一位老手写的爬虫程序。
今天接到一个活，需要统计人员的工号信息，由于种种原因不能直接连数据库 [无奈]、[无奈]、[无奈]。采取迂回方案，写个工具自动登录网站，采集用户信息。
这也不是第一次采集ASP.NET网站，以前采集的时候就知道，这种网站采集比较麻烦，尤其是WebForm的ASP.NET 网站，那叫一个费劲。
喜欢现在流行的Restful模式的网站，数据接口采集那才叫舒服。
闲话少说，开干
工作量不大，HTTP纯手写
先准备下一个GET/POST预备使用
public static string Get(string url, Action<string> SuccessCallback, Action<string> FailCallback) {
HttpWebRequest req = WebRequest.Create(url) as HttpWebRequest;
req.Method = "GET";
req.UserAgent = "Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/65.0.3325.181 Safari/537.36";
req.Accept = "*/*";
req.KeepAlive = true;
req.ServicePoint.ConnectionLimit = int.MaxValue;
req.ServicePoint.Expect100Continue = false;
req.CookieContainer = sznyCookie; #静态变量
req.Credentials = System.Net.CredentialCache.DefaultCredentials;
string msg = "";
using (HttpWebResponse rsp = req.GetResponse() as HttpWebResponse)
{
using (StreamReader reader = new StreamReader(rsp.GetResponseStream()))
{
msg = reader.ReadToEnd();
}
}
return msg;
}
public static string Post(string url, Dictionary<string, string> dicParms, Action<string> SuccessCallback, Action<string> FailCallback) {
StringBuilder data = new StringBuilder();
foreach (var kv in dicParms) {
if (kv.Key.StartsWith("header"))
continue;
data.Append($"&{Common.UrlEncode( kv.Key,Encoding.UTF8)}={ Common.UrlEncode( kv.Value,Encoding.UTF8)}");
}
if (data.Length > 0)
data.Remove(0, 1);
HttpWebRequest req = WebRequest.Create(url) as HttpWebRequest;
req.Method = "POST";
req.KeepAlive = true;
req.CookieContainer = sznyCookie;
req.Connection = "KeepAlive";
req.KeepAlive = true;
req.ContentType = "application/x-www-form-urlencoded";
req.Accept = "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9";
req.Referer = url;
if (dicParms.ContainsKey("ScriptManager1"))
{
req.Headers.Add("X-MicrosoftAjax", "Delta=true");
req.Headers.Add("X-Requested-With", "XMLHttpRequest");
req.ContentType = "application/x-www-form-urlencoded; charset=UTF-8";
req.Accept = "*/*";
}
req.Headers.Add("Cache-Control", "no-cache");
req.UserAgent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/80.0.3987.163 Safari/537.36";
req.ServicePoint.ConnectionLimit = int.MaxValue;
req.ServicePoint.Expect100Continue = false;
req.AllowAutoRedirect = true;
req.Credentials = System.Net.CredentialCache.DefaultCredentials;
byte[] buffer = Encoding.UTF8.GetBytes(data.ToString());
using (Stream reqStream = req.GetRequestStream())
{
reqStream.Write(buffer, 0, buffer.Length);
}
string msg = "";
using (HttpWebResponse rsp = req.GetResponse() as HttpWebResponse)
{
using (StreamReader reader = new StreamReader(rsp.GetResponseStream()))
{
msg = reader.ReadToEnd();
if (msg.Contains("images/dl.jpg") || msg.Contains("pageRedirect||%2flogin.aspx"))
{
//登录失败
if (FailCallback != null)
FailCallback(msg);
}
else {
if (SuccessCallback!=null)
SuccessCallback(msg);
}
}
}
return msg;
}
整个过程分为登陆、用户信息列表、用户信息详情，分三步走来完成这个项目
登陆
根据Chrome抓包结果编写Login，帐号密码没有任何加密，直接明文显示了，直接用了，根据是否跳转页面判断是否登陆成功。调试查看结果登陆成功了。
根据上面的抓包数据，可以调用下面的代码确定是否登陆成功。
public static bool SznyLogin(string username, string password, Action<string> SuccessCallback, Action<string> FailCallback) {
string url = "http://127.0.0.1/login.aspx";
string msg = Get(url, SuccessCallback, FailCallback);
if (msg.Trim().Length > 0) {
Dictionary<string, string> dicParms = new Dictionary<string, string>();
dicParms.Add("__VIEWSTATE", "");
dicParms.Add("__EVENTVALIDATION", "");
dicParms.Add("Text_Name", "");
dicParms.Add("Text_Pass", "");
dicParms.Add("btn_Login.x", new Random().Next(100).ToString());
dicParms.Add("btn_Login.y", new Random().Next(200).ToString());
MatchCollection mc = Regex.Matches(msg, @"<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?value=""(?<val>[^""]*?)""[^<>]*?/?>|<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>|<select[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
if (dicParms.ContainsKey(mi.Groups["name"].Value.Trim()))
dicParms[mi.Groups["name"].Value.Trim()] = mi.Groups["val"].Value.Trim();
}
dicParms["Text_Name"] = username;
dicParms["Text_Pass"] = password;
msg=Post(url, dicParms, SuccessCallback, FailCallback);
if (msg.Contains("images/dl.jpg") || msg.Contains("pageRedirect||%2flogin.aspx"))
{
return false;
}
else
return true;
}
return false;
}
抓取人员信息
看到下面这个页面，失望了，列表上没有工号，如果列表上有工号 设置一页显示全部信息就可以把所有的数据都抓取到了。
换个思路：是不是我直接设置一页显示所有的数据后，然后根据员工ID可以获取到所有的信息呢？
接下来点击任意一条信息后，查看详情，显示下面的调用结果。Url上没有ID，Get这条路走不通了，查看Post的数据，更失望，没有ID,通过行信息绑定。传统的WebForm 提交模式…
把所有的数据显示到一页，把列表的数据先采集完，然后最后一个页面一个页面的采集工号信息。
public static CookieContainer sznyCookie = new CookieContainer();
/// <summary>
/// 员工信息
/// </summary>
public static Dictionary<int, Dictionary<string,string>> dicSznyEmployees = new Dictionary<int, Dictionary<string, string>>();
public static Dictionary<string, string> dicSznyEmployeeParms = new Dictionary<string, string>();
/// <summary>
/// 人员顺序号
/// </summary>
public static ConcurrentQueue<int> queueSznyEmployeeInfo = new ConcurrentQueue<int>();
public static ConcurrentQueue<int> queueSuccessEmployeeInfo = new ConcurrentQueue<int>();
public static bool SznyEmployeeList(Action<string> SuccessCallback, Action<string> FailCallback)
{
string url = $"http://127.0.0.1/HumanResources/EmployeeManage/EmployeeInfoList.aspx";
string msg = Get(url, SuccessCallback, FailCallback);
if (msg.Trim().Length > 100)
{
//统计参数
//__doPostBack\('(?<name>[^']*?)'
//new Regex(@"__doPostBack\('(?<name>[^']*?)'", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
string name = "";
MatchCollection mc = Regex.Matches(msg, @"__doPostBack\('(?<name>[^']*?)'", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
name = mi.Groups["name"].Value.Trim();
break;
}
//(?<=<a[^<>]*?href="javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)"[^<>]*?>条/页)
//new Regex(@"(?<=<a[^<>]*?href=""javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)""[^<>]*?>条/页)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
string smname = "";
Match m = Regex.Match(msg, @"(?<=<a[^<>]*?href=""javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)""[^<>]*?>条/页)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
if (m.Success)
smname = m.Value.Trim().Replace("'", "").Replace("'", "");
//<input[^<>]*?name="(?<name>[^"]*?)"[^<>]*?value="(?<val>[^"]*?)"[^<>]*?/?>|<input[^<>]*?name="(?<name>[^"]*?)"[^<>]*?"[^<>]*?/?>|<select[^<>]*?name="(?<name>[^"]*?)"[^<>]*?"[^<>]*?/?>
//new Regex(@"<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?value=""(?<val>[^""]*?)""[^<>]*?/?>|<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>|<select[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
Dictionary<string, string> dicParms = new Dictionary<string, string>();
dicParms.Add("ScriptManager1", $"UpdatePanel1|{smname}");
dicParms.Add("__EVENTTARGET", smname);
dicParms.Add("__EVENTARGUMENT", "");
dicParms.Add("__VIEWSTATE", "");
dicParms.Add("__EVENTVALIDATION", "");
dicParms.Add("MdGridView_t_unitemployees_dwyg_GridViewID", $"{name}");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentPage", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalPage", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageSize", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentNum", "0");
dicParms.Add("XM", "ZXMCHECK");
List<string> lstParms = new List<string>() { "XM", "MdGridView_t_unitemployees_dwyg_iCurrentPage", "MdGridView_t_unitemployees_dwyg_GridViewID", "MdGridView_t_unitemployees_dwyg_iCurrentNum", "MdGridView_t_unitemployees_dwyg_iPageCount", "MdGridView_t_unitemployees_dwyg_iPageSize", "Button_Query", "__EVENTTARGET", "__EVENTARGUMENT", "Button_SelQuery", "Button_view", "Button_edit", "Button_out", "ImageButton_Tx", "ImageButton_xx1", "Button_qd", "MdGridView_t_unitemployees_dwyg_GridViewID", "__ASYNCPOST", "MdGridView_t_unitemployees_dwyg__PageSetText" };
mc = Regex.Matches(msg, @"<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?value=""(?<val>[^""]*?)""[^<>]*?/?>|<input[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>|<select[^<>]*?name=""(?<name>[^""]*?)""[^<>]*?""[^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
if (lstParms.Contains(mi.Groups["name"].Value.Trim()))
continue;
if (dicParms.ContainsKey(mi.Groups["name"].Value.Trim()))
dicParms[mi.Groups["name"].Value.Trim()] = mi.Groups["val"].Value.Trim();
else
dicParms.Add(mi.Groups["name"].Value.Trim(), mi.Groups["val"].Value.Trim());
}
if (dicParms.ContainsKey("MdGridView_t_unitemployees_dwyg$_PageSetText"))
dicParms["MdGridView_t_unitemployees_dwyg$_PageSetText"] = "1200";
else
dicParms.Add("MdGridView_t_unitemployees_dwyg$_PageSetText", "100");//1200条 每页
msg = Post(url, dicParms, SuccessCallback, FailCallback);
dicSznyEmployees.Clear();
dicSznyEmployeeParms.Clear();
dicSznyEmployeeParms.Clear();
dicSznyEmployeeParms.Add("__EVENTTARGET", "");
dicSznyEmployeeParms.Add("__EVENTARGUMENT", "");
dicSznyEmployeeParms.Add("__VIEWSTATE", dicParms["__VIEWSTATE"]);
dicSznyEmployeeParms.Add("__EVENTVALIDATION", dicParms["__EVENTVALIDATION"]);
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_GridViewID", $"{name}");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentPage", "1");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iTotalPage", "1");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iTotalCount", "1");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iPageSize", "1");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iPageCount", "1");
dicSznyEmployeeParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentNum", "0");
dicSznyEmployeeParms.Add("XM", "ZXMCHECK");
lstParms.Clear();
lstParms = new List<string>() { "XM", "__EVENTTARGET", "__EVENTARGUMENT", "Button_Query", "Button_SelQuery", };
lstParms.Add("Button_edit");
lstParms.Add("Button_out");
lstParms.Add("ImageButton_Tx");
lstParms.Add("ImageButton_xx1");
lstParms.Add("Button_qd");
lstParms.Add("MdGridView_t_unitemployees_dwyg_GridViewID");
lstparms.add("mdgridview_t_unitemployees_dwyg_icurrentpage");
lstparms.add("mdgridview_t_unitemployees_dwyg_itotalpage");
lstparms.add("mdgridview_t_unitemployees_dwyg_itotalcount");
lstparms.add("mdgridview_t_unitemployees_dwyg_ipagesize");
lstparms.add("mdgridview_t_unitemployees_dwyg_ipagecount");
lstParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentNum");
mc = Regex.Matches(msg, @"<input[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?value=[""'](?<val>[^'""]*?)[""'][^<>]*?/?>|<input[^<>]*?name=[""'](?<name>[^'""]*?)[""'][^<>]*?[""'][^<>]*?/?>|<select[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?[""'][^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
if (lstParms.Contains(mi.Groups["name"].Value.Trim()))
continue;
if (dicSznyEmployeeParms.ContainsKey(mi.Groups["name"].Value.Trim()))
dicSznyEmployeeParms[mi.Groups["name"].Value.Trim()] = mi.Groups["val"].Value.Trim();
else
dicSznyEmployeeParms.Add(mi.Groups["name"].Value.Trim(), mi.Groups["val"].Value.Trim());
}
int cnt = int.Parse(dicSznyEmployeeParms["MdGridView_t_unitemployees_dwyg_iTotalCount"]);
for (int i = 1; i <= cnt; i++)
queueSznyEmployeeInfo.Enqueue(i);
//获取TR
//< tr[^<>] *? name = "SelectTR"[^<>] *?>.*?</ tr >
//new Regex(@"<tr[^<>]*?name=""SelectTR""[^<>]*?>.*?</tr>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
mc = Regex.Matches(msg, @"<tr[^<>]*?name=""SelectTR""[^<>]*?>.*?</tr>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
//获取td
//(?<=<td[^<>]*?>).*?(?=</td>)
//new Regex("(?<=<td[^<>]*?>).*?(?=</td>)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
MatchCollection mic = Regex.Matches(mi.Value, "(?<=<td[^<>]*?>).*?(?=</td>)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
int ix = int.Parse(mic[1].Value.Trim());
if (!dicSznyEmployees.ContainsKey(ix))
{
dicSznyEmployees.Add(ix, new Dictionary<string, string>());
}
queueSznyEmployeeInfo.Enqueue(ix);
dicSznyEmployees[ix].Add("UserName", mic[2].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("PersonID", mic[3].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Birthday", mic[4].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Sex", mic[5].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("HomePhone", mic[6].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("TelPhone", mic[7].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Mail", mic[8].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Address", mic[9].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("MinZu", mic[10].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("AddressJiGuan", mic[11].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("ZhengZhiMianmao", mic[12].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Paiqianshijian", mic[13].Value.Trim().Replace(" ", ""));
dicSznyEmployees[ix].Add("Remark", mic[14].Value.Trim().Replace(" ", ""));
}
}
return true;
}
这样所有的人员信息一次性采集到静态变量字典中了，剩下的一个工号可以慢慢获取了。
既然是这样，老实的分析Post数据，按照格式Post数据把。
分析完Post的数据后，突发奇想，我是不是可以通过相同的__ViewState和__EVENTVALIDATION POST数据呢？说干就干。
写代码跳转到员工列表页面，然后POST数据设置一页显示所有数据。
所有的POST的参数，保存到一个静态变量中。
发现POST批量提交的时候，前3次正常，以后就直接未登录。
果断放弃，换思路。
那如果这样不行 可不可以把所有的数据放到一个页面上，然后每次获取一次页面，然后根据顺序号POST数据呢。
上面已经把所有的列表数据都采集完了，顺序号也固定了，然后在POST数据的时候，发现有的人员和工号不对应。
这时候去分析为什么数据会出现不对应的情况呢？发现正则表达式写的还有问题。获取页面的Input的时候，属性有可能使用双引号，也有可能使用单引号。
正则表达式由原来的
<input[^<>]*?name="(?<name>[^"]*?)"[^<>]*?value="(?<val>[^"]*?)"[^<>]*?/?>|<input[^<>]*?name="(?<name>[^"]*?)"[^<>]*?"[^<>]*?/?>|<select[^<>]*?name="(?<name>[^"]*?)"[^<>]*?"[^<>]*?/?>
修改为
<input[^<>]*?name=["'](?<name>[^"']*?)["'][^<>]*?value=["'](?<val>[^'"]*?)["'][^<>]*?/?>|<input[^<>]*?name=["'](?<name>[^'"]*?)["'][^<>]*?["'][^<>]*?/?>|<select[^<>]*?name=["'](?<name>[^"']*?)["'][^<>]*?["'][^<>]*?/?>
由于网站异步提交，也就是以前WEBForm采用的ScriptManager,提交的时候返回的HTML不是整个Document,没有注意，以为没有返回__ViewState。所以采用GET的时候获取的__ViewState继续执行获取工号的操作。发现获取的工号都是错误，人员与工号对不上
麻爪了，不知道该咋办了。犹豫了一下后，上Fiddler吧，一点点的看提交的参数是否有区别。发现正常网站在Get到页面后，通过调整每页x条数据后，提交的ViewState与原来的不一致。寻寻觅觅 觅觅寻寻 最后发现异步返回的HTML中，最后有ViewState….
由于返回的数据顺序，每次也不一样，也是造成人员、工号不一致的原因。
提交后正常了，但是1000多条的员工信息，每次提交都是2000多个参数。看着冗长的POST数据，无语了。这样提交 先不说网站本身就慢。我提交这么多网站会不会更慢，我的系统是不是也会更慢。
怎么办？
是不是有可能把分页设置成每页只有一条数据，然后每次翻页，采集数据。简单试试把
先修改获取列表页面数据，把数据设置成一条每页，此时不再采集列表中的信息。而是记录总共多少页，放入队列中，共定时任务去分页采集数据。列表信息通过后面的分页数据采集。
由于网站是内部系统，为了不影响系统的正常运行，每次只采集一条信息，等待这条信息采集完成后，在采集下一页信息。
采集列表
public static void ReqSznyEmployeeList(int ix,Action<string> SuccessCallback, Action<string> FailCallback)
{
string url = $"http://127.0.0.1/HumanResources/EmployeeManage/EmployeeInfoList.aspx";
HttpWebRequest req = WebRequest.Create(url) as HttpWebRequest;
req.Method = "GET";
req.UserAgent = "Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/65.0.3325.181 Safari/537.36";
req.Accept = "*/*";
req.KeepAlive = true;
req.ServicePoint.ConnectionLimit = int.MaxValue;
req.ServicePoint.Expect100Continue = false;
req.CookieContainer = sznyCookie;
req.Credentials = System.Net.CredentialCache.DefaultCredentials;
req.BeginGetResponse(new AsyncCallback(RspSznyEmployeeList), new object[] { req, url,ix, SuccessCallback, FailCallback });
}
private static void RspSznyEmployeeList(IAsyncResult result)
{
object[] parms = result.AsyncState as object[];
HttpWebRequest req = parms[0] as HttpWebRequest;
string url = parms[1].ToString();
int ix = int.Parse(parms[2].ToString());
Action<string> SuccessCallback = parms[3] as Action<string>;
Action<string> FailCallback = parms[4] as Action<string>;
try
{
using (HttpWebResponse rsp = req.EndGetResponse(result) as HttpWebResponse)
{
using (StreamReader reader = new StreamReader(rsp.GetResponseStream()))
{
string msg = "";
msg = reader.ReadToEnd();
//统计参数
//__doPostBack\('(?<name>[^']*?)'
//new Regex(@"__doPostBack\('(?<name>[^']*?)'", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
string name = "";
MatchCollection mc = Regex.Matches(msg, @"__doPostBack\('(?<name>[^']*?)'", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
name = mi.Groups["name"].Value.Trim();
break;
}
//(?<=<a[^<>]*?href="javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)"[^<>]*?>条/页)
//new Regex(@"(?<=<a[^<>]*?href=""javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)""[^<>]*?>条/页)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
string smname = "MdGridView_t_unitemployees_dwyg$_SearchGo";
//Match m = Regex.Match(msg, @"(?<=<a[^<>]*?href=""javascript:__dopostback\()[^<>]*?(?=,[^<>]*?\)""[^<>]*?>条/页)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
//<input[^<>]*?name=["'](?<name>[^"']*?)["'][^<>]*?value=["'](?<val>[^'"]*?)["'][^<>]*?/?>|<input[^<>]*?name=["'](?<name>[^'"]*?)["'][^<>]*?["'][^<>]*?/?>|<select[^<>]*?name=["'](?<name>[^"']*?)["'][^<>]*?["'][^<>]*?/?>
//new Regex(@"<input[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?value=[""'](?<val>[^'""]*?)[""'][^<>]*?/?>|<input[^<>]*?name=[""'](?<name>[^'""]*?)[""'][^<>]*?[""'][^<>]*?/?>|<select[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?[""'][^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
Dictionary<string, string> dicParms = new Dictionary<string, string>();
dicParms.Add("ScriptManager1", $"UpdatePanel1|{smname}");
dicParms.Add("__EVENTTARGET", smname);
dicParms.Add("__EVENTARGUMENT", "");
dicParms.Add("__VIEWSTATE", "");
dicParms.Add("__EVENTVALIDATION", "");
dicParms.Add("MdGridView_t_unitemployees_dwyg_GridViewID", $"{name}");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentPage", ix.ToString());
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalPage", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageSize", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentNum", "0");
//dicParms.Add("MdGridView_t_unitemployees_dwyg$_SearchTextBox", ix.ToString());
dicParms.Add("XM", "ZXMCHECK");
List<string> lstParms = new List<string>() { "ScriptManager1", "XM", "MdGridView_t_unitemployees_dwyg_iCurrentNum", "Button_Query", "__EVENTTARGET", "__EVENTARGUMENT", "Button_SelQuery", "Button_view", "Button_edit", "Button_out", "ImageButton_Tx", "ImageButton_xx1", "Button_qd", "__ASYNCPOST", "MdGridView_t_unitemployees_dwyg__PageSetText" };
mc = Regex.Matches(msg, @"<input[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?value=[""'](?<val>[^'""]*?)[""'][^<>]*?/?>|<input[^<>]*?name=[""'](?<name>[^'""]*?)[""'][^<>]*?[""'][^<>]*?/?>|<select[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?[""'][^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
if (lstParms.Contains(mi.Groups["name"].Value.Trim()))
continue;
if (dicParms.ContainsKey(mi.Groups["name"].Value.Trim()))
dicParms[mi.Groups["name"].Value.Trim()] = mi.Groups["val"].Value.Trim();
else
dicParms.Add(mi.Groups["name"].Value.Trim(), mi.Groups["val"].Value.Trim());
}
if (dicParms.ContainsKey("MdGridView_t_unitemployees_dwyg$_PageSetText"))
dicParms["MdGridView_t_unitemployees_dwyg$_PageSetText"] = "1";
else
dicParms.Add("MdGridView_t_unitemployees_dwyg$_PageSetText", "1");//1200条 每页
if (dicParms.ContainsKey("MdGridView_t_unitemployees_dwyg_iPageCount"))
dicParms["MdGridView_t_unitemployees_dwyg_iPageCount"] = "1";
else
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageCount", "1");
if (dicParms.ContainsKey("MdGridView_t_unitemployees_dwyg_iPageSize"))
dicParms["MdGridView_t_unitemployees_dwyg_iPageSize"] = "1";
else
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageSize", "1");
if (dicParms.ContainsKey("MdGridView_t_unitemployees_dwyg$_SearchTextBox"))
dicParms["MdGridView_t_unitemployees_dwyg$_SearchTextBox"] = $"{ix}";
else
dicParms.Add("MdGridView_t_unitemployees_dwyg$_SearchTextBox", $"{ix}");/*第几页*/
dicParms["MdGridView_t_unitemployees_dwyg_iTotalPage"] = dicParms["MdGridView_t_unitemployees_dwyg_iTotalCount"];
msg = Post(url, dicParms, SuccessCallback, FailCallback);
//获取TR
//<tr[^<>]*?name="SelectTR"[^<>]*?>.*?</tr>
//new Regex(@"<tr[^<>]*?name=""SelectTR""[^<>]*?>.*?</tr>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
mc = Regex.Matches(msg, @"<tr[^<>]*?name=""SelectTR""[^<>]*?>.*?</tr>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
//获取td
//(?<=<td[^<>]*?>).*?(?=</td>)
//new Regex("(?<=<td[^<>]*?>).*?(?=</td>)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
MatchCollection mic = Regex.Matches(mi.Value, "(?<=<td[^<>]*?>).*?(?=</td>)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
if (!dicSznyEmployees.ContainsKey(ix))
{
dicSznyEmployees.Add(ix, new Dictionary<string, string>());
}
//queueSznyEmployeeInfo.Enqueue(ix);
if (!dicSznyEmployees[ix].ContainsKey("UserName"))
dicSznyEmployees[ix].Add("UserName", mic[2].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["UserName"] = mic[2].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("PersonID"))
dicSznyEmployees[ix].Add("PersonID", mic[3].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["PersonID"] = mic[3].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Birthday"))
dicSznyEmployees[ix].Add("Birthday", mic[4].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Birthday"] = mic[4].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Sex"))
dicSznyEmployees[ix].Add("Sex", mic[5].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Sex"] = mic[5].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("HomePhone"))
dicSznyEmployees[ix].Add("HomePhone", mic[6].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["HomePhone"] = mic[6].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("TelPhone"))
dicSznyEmployees[ix].Add("TelPhone", mic[7].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["TelPhone"] = mic[7].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Mail"))
dicSznyEmployees[ix].Add("Mail", mic[8].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Mail"] = mic[8].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Address"))
dicSznyEmployees[ix].Add("Address", mic[9].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Address"] = mic[9].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("MinZu"))
dicSznyEmployees[ix].Add("MinZu", mic[10].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["MinZu"] = mic[10].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("AddressJiGuan"))
dicSznyEmployees[ix].Add("AddressJiGuan", mic[11].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["AddressJiGuan"] = mic[11].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("ZhengZhiMianmao"))
dicSznyEmployees[ix].Add("ZhengZhiMianmao", mic[12].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["ZhengZhiMianmao"] = mic[12].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Paiqianshijian"))
dicSznyEmployees[ix].Add("Paiqianshijian", mic[13].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Paiqianshijian"] = mic[13].Value.Trim().Replace(" ", "");
if (!dicSznyEmployees[ix].ContainsKey("Remark"))
dicSznyEmployees[ix].Add("Remark", mic[14].Value.Trim().Replace(" ", ""));
else
dicSznyEmployees[ix]["Remark"] = mic[14].Value.Trim().Replace(" ", "");
}
dicParms.Clear();
mc = Regex.Matches(msg, @"(?<name>__VIEWSTATE)\|(?<v>[^\|]+)|(?<name>__EVENTVALIDATION)\|(?<v>[^\|]+)", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
dicParms.Add(mi.Groups["name"].Value.Trim(), mi.Groups["v"].Value.Trim());
}
dicParms.Add("HiddenField_param", "");
dicParms.Add("__EVENTTARGET", "");
dicParms.Add("__EVENTARGUMENT", "");
dicParms.Add("MdGridView_t_unitemployees_dwyg_GridViewID", $"{name}");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentPage", ix.ToString());
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalPage", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iTotalCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageSize", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iPageCount", "1");
dicParms.Add("MdGridView_t_unitemployees_dwyg_iCurrentNum", "0");
dicParms.Add("XM", "ZXMCHECK");
lstParms.Clear();
lstParms = new List<string>() { "XM", "__EVENTTARGET", "__EVENTARGUMENT", "Button_Query", "Button_SelQuery", };
lstParms.Add("Button_edit");
lstParms.Add("Button_out");
lstParms.Add("ImageButton_Tx");
lstParms.Add("ImageButton_xx1");
lstParms.Add("Button_qd");
mc = Regex.Matches(msg, @"<input[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?value=[""'](?<val>[^'""]*?)[""'][^<>]*?/?>|<input[^<>]*?name=[""'](?<name>[^'""]*?)[""'][^<>]*?[""'][^<>]*?/?>|<select[^<>]*?name=[""'](?<name>[^""']*?)[""'][^<>]*?[""'][^<>]*?/?>", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
foreach (Match mi in mc)
{
if (lstParms.Contains(mi.Groups["name"].Value.Trim()))
continue;
if (dicParms.ContainsKey(mi.Groups["name"].Value.Trim()))
dicParms[mi.Groups["name"].Value.Trim()] = mi.Groups["val"].Value.Trim();
else
dicParms.Add(mi.Groups["name"].Value.Trim(), mi.Groups["val"].Value.Trim());
}
ReqSznyEmployeeInfo(ix, dicParms, SuccessCallback, FailCallback);
}
}
}
catch (Exception ex) {
Business.queueSznyEmployeeInfo.Enqueue(ix);
Business.queueMsg.Enqueue($"{DateTime.Now.ToString("yyy-MM-dd HH:mm:ss")}{ex.Message}");
}
}
获取工号
public static void ReqSznyEmployeeInfo(int ix,Dictionary<string,string> dicParms, Action<string> SuccessCallback, Action<string> FailCallback) {
StringBuilder data = new StringBuilder();
foreach (var kv in dicParms)
{
if (kv.Key.StartsWith("header"))
continue;
data.Append($"&{Common.UrlEncode(kv.Key, Encoding.UTF8)}={ Common.UrlEncode(kv.Value, Encoding.UTF8)}");
}
if (data.Length > 0)
data.Remove(0, 1);
HttpWebRequest req = WebRequest.Create("http://127.0.0.1/HumanResources/EmployeeManage/EmployeeInfoList.aspx") as HttpWebRequest;
req.Method = "POST";
req.KeepAlive = true;
req.CookieContainer = sznyCookie;
req.ContentType = "application/x-www-form-urlencoded";
req.Accept = "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9";
if (dicParms.ContainsKey("ScriptManager1"))
{
req.Headers.Add("X-MicrosoftAjax", "Delta=true");
req.Headers.Add("X-Requested-With", "XMLHttpRequest");
req.ContentType = "application/x-www-form-urlencoded; charset=UTF-8";
req.Accept = "*/*";
}
req.Headers.Add("Cache-Control", "max-age=0");
req.UserAgent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/80.0.3987.163 Safari/537.36";
req.ServicePoint.ConnectionLimit = int.MaxValue;
req.ServicePoint.Expect100Continue = false;
req.AllowAutoRedirect = true;
req.Credentials = System.Net.CredentialCache.DefaultCredentials;
byte[] buffer = Encoding.UTF8.GetBytes(data.ToString());
using (Stream reqStream = req.GetRequestStream())
{
reqStream.Write(buffer, 0, buffer.Length);
}
req.BeginGetResponse(new AsyncCallback(RspSznyEmployeeInfo), new object[] { req,ix, dicParms, SuccessCallback, FailCallback });
}
private static void RspSznyEmployeeInfo(IAsyncResult result)
{
object[] parms = result.AsyncState as object[];
HttpWebRequest req = parms[0] as HttpWebRequest;
int ix =int.Parse( parms[1].ToString());
Dictionary<string, string> dicParms = parms[2] as Dictionary<string, string>;
Action<string> SuccessCallback = parms[3] as Action<string>;
Action<string> FailCallback = parms[4] as Action<string>;
try
{
using (HttpWebResponse rsp = req.EndGetResponse(result) as HttpWebResponse)
{
using (StreamReader reader = new StreamReader(rsp.GetResponseStream()))
{
string msg = "";
msg = reader.ReadToEnd();
string code = "无";
//<input[^<>]*?name\s*?=\s*?["']TextBox_YG_Code_str["'][^<>]*?value\s*?=\s*?["'](?<code>[^"']*?)["']|<input[^<>]*?value\s*?=\s*?["'](?<code>[^"']*?)["'][^<>]*?name\s*?=\s*?["']TextBox_YG_Code_str["']
//new Regex(@"<input[^<>]*?name\s*?=\s*?[""']TextBox_YG_Code_str[""'][^<>]*?value\s*?=\s*?[""'](?<code>[^""']*?)[""']|<input[^<>]*?value\s*?=\s*?[""'](?<code>[^""']*?)[""'][^<>]*?name\s*?=\s*?[""']TextBox_YG_Code_str[""']", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture)
Match m = Regex.Match(msg, @"<input[^<>]*?name\s*?=\s*?[""']TextBox_YG_Code_str[""'][^<>]*?value\s*?=\s*?[""'](?<code>[^""']*?)[""']|<input[^<>]*?value\s*?=\s*?[""'](?<code>[^""']*?)[""'][^<>]*?name\s*?=\s*?[""']TextBox_YG_Code_str[""']", RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Multiline | RegexOptions.IgnorePatternWhitespace | RegexOptions.ExplicitCapture);
if (m.Success)
code = m.Groups["code"].Value.Trim();
if (dicSznyEmployees[ix].ContainsKey("Code"))
dicSznyEmployees[ix]["Code"] = code;
else
dicSznyEmployees[ix].Add("Code", code);
queueSuccessEmployeeInfo.Enqueue(ix);
}
}
}
catch (Exception ex) {
Business.queueSznyEmployeeInfo.Enqueue(ix);
Business.queueMsg.Enqueue($"{DateTime.Now.ToString("yyy-MM-dd HH:mm:ss")}{ex.Message}");
}
}
入库
采集到的信息，通过定时任务保存到数据库。
Task.Factory.StartNew(() => {
while (true) {
if (Business.queueSuccessEmployeeInfo.Count <= 0) {
Thread.Sleep(1000);
continue;
}
List<Dictionary<string, string>> lst = new List<Dictionary<string, string>>();
while (Business.queueSuccessEmployeeInfo.Count > 0) {
Business.queueSuccessEmployeeInfo.TryDequeue(out int ix);
lst.Add(Business.dicSznyEmployees[ix]);
if (lst.Count >= 50)
break;
}
DbAccess.AddTran(lst, "SznyEmployee",new List<string>() { "UserName", "PersonID" });
Thread.Sleep(1);
}
});
总结
采集的时候，为了能利用已经采集到的信息，而不是重复采集，在采集的时候对数据库数据进行判断是否存在。纯粹是为了提高效率，WebForm的网站真是太慢，太慢了
以前写异步纯粹是为了提高线程效率，在.NET中感觉不到快乐。
终于搞定了，数据已经成功入库了。
.NET的没落也是有原因的，网站的速度的确是慢，.net押宝.net core的新体验了。
我讨厌采集WEBForm网站，写了这么久的爬虫，祈祷永远不要在碰到WEBFORM了。
