public static IEnumerable <List <CrawlItem> > SearchPropertiesSmart(this HtmlDocument doc2,
ICollection <CrawlItem> existItems = null, bool isAttrEnabled = false)
{
if (existItems == null)
{
existItems = new List <CrawlItem>();
}
var shortv = "";
if (existItems.Count > 1)
{
shortv =
XPath.GetMaxCompareXPath(existItems.Select(d => new XPath(d.XPath)).ToList()).ToString();
yield return(GetDiffNodes(doc2, shortv, isAttrEnabled, existItems));
}
else if (existItems.Count == 1)
{
var realPath = new XPath(existItems.First().XPath);
var array =
realPath.SelectAll(d => true)
.Select(d => new XPath(realPath.Take(d)).ToString()).ToList();
var dict = new Dictionary <string, double>();
foreach (var item in array)
{
GetTableRootProbability(
doc2.DocumentNode.SelectSingleNode(item)
.ChildNodes.Where(d2 => d2.Name.Contains("#") == false)
.ToList(), dict, false);
}
foreach (
var item in
dict.OrderByDescending(d => d.Value))
{
shortv = item.Key;
yield return(GetDiffNodes(doc2, shortv, isAttrEnabled, existItems));
}
}
else
{
var dict = new Dictionary <string, double>();
GetTableRootProbability(
doc2.DocumentNode.ChildNodes.Where(d => d.Name.Contains("#") == false).ToList(), dict, true);
IEnumerable <KeyValuePair <string, double> > p = dict.OrderByDescending(d => d.Value);
foreach (var keyValuePair in p)
{
var items = GetDiffNodes(doc2, keyValuePair.Key, isAttrEnabled, existItems);
if (items.Count > 1)
{
yield return(items);
}
}
}
}
public static string CompileCrawItems(this HtmlDocument doc2, IList <CrawlItem> crawlItem)
{
var shortv =
XPath.GetMaxCompareXPath(crawlItem.Select(d => new XPath(d.XPath)).ToList()).ToString();
if (!string.IsNullOrEmpty(shortv))
{
crawlItem.Execute(d => d.XPath = new XPath(d.XPath).TakeOff(shortv).ToString());
return(shortv);
}
return("");
}
public override bool Init(IEnumerable <IFreeDocument> docus)
{
base.Init(docus);
if (Crawler == null)
{
return(false);
}
IsMultiYield = true;
xpaths = Crawler.CrawlItems.GroupBy(d => d.Name).Select(d =>
{
var column = d.Key;
var path = XPath.GetMaxCompareXPath(d.Select(d2 => d2.XPath).ToList());
return(new { Column = column, XPath = path });
}).ToDictionary(d => d.Column, d => d.XPath);
return(true);
}
public static List <FreeDocument> GetDataFromXPath(this HtmlDocument doc2, IList <CrawlItem> crawlItems,
ListType type = ListType.List, string rootXPath = "")
{
if (crawlItems.Count == 0)
{
return(new List <FreeDocument>());
}
var documents = new List <FreeDocument>();
switch (type)
{
case ListType.List:
var root = "";
var takeoff = "";
if (string.IsNullOrEmpty(rootXPath))
{
root =
XPath.GetMaxCompareXPath(crawlItems.Select(d => new XPath(d.XPath)).ToList()).ToString();
takeoff = root;
}
else
{
root = rootXPath;
}
var nodes = doc2.DocumentNode.SelectNodes(root);
if (nodes == null)
{
break;
}
foreach (var node in nodes)
{
var document = new FreeDocument();
foreach (var r in crawlItems)
{
string path;
if (string.IsNullOrEmpty(takeoff))
{
path = node.XPath + r.XPath;
}
else
{
path = node.XPath + new XPath(r.XPath).TakeOff(takeoff);
}
var result = node.GetDataFromXPath(path, r.IsHTML);
document.SetValue(r.Name, result);
}
documents.Add(document);
}
return(documents);
case ListType.One:
var freeDocument = new FreeDocument();
foreach (var r in crawlItems)
{
doc2.GetDataFromXPath(r, freeDocument);
}
return(new List <FreeDocument> {
freeDocument
});
}
return(new List <FreeDocument>());
}
public static IEnumerable <CrawTarget> SearchPropertiesSmart(this HtmlDocument doc2,
ICollection <CrawlItem> existItems = null, string rootPath = null, bool isAttrEnabled = false)
{
if (existItems == null)
{
existItems = new List <CrawlItem>();
}
var shortv = "";
var dict = new Dictionary <string, double>();
if (string.IsNullOrEmpty(rootPath))
{
var isForceItemOne = false; //强制进入只有一个属性的模式
if (existItems.Count > 1)
{
shortv =
XPath.GetMaxCompareXPath(existItems.Select(d => d.XPath));
var nodes = doc2.DocumentNode.SelectNodes(shortv);
if (nodes == null || nodes.Count == 0)
{
yield break;
}
if (nodes.Count == 1)
{
isForceItemOne = true;
}
else
{
var items = GetDiffNodes(doc2, shortv, isAttrEnabled, existItems, 1);
var target = getCrawTarget(items, shortv);
if (target != null)
{
yield return(target);
}
}
}
if (isForceItemOne || existItems.Count == 1)
{
var realPath = existItems.First().XPath;
var items = XPath.Split(realPath);
var array =
items.SelectAll(d => true)
.Select(d => XPath.SubXPath(realPath, d)).ToList();
foreach (var item in array)
{
GetTableRootProbability(
doc2.DocumentNode.SelectSingleNode(item)
.ChildNodes.Where(d2 => d2.Name.Contains("#") == false)
.ToList(), dict, false);
}
}
else
{
GetTableRootProbability(
doc2.DocumentNode.ChildNodes.Where(d => d.Name.Contains("#") == false).ToList(), dict, true);
}
if (isForceItemOne || existItems.Count < 2)
{
IEnumerable <KeyValuePair <string, double> > p = dict.OrderByDescending(d => d.Value);
foreach (var keyValuePair in p)
{
var items = GetDiffNodes(doc2, keyValuePair.Key, isAttrEnabled, existItems, 4);
if (items.Count == 0)
{
continue;
}
var target = getCrawTarget(items, keyValuePair.Key);
var rootNode = doc2.DocumentNode.SelectSingleNode(keyValuePair.Key).ParentNode;
if (rootNode == null)
{
continue;
}
target.Html = rootNode.InnerHtml;
target.Text = rootNode.InnerText;
target.NodeCount = doc2.DocumentNode.SelectNodes(keyValuePair.Key).Count;
target.Score = keyValuePair.Value;
target.ColumnCount = items.Count;
yield return(target);
}
}
}
else
{
var items = GetDiffNodes(doc2, rootPath, isAttrEnabled, new List <CrawlItem>());
if (items.Count > 0)
{
var root = doc2.DocumentNode.SelectSingleNode(rootPath);
var xpath = XPath.RemoveFinalNum(root.XPath);
var target = getCrawTarget(items, xpath);
target.RootXPath = rootPath;
yield return(target);
}
}
}
public static IEnumerable <CrawTarget> SearchPropertiesSmart(this HtmlDocument doc2,
ICollection <CrawlItem> existItems = null, string rootPath = null, bool isAttrEnabled = false)
{
if (existItems == null)
{
existItems = new List <CrawlItem>();
}
var shortv = "";
var dict = new Dictionary <string, double>();
if (string.IsNullOrEmpty(rootPath))
{
if (existItems.Count > 1)
{
shortv =
XPath.GetMaxCompareXPath(existItems.Select(d => d.XPath));
var items = GetDiffNodes(doc2, shortv, isAttrEnabled, existItems, 1);
var target = getCrawTarget(items, shortv);
if (target != null)
{
yield return(target);
}
yield break;
}
if (existItems.Count == 1)
{
var realPath = existItems.First().XPath;
var items = XPath.Split(realPath);
var array =
items.SelectAll(d => true)
.Select(d => XPath.SubXPath(realPath, d)).ToList();
foreach (var item in array)
{
GetTableRootProbability(
doc2.DocumentNode.SelectSingleNode(item)
.ChildNodes.Where(d2 => d2.Name.Contains("#") == false)
.ToList(), dict, false);
}
}
else
{
GetTableRootProbability(
doc2.DocumentNode.ChildNodes.Where(d => d.Name.Contains("#") == false).ToList(), dict, true);
}
if (existItems.Count < 2)
{
IEnumerable <KeyValuePair <string, double> > p = dict.OrderByDescending(d => d.Value);
foreach (var keyValuePair in p)
{
var items = GetDiffNodes(doc2, keyValuePair.Key, isAttrEnabled, existItems, 4);
var target = getCrawTarget(items, keyValuePair.Key);
if (target != null)
{
yield return(target);
}
}
}
}
else
{
var items = GetDiffNodes(doc2, rootPath, isAttrEnabled, new List <CrawlItem>());
if (items.Count > 0)
{
var root = doc2.DocumentNode.SelectSingleNode(rootPath);
foreach (var crawlItem in items)
{
crawlItem.XPath = XPath.TakeOff(crawlItem.XPath, root.XPath);
}
var target = getCrawTarget(items, rootPath);
if (target != null)
{
yield return(target);
}
}
}
}