//过滤站点
html = FilterHtml(html, sex.SiteFilter);
html = ReplaceHtml(html, sex.SiteReplace);
FilterChain chain = LoadFilter(sex.PageFilter);
html = chain.DoFilter(html);
//分页的时候取当前页
string _domain = url.Substring(0, url.LastIndexOf('/') + 1);
var parser = new HtmlParser();
var _document = parser.Parse(html);
var content = _document.QuerySelectorAll(sex.PageDiv);
foreach (var item in content)
{
//分页时点击类型没有内容要注释，此处与获取多分页时冲突
if (!Regex.IsMatch(item.InnerHtml, @"^\d*$"))
{
continue;
}
string _link = item.GetAttribute("href");
if (_link == null || _link == "#" || _link.Contains("javascript"))
{
continue;
}
_link = GetLink(_link, _domain, sex.Domain);
yield return(new PageModel
{
PageUrl = _link
});
}
}
/// <summary>
/// 取图片总页数
/// </summary>
public static string GetPageTotal(SexSpider sex, string url)
{
string total = "";
string html = sex.ImgType != null && sex.ImgType.Contains("ajax") ? GetJSContent(url, sex.PageEncode) : GetHtmlContent(url, sex.PageEncode, sex.Domain);
//过滤站点
html = FilterHtml(html, sex.SiteFilter);
html = ReplaceHtml(html, sex.SiteReplace);
var parser = new HtmlParser();
var _document = parser.Parse(html);
var content = _document.QuerySelectorAll(sex.PageFilter);//取总页数
foreach (var item in content)
{
string str = System.Net.WebUtility.HtmlDecode(item.InnerHtml);
total = Regex.Replace(str, "[^\\d]", "");
}
return(total);
}
/// <summary>
/// 取有分页的图片
/// </summary>
public static IEnumerable <ImageModel> GetListImagePage(SexSpider sex, string url)
{
var images = new List <ImageModel>();
var pages = new List <PageModel>();
var newPages = SiteHelper.GetImagePage(sex, url).ToList();
//1：默认 2：总页数[1][2]..[10] 3：先通过filter取总页数 4:ajax #page
if (sex.PageLevel == 4)
{
string total = GetPageTotal(sex, url);
pages = GetPageAjax(url, total);
}
else if (sex.PageLevel == 3)
{
string total = GetPageTotal(sex, url);
pages = GetPageMany(url, newPages, total);
}
else if (sex.PageLevel == 2)
{
pages = GetPageMany(url, newPages, "");
}
else
{
pages = newPages;
}
//添加原始页面
pages.Insert(0, new PageModel {
PageUrl = url
});
foreach (var p in pages)
{
var image = SiteHelper.GetListImage(sex, p.PageUrl).ToList();
images.AddRange(image);
}
return(images);
}
/// <summary>
/// 取图片页面
/// </summary>
public static IEnumerable <ImageModel> GetListImage(SexSpider sex, string url)
{
string html = sex.ImgType != null && sex.ImgType.Contains("ajax") ? GetJSContent(url, sex.PageEncode) : GetHtmlContent(url, sex.PageEncode, sex.Domain);
//过滤站点
html = FilterHtml(html, sex.SiteFilter);
html = ReplaceHtml(html, sex.SiteReplace);
FilterChain chain = LoadFilter(sex.ImageFilter);
var parser = new HtmlParser();
var _document = parser.Parse(html);
var content = _document.QuerySelectorAll(sex.ImageDiv);
foreach (var item in content)
{
string link = "";
if (chain.Count() > 0)
{
link = chain.DoFilter(item.OuterHtml);
}
else
{
link = item.GetAttribute("src");
}
if (String.IsNullOrEmpty(link))
{
continue;
}
string _image = GetLink(link, sex.Domain);
yield return(new ImageModel
{
ImageUrl = _image,
ImageDomain = sex.Domain
});
}
}
/// <summary>
/// 取站点列表
/// </summary>
public static IEnumerable <ListModel> GetSiteList(SexSpider sex)
{
string html = sex.DocType != null && sex.DocType.Contains("ajax") ? GetJSContent(sex.SiteLink, sex.PageEncode) : GetHtmlContent(sex.SiteLink, sex.PageEncode, sex.Domain);
//过滤站点
html = FilterHtml(html, sex.SiteFilter);
html = ReplaceHtml(html, sex.SiteReplace);
FilterChain chain = LoadFilter(sex.ListFilter);
if (sex.DocType != null && sex.DocType.Contains("json"))
{
string[] root = Regex.Split(sex.ListDiv, "\\|\\|");
var jObject = Newtonsoft.Json.Linq.JObject.Parse(html);
var jToken = jObject[root[0]];
string[] m = root[0].Split('&');
switch (m.Length)
{
case 2:
jToken = jObject[m[0]][m[1]];
break;
case 3:
jToken = jObject[m[0]][m[1]][m[2]];
break;
}
foreach (var item in jToken)
{
string[] child = root[1].Split('&');
yield return(new ListModel
{
Thumb = item.Value <string>(child[2]),
Title = System.Net.WebUtility.HtmlDecode(item.Value <string>(child[0])),
Link = GetLink(item.Value <string>(child[1]), sex.Domain),
Domain = sex.Domain,
LastStart = item.Value <string>(child[3])
});
}
}
else
{
var parser = new HtmlParser();
var _document = parser.Parse(html);
if (!string.IsNullOrWhiteSpace(sex.MainDiv))
{
var main = _document.QuerySelectorAll(sex.MainDiv);
foreach (var m in main)
{
var ctx = parser.Parse(m.InnerHtml);
var item = ctx.QuerySelector(sex.ListDiv);
if (item == null)
{
continue;
}
string _title = chain.DoFilter(item.InnerHtml);
string _link = GetLink(item.GetAttribute("href"), sex.Domain);
if (String.IsNullOrEmpty(_title))
{
continue;
}
var thumb = ctx.QuerySelector(sex.ThumbDiv);
var imgtext = thumb == null ? "" : thumb.OuterHtml;
yield return(new ListModel
{
Thumb = GetThumb(imgtext, sex.Domain),
Title = System.Net.WebUtility.HtmlDecode(_title),
Link = _link,
Domain = sex.Domain
});
}
}
else
{
var content = _document.QuerySelectorAll(sex.ListDiv);
foreach (var item in content)
{
string _title = chain.DoFilter(item.InnerHtml);
string _link = GetLink(item.GetAttribute("href"), sex.Domain);
if (String.IsNullOrEmpty(_title))
{
continue;
}
yield return(new ListModel
{
Thumb = GetThumb(item.InnerHtml, sex.Domain),
Title = System.Net.WebUtility.HtmlDecode(_title),
Link = _link,
Domain = sex.Domain
});
}
}
}
}