private void UpdataPartWord(string dataName, string dataValue, string instr)
{
string sql = "update partword set " + dataName + " = '" + dataValue + "' where uid in (" + instr + ")";
cmd.ExecuteNonQuery(sql);
GetPartWordData();
}
private bool AddKeyWord(string kw, int meal, int mual, int has)
{
if (checkHasData("keyword", kw))
{
MessageBox.Show("关键词 ：" + kw + " 已存在！", "提示：", MessageBoxButtons.OK, MessageBoxIcon.Information);
return(false);
}
string sql = "insert into keywords (keyword,messagealarm,musicalarm,has) values ('" + kw + "'," + meal + "," + mual + "," + has + ")";
cmd.ExecuteNonQuery(sql);
return(true);
}
private bool AddKeyWord(string url, string uname, string likeurl, int upid)
{
if (checkHasData("url", url))
{
MessageBox.Show("网址 ：" + url + " 已存在！", "提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
return(false);
}
string sql = "insert into webaddress (url,name,likeurl,pid) values ('" + url + "','" + uname + "','" + likeurl + "'," + upid + ")";
cmd.ExecuteNonQuery(sql);
return(true);
}
private void GetBaiduInfo()
{
lbAll.Text = "";
lbAll.Visible = true;
MySqlCmd cmd = new MySqlCmd();
;
//得到关键字列表
DataTable dtkey = new DataTable();
dtkey = cmd.GetTabel("select * from Keywords");
dtParts = cmd.GetTabel("SELECT * FROM partword");
//链接的正则
string aa = "http://.[^\"]+";
string[] sDate;
sb = new StringBuilder();
sb.Append("");
//TbReleaseInfo ri = new TbReleaseInfo();
//按关键字循环
for (int kw = 0; kw < dtkey.Rows.Count; kw++)
{
lbAll.Text = "正在搜索关键字为<" + dtkey.Rows[kw]["KeyWord"].ToString().Trim() + ">的数据.";
lbAll.ForeColor = Color.DarkBlue;
//取得关键字
string keys = dtkey.Rows[kw]["KeyWord"].ToString().Trim();
//组成查询字串
string url = "http://www.baidu.com/s?wd=\"" + keys + "\"&rn=50";
//得到结果放在数组内
List<string> lis = new List<string>();
lis = HtmlUtil.GetElementsByClassList(getHtml(url, "utf-8"), "result");
//如果没取到,就结束本次循环
if (lis == null) return;
//webBrowser1.Navigate(url);
//循环时判断是否要验证
bool isThere = false;
for (int i = 0; i < lis.Count; i++)
{
if (Program.ProClose == true) break;
ModelReleaseInfo mri = new ModelReleaseInfo();
//发布日期的赋值
sDate = HtmlUtil.GetElementsByTagAndClass(lis[i], "span", "g");
if (sDate.Length <= 0) continue;
mri.ReleaseDate = HtmlUtil.NoHTML(sDate[0]);
mri.ReleaseDate = mri.ReleaseDate.Substring(mri.ReleaseDate.Length - 10, 10);
//判断日期
DateTime ddt;
if (DateTime.TryParse(mri.ReleaseDate, out ddt))
{
}
else
{
//百度的快照日期有时会是9位或8位,如果是这种情况,那么按规则去掉
mri.ReleaseDate = mri.ReleaseDate.Substring(1, 9);
if (DateTime.TryParse(mri.ReleaseDate, out ddt))
{
}
else
{
mri.ReleaseDate = mri.ReleaseDate.Substring(1, 8);
}
}
//处理日期
try
{
mri.ReleaseDate = DateTime.Parse(mri.ReleaseDate).ToString("yyyy-MM-dd HH:mm:ss");
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
//只拿取三天的内的数据
try
{
if (DateTime.Parse(mri.ReleaseDate) < DateTime.Now.AddDays(-3)) continue;
}
catch (Exception ex) { continue; }
try
{
//得到标题
mri.Title = HtmlUtil.NoHTML(HtmlUtil.GetElementsByTagName(lis[i], "h3")[0]);
string[] temp = HtmlUtil.GetElementsByClass(lis[i], "c-abstract");
//如果未取到内容部分,就跳出
if (temp.Length == 0)
continue;
mri.Contexts = HtmlUtil.NoHTML(temp[0]);
mri.InfoSource = HtmlUtil.GetListByHtml("", HtmlUtil.GetElementsByTagName(lis[i], "a")[0], aa)[0];
//去掉重复
if (isThere)
{
continue;
}
else
{
if (UrlThereare(mri.Title, this.dtqueryinfo, dtWebQueryInfo, false) != 0)
{ isThere = true; continue; }
}
mri.KeyWords = dtkey.Rows[kw]["KeyWord"].ToString().Trim();
mri.CollectDate = string.Format(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss"));
mri.Snapshot = "";
mri.ReleaseName = "";
mri.WebName = "百度";
mri.Pid = 0;
mri.Part = GetParts(mri.Contexts);
mri.Comments = 0;
mri.Reposts = 0;
DataRow dr = dtqueryinfo.NewRow();
if (dvAll.RowCount == 0)
{
dr[0] = 1;
}
else
{
dr[0] = int.Parse(dvAll.Rows[dvAll.RowCount - 1].Cells[0].Value.ToString()) + 1;
}
dr[1] = mri.Title;
dr[2] = mri.Contexts;
dr[3] = mri.ReleaseDate;
dr[4] = mri.InfoSource;
dr[5] = mri.KeyWords;
dr[6] = mri.ReleaseName;
dr[7] = mri.CollectDate;
dr[8] = mri.Snapshot;
dr[9] = mri.WebName;
dr[10] = mri.Pid;
dr[11] = mri.Part;
dr[12] = mri.Reposts;
dr[13] = mri.Comments;
dtqueryinfo.Rows.InsertAt(dr, 0);
if (dtqueryinfo.Rows.Count >= 500)
{
dtqueryinfo.Rows.RemoveAt(500);
}
dvAll.Refresh();
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
try
{
//得到插入语句
if (isThere)
{
continue;
}
else
{
sb.Append(tri.GetInsString(mri) + ";");
}
//每10次执行一次插入数据库
if (sb.ToString().Length != 0)
{
if (i % 10 == 0)
{
//执行插入
cmd.ExecuteNonQuery(sb.ToString());
//清除插入字段串
sb.Clear();
}
}
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
}
}
try
{
if (sb.ToString().Length != 0)
{
//执行插入
cmd.ExecuteNonQuery(sb.ToString());
//清除插入字段串
sb.Clear();
}
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
//执行完毕后,重新获取一次数据库的数据
dtWebQueryInfo = tri.SelReleaseInfo(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss"), DateTime.Now.AddDays(-15).ToString("yyyy-MM-dd HH:mm:ss"), "0 AND webName='百度'");
//MessageBox.Show("ok");
lbAll.Text = "一轮搜索完毕！";
lbAll.ForeColor = Color.Red;
}
/// <summary>
/// 得到网站的新闻类数据
/// </summary>
private void GetWebNewsInfo()
{
lbWeb.Text = "";
lbWeb.Visible = true;
//相似链接
string Similar = "";
DataBaseServer.MySqlCmd cmd = new MySqlCmd();
//得到相似表
DataTable dtXs = new DataTable();
dtXs = cmd.GetTabel("Select * from WebAddress WHERE pid=0");
dtParts = cmd.GetTabel("SELECT * FROM partword");
DataTable dtkey = new DataTable();
dtkey = cmd.GetTabel("select * from Keywords");
//相似表中的被抓取网址
string webInfo = "";
//要过滤链接中首页的正则
string strTopFormat = "http://.+/";
List<string> strTop = new List<string>();
sb = new StringBuilder();
sb.Append("");
string filterStr = "";
#region 读取相似度表中的数据据,循环抓取
for (int xs = 0; xs < dtXs.Rows.Count; xs++)
{
lbWeb.ForeColor = Color.DarkBlue;
lbWeb.Text = "正在搜索:" + dtXs.Rows[xs]["name"].ToString();
//读取相似表中要抓取的网址
webInfo = getHtml(dtXs.Rows[xs]["url"].ToString(), "");
//读取相似链接
Similar = dtXs.Rows[xs]["likeurl"].ToString();
//取出
//string[] strA = HtmlUtil.GetElementsByTagName(webInfo, "a");
List<string> strList = HtmlUtil.GetElementsByTagNameList(webInfo, "a");
string strURLformat = "http://.[^\"]+";
TbReleaseInfo ri = new TbReleaseInfo();
string[] strA = GetLIstDate(strList.Distinct());
#region 逐个链接判断
//循环时判断是否要验证
bool isThere = false;
for (int i = 0; i < strA.Length; i++)
{
if (Program.ProClose == true) break;
Application.DoEvents();
try
{
//得到目标网址中的所有链接,如果未得到,那么就继续读取下一个
strA[i] = HtmlUtil.GetListByHtml(dtXs.Rows[xs]["url"].ToString(), strA[i], strURLformat)[0];
//处理含有单引号的链接
strA[i] = UrlCl(strA[i]);
//处理单引号的链接
if (strA[i].IndexOf("'") != -1)
{
strA[i] = GetstringByHtmlArray(strA[i], "http://.[^\']+");
}
}
catch (Exception)
{
continue;
}
//得到相似值,大于0.70的认为相同,并开始抓取
if (HtmlUtil.getSimilarDegree(Similar, strA[i]) >= 0.60)
{
//判断这个链接是否已经在库中或者列表中,如果存在,此次就不再执行
strTop = HtmlUtil.GetListByHtmlArray(strA[i], strTopFormat);
if (strTop.Count != 0)
{
//if (strTop[0] == "http://blog.sohu.com/")
continue;//同新闻,如果将首页去掉
}
if (isThere)
{
continue;
}
else
{
//if (strA[i] == "http://news.ifeng.com/mainland/detail_2013_10/18/30459577_0.shtml'>[详细]</a>")
//{
// strA[i] = strA[i];
//}
if (UrlThereare(strA[i], this.dtnewsinfo, dtWebNewsInfo, true) != 0) { isThere = true; continue; }
}
//得到此链接的源码
webInfo = getHtml(strA[i], "");
if (webInfo.Length == 0) { continue; }
//创建数据对象
ModelReleaseInfo newsInfo = new ModelReleaseInfo();
try
{
//流水+1
newsInfo.Uid = this.dvAll.Rows.Count + 1;
//标题
string[] strT = HtmlUtil.GetElementsByTagName(webInfo, "title");
if (strT.Length == 0)
{
continue;
}
else
{
newsInfo.Title = HtmlUtil.NoHTML(HtmlUtil.GetElementsByTagName(webInfo, "title")[0]);
}
//得到正文,以P标签来区分
string[] strContext = HtmlUtil.GetElementsByTagName(webInfo, "p");
newsInfo.Contexts = "";
for (int j = 0; j < strContext.Length; j++)
{
//循环累加正文信息
newsInfo.Contexts += HtmlUtil.NoHTML(strContext[j]);
}
//如果正文信息为空,那么将无法做关键字对照,此条数据舍弃
if (newsInfo.Contexts.Length == 0)
{
continue;
}
//网站链接
newsInfo.InfoSource = strA[i].Trim();
//关键字的设置
newsInfo.KeyWords = "";
for (int j = 0; j < dtkey.Rows.Count; j++)
{
Application.DoEvents();
if (newsInfo.Contexts.IndexOf(dtkey.Rows[j][1].ToString()) > 0)
{ newsInfo.KeyWords += dtkey.Rows[j][1].ToString() + ","; }
else
{
}
}
if (newsInfo.KeyWords.Length == 0) { continue; }
newsInfo.KeyWords = newsInfo.KeyWords.Substring(0, newsInfo.KeyWords.Length - 1);
//收集日期
newsInfo.CollectDate = string.Format(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss"));
//发布人和发布日期暂时无法取到,手工赋值为空
newsInfo.ReleaseDate = "";
newsInfo.ReleaseName = "";
//网页快照,这里为用户指定生成,如果未选择生成,那么为空
newsInfo.Snapshot = "";
//网站名
newsInfo.WebName = dtXs.Rows[xs]["Name"].ToString();
//pid
newsInfo.Pid = 0;
//part正负判断
newsInfo.Part = GetParts(newsInfo.Contexts);
//reposts
newsInfo.Reposts = 0;
//comments
newsInfo.Comments = 0;
//新建数据行
DataRow dr = dtnewsinfo.NewRow();
if (dvWeb.RowCount == 0)
{
dr[0] = 1;
}
else
{
dr[0] = int.Parse(dvWeb.Rows[dvWeb.RowCount - 1].Cells[0].Value.ToString()) + 1;
}
//dr[0] = newsInfo.Uid;
dr[1] = newsInfo.Title;
dr[2] = newsInfo.Contexts;
dr[3] = newsInfo.ReleaseDate;
dr[4] = newsInfo.InfoSource;
dr[5] = newsInfo.KeyWords;
dr[6] = newsInfo.ReleaseName;
dr[7] = newsInfo.CollectDate;
dr[8] = newsInfo.Snapshot;
dr[9] = newsInfo.WebName;
dr[10] = newsInfo.Pid;
dr[11] = newsInfo.Part;
dr[12] = newsInfo.Reposts;
dr[13] = newsInfo.Comments;
//把行加到DT中
dtnewsinfo.Rows.InsertAt(dr, 0);
//数据源刷新
if (dtnewsinfo.Rows.Count >= 500)
{
dtnewsinfo.Rows.RemoveAt(500);
}
dvWeb.Refresh();
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
////总表刷新
//dt.Rows.Add(dr);
//dvAll.Refresh();
//得到插入语句
try
{
if (isThere)
{
continue;
}
else
{
sb.Append(ri.GetInsString(newsInfo) + ";");
}
//每10次执行一次插入数据库
if (sb.ToString().Length != 0)
{
if (i % 10 == 0)
{
filterStr = sb.ToString();
filterStr = filterStr.Replace("[ ", "[");
filterStr = filterStr.Replace(" ]", "]");
//执行插入
cmd.ExecuteNonQuery(filterStr);
//清除插入字段串
sb.Clear();
filterStr = "";
}
}
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
}
}
#endregion
}
#endregion
try
{
if (sb.ToString().Length != 0)
{
filterStr = sb.ToString();
filterStr = filterStr.Replace("[ ", "[");
filterStr = filterStr.Replace(" ]", "]");
//执行插入
cmd.ExecuteNonQuery(filterStr);
//清除插入字段串
sb.Clear();
filterStr = "";
}
}
catch (Exception ex)
{
StreamWriter sw = File.AppendText("log.txt");
sw.WriteLine(DateTime.Now.ToLongDateString());
sw.WriteLine("begin");
sw.WriteLine(ex.Message);
sw.WriteLine(sb.ToString());
sw.WriteLine("end");
sw.WriteLine("");
sw.Close();
}
//执行完毕后,重新获取一次数据库的数据
dtWebNewsInfo = tri.SelReleaseInfo(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss"), DateTime.Now.AddDays(-15).ToString("yyyy-MM-dd HH:mm:ss"), "0 AND webName<>'百度'");
//MessageBox.Show("ok");
lbWeb.Text = "一轮搜索完毕！";
lbWeb.ForeColor = Color.Red;
}
public void StartWrite()
{
Action actionG = () =>
{
List<ModelReleaseInfo> data;
while (true)
{
if (Program.ProClose)
{
break;
}
if (Queue.TryDequeue(out data))
{
//// 写入数据库
#region 数据入库
try
{
StringBuilder sb = new StringBuilder();
sb.Append("");
TbReleaseInfo tri = new TbReleaseInfo();
MySqlCmd cmd = new MySqlCmd();
foreach (var mri in data)
{
#region 2016.11.16 加入精确匹配的判断
string keywords = mri.KeyWords;
string title = mri.Title;
string context = mri.Contexts;
if (!string.IsNullOrEmpty(keywords))
{
bool isFundTitle = true;
bool isFundContext = true;
string[] keyw = keywords.Split(' ');
if (keyw != null && keyw.Count() > 0)
{
foreach (string key in keyw)
{
if (title.IndexOf(key) < 0)
{
isFundTitle = false;
}
if (context.IndexOf(key) < 0)
{
isFundContext = false;
}
}
}
if (!isFundTitle && !isFundContext)
{
//如果标题或者内容没有匹配全部关键字则去掉该条数据
continue;
}
}
#endregion
if (tri.GetReleaseInfoCount(mri.InfoSource, mri.KeyWords) > 0) continue;
string sql = tri.GetInsertStr(mri);
if (!sql.Trim().EndsWith(";"))
{
sql += sql + ";";
}
sb.Append(sql);
}
if (sb.ToString().Length > 0)
{
//执行插入
cmd.ExecuteNonQuery(sb.ToString());
//清除插入字段串
sb.Clear();
}
}
catch (Exception ex)
{
Comm.WriteErrorLog(ex.Message);
Comm.WriteErrorLog(ex.StackTrace);
}
#endregion
log.Info("数据层写入数据库成功");
}
else
{
System.Threading.Thread.Sleep(1000);
}
}
};
Parallel.Invoke(actionG, actionG);
}