public virtual List <JObject> Extract(Page page)
{
List <JObject> result = new List <JObject>();
if (GlobalValues != null && GlobalValues.Count > 0)
{
foreach (var enviromentValue in GlobalValues)
{
string name = enviromentValue.Name;
var value = page.Selectable.Select(SelectorUtil.Parse(enviromentValue)).GetValue();
page.Request.PutExtra(name, value);
}
}
ISelector selector = SelectorUtil.Parse(EntityMetadata.Selector);
if (selector != null && EntityMetadata.Multi)
{
var list = page.Selectable.SelectList(selector).Nodes();
if (list == null || list.Count == 0)
{
result = null;
}
else
{
if (EntityMetadata.Take > 0)
{
list = list.Take(EntityMetadata.Take).ToList();
}
int index = 0;
foreach (var item in list)
{
var obj = ExtractSingle(page, item, index);
if (obj != null)
{
result.Add(obj);
}
index++;
}
}
}
else
{
ISelectable select = selector == null ? page.Selectable : page.Selectable.Select(selector);
if (select != null)
{
var singleResult = ExtractSingle(page, select, 0);
result = singleResult != null ? new List <JObject> {
singleResult
} : null;
}
else
{
result = null;
}
};
return(result);
}
/// <summary>
/// 解析成爬虫实体对象
/// </summary>
/// <param name="page">页面数据</param>
/// <returns>爬虫实体对象</returns>
public List <T> Extract(Page page)
{
List <T> result = new List <T>();
if (EntityDefine.SharedValues != null && EntityDefine.SharedValues.Count > 0)
{
foreach (var enviromentValue in EntityDefine.SharedValues)
{
string name = enviromentValue.Name;
var value = page.Selectable.Select(SelectorUtil.Parse(enviromentValue)).GetValue();
page.Request.PutExtra(name, value);
}
}
ISelector selector = SelectorUtil.Parse(EntityDefine.Selector);
if (selector != null && EntityDefine.Multi)
{
var list = page.Selectable.SelectList(selector).Nodes();
if (list == null || list.Count == 0)
{
result = null;
}
else
{
if (EntityDefine.Take > 0)
{
list = list.Take(EntityDefine.Take).ToList();
}
for (int i = 0; i < list.Count; ++i)
{
var item = list[i];
var obj = ExtractSingle(page, item, i);
if (obj != null)
{
result.Add(obj);
}
}
}
}
else
{
ISelectable select = selector == null ? page.Selectable : page.Selectable.Select(selector);
if (select != null)
{
var item = ExtractSingle(page, select, 0);
result = item != null ? new List <T> {
item
} : null;
}
else
{
result = null;
}
}
return(result);
}
public bool NeedStop(Page page, BaseTargetUrlsCreator creator)
{
var tmps = page.Selectable.SelectList(SelectorUtil.Parse(TimeSelector)).GetValues();
if (tmps == null)
{
return(true);
}
List <string> timeStrings = new List <string>();
foreach (var c in tmps)
{
var s = c;
if (TimeFormatters != null)
{
foreach (var formatter in TimeFormatters)
{
s = formatter.Formate(s);
}
}
timeStrings.Add(s);
}
foreach (var c in timeStrings)
{
var dt = DateTime.Parse(c);
if (IsBefore)
{
foreach (var stopper in Times)
{
var stopDate = DateTime.Parse(stopper);
if (dt < stopDate)
{
return(true);
}
}
}
else
{
foreach (var stopper in Times)
{
var stopDate = DateTime.Parse(stopper);
if (dt > stopDate)
{
return(true);
}
}
}
}
return(false);
}
public override bool CanStop(Page page)
{
var current = page.Selectable.SelectList(SelectorUtil.Parse(CurrenctPageSelector)).GetValues();
if (current == null)
{
return(true);
}
List <string> timeStrings = new List <string>();
foreach (var c in current)
{
var s = c;
if (CurrenctPageFormatters != null)
{
foreach (var formatter in CurrenctPageFormatters)
{
s = formatter.Formate(s);
}
}
timeStrings.Add(s);
}
foreach (var c in timeStrings)
{
var dt = DateTime.Parse(c.ToString());
if (IsBefore)
{
foreach (var stopper in Stoppers)
{
var stopDate = DateTime.Parse(stopper);
if (dt < stopDate)
{
return(true);
}
}
}
else
{
foreach (var stopper in Stoppers)
{
var stopDate = DateTime.Parse(stopper);
if (dt > stopDate)
{
return(true);
}
}
}
}
return(false);
}
internal Model.TargetUrlExtractor GetTargetUrlExtractInfo()
{
var t = new Model.TargetUrlExtractor
{
Formatters = Formatters,
Region = SelectorUtil.Parse(Region)
};
foreach (var p in Patterns)
{
if (!string.IsNullOrEmpty(p?.Trim()))
{
t.Patterns.Add(new Regex(p));
}
}
return(t);
}
public bool NeedStop(Page page, BaseTargetUrlsCreator creator)
{
int totalPage = -2000;
if (TotalPageSelector != null)
{
string totalStr = page.Selectable.Select(SelectorUtil.Parse(TotalPageSelector)).GetValue();
if (TotalPageFormatters != null)
{
foreach (var formatter in TotalPageFormatters)
{
totalStr = formatter.Formate(totalStr);
}
}
if (!string.IsNullOrEmpty(totalStr))
{
totalPage = int.Parse(totalStr);
}
}
int currentPage = -1000;
if (CurrenctPageSelector != null)
{
string currentStr = page.Selectable.Select(SelectorUtil.Parse(CurrenctPageSelector)).GetValue();
if (CurrnetPageFormatters != null)
{
foreach (var formatter in CurrnetPageFormatters)
{
currentStr = formatter.Formate(currentStr);
}
}
if (!string.IsNullOrEmpty(currentStr))
{
currentPage = int.Parse(currentStr);
}
}
if (currentPage == totalPage)
{
return(true);
}
return(false);
}
public int?Interval(Page page)
{
var intervalStr = page.Selectable.Select(SelectorUtil.Parse(Selector)).GetValue();
if (!string.IsNullOrEmpty(intervalStr))
{
if (IntervalFormatters != null)
{
foreach (var formatter in IntervalFormatters)
{
intervalStr = formatter.Formate(intervalStr);
}
}
if (!string.IsNullOrEmpty(intervalStr))
{
return(int.Parse(intervalStr));
}
}
return(null);
}
private dynamic ExtractField(ISelectable item, Page page, Field field, int index)
{
if (field == null)
{
return(null);
}
ISelector selector = SelectorUtil.Parse(field.Selector);
if (selector == null)
{
return(null);
}
string tmpValue;
if (selector is EnviromentSelector)
{
var enviromentSelector = selector as EnviromentSelector;
tmpValue = GetEnviromentValue(enviromentSelector.Field, page, index);
foreach (var formatter in field.Formatters)
{
tmpValue = formatter.Formate(tmpValue);
}
return(tmpValue);
}
else
{
bool needPlainText = field.Option == PropertyDefine.Options.PlainText;
if (field.Multi)
{
var propertyValues = item.SelectList(selector).Nodes();
List <string> results = new List <string>();
foreach (var propertyValue in propertyValues)
{
results.Add(propertyValue.GetValue(needPlainText));
}
foreach (var formatter in field.Formatters)
{
results = formatter.Formate(results);
}
return(new JArray(results));
}
else
{
bool needCount = field.Option == PropertyDefine.Options.Count;
if (needCount)
{
var propertyValues = item.SelectList(selector).Nodes();
string count = propertyValues?.Count.ToString();
count = string.IsNullOrEmpty(count) ? "-1" : count;
return(count);
}
else
{
tmpValue = item.Select(selector)?.GetValue(needPlainText);
foreach (var formatter in field.Formatters)
{
tmpValue = formatter.Formate(tmpValue);
}
return(tmpValue);
}
}
}
}
private object ExtractField(ISelectable item, Page page, Column field, int index)
{
if (field == null)
{
return(null);
}
ISelector selector = SelectorUtil.Parse(field.Selector);
if (selector == null)
{
return(null);
}
if (selector is EnviromentSelector)
{
var enviromentSelector = selector as EnviromentSelector;
var value = SelectorUtil.GetEnviromentValue(enviromentSelector.Field, page, index);
foreach (var formatter in field.Formatters)
{
#if DEBUG
try
{
#endif
value = formatter.Formate(value);
#if DEBUG
}
catch (Exception e)
{
Logger.NLog(e.ToString(), Level.Error);
}
#endif
}
return(TryConvert(value, field.DataType));
}
else
{
bool needCount = field.Option == PropertyDefine.Options.Count;
if (needCount)
{
var values = item.SelectList(selector).Nodes();
return(values.Count);
}
else
{
var value = (object)item.Select(selector)?.GetValue(field.Option == PropertyDefine.Options.PlainText);
foreach (var formatter in field.Formatters)
{
#if DEBUG
try
{
#endif
value = formatter.Formate(value);
#if DEBUG
}
catch (Exception e)
{
Logger.NLog(e.ToString(), Level.Error);
}
#endif
}
return(TryConvert(value, field.DataType));
}
}
}
public bool NeedStop(Page page, BaseTargetUrlsCreator creator)
{
int totalPage = -2000;
if (TotalPageSelector != null)
{
string totalStr = string.Empty;
if (TotalPageSelector.Type == SelectorType.Enviroment)
{
var selector = SelectorUtil.Parse(TotalPageSelector) as EnviromentSelector;
if (selector != null)
{
totalStr = EntityExtractor.GetEnviromentValue(selector.Field, page, 0);
}
}
else
{
totalStr = page.Selectable.Select(SelectorUtil.Parse(TotalPageSelector)).GetValue();
}
if (!string.IsNullOrEmpty(totalStr))
{
if (TotalPageFormatters != null)
{
foreach (var formatter in TotalPageFormatters)
{
totalStr = formatter.Formate(totalStr);
}
}
if (!string.IsNullOrEmpty(totalStr))
{
totalPage = int.Parse(totalStr);
}
}
}
int currentPage = -1000;
if (CurrenctPageSelector != null)
{
string currentStr = string.Empty;
if (CurrenctPageSelector.Type == SelectorType.Enviroment)
{
var selector = SelectorUtil.Parse(CurrenctPageSelector) as EnviromentSelector;
if (selector != null)
{
currentStr = EntityExtractor.GetEnviromentValue(selector.Field, page, 0);
}
}
else
{
currentStr = page.Selectable.Select(SelectorUtil.Parse(CurrenctPageSelector)).GetValue();
}
if (!string.IsNullOrEmpty(currentStr))
{
if (CurrnetPageFormatters != null)
{
foreach (var formatter in CurrnetPageFormatters)
{
currentStr = formatter.Formate(currentStr);
}
}
if (!string.IsNullOrEmpty(currentStr))
{
currentPage = int.Parse(currentStr);
}
}
}
if (currentPage == totalPage)
{
return(true);
}
return(false);
}
private dynamic ExtractField(ISelectable item, Page page, DataToken field, int index)
{
ISelector selector = SelectorUtil.Parse(field.Selector);
if (selector == null)
{
return(null);
}
var f = field as Field;
bool isEntity = field is Entity;
if (!isEntity)
{
string tmpValue;
if (selector is EnviromentSelector)
{
var enviromentSelector = selector as EnviromentSelector;
tmpValue = GetEnviromentValue(enviromentSelector.Field, page, index);
if (f != null)
{
foreach (var formatter in f.Formatters)
{
tmpValue = formatter.Formate(tmpValue);
}
}
return(tmpValue);
}
else
{
bool needPlainText = ((Field)field).Option == PropertySelector.Options.PlainText;
if (field.Multi)
{
var propertyValues = item.SelectList(selector).Nodes();
List <string> results = new List <string>();
foreach (var propertyValue in propertyValues)
{
results.Add(propertyValue.GetValue(needPlainText));
}
if (f != null)
{
foreach (var formatter in f.Formatters)
{
results = formatter.Formate(results);
}
}
return(new JArray(results));
}
else
{
bool needCount = (((Field)field).Option == PropertySelector.Options.Count);
if (needCount)
{
var propertyValues = item.SelectList(selector).Nodes();
return(propertyValues?.Count.ToString() ?? "-1");
}
else
{
tmpValue = item.Select(selector)?.GetValue(needPlainText);
if (f != null)
{
foreach (var formatter in f.Formatters)
{
tmpValue = formatter.Formate(tmpValue);
}
}
return(tmpValue);
}
}
}
}
else
{
if (field.Multi)
{
JArray objs = new JArray();
var selectables = item.SelectList(selector).Nodes();
foreach (var selectable in selectables)
{
JObject obj = new JObject();
foreach (var child in ((Entity)field).Fields)
{
obj.Add(child.Name, ExtractField(selectable, page, child, 0));
}
objs.Add(obj);
}
return(objs);
}
else
{
JObject obj = new JObject();
var selectable = item.Select(selector);
foreach (var child in ((Entity)field).Fields)
{
obj.Add(child.Name, ExtractField(selectable, page, field, 0));
}
return(obj);
}
}
}
public List <JObject> Process(Page page)
{
List <JObject> result = new List <JObject>();
bool isTarget = true;
foreach (var targetUrlExtractor in EntityMetadata.TargetUrlExtractors)
{
foreach (var regex in targetUrlExtractor.Regexes)
{
isTarget = regex.IsMatch(page.Url);
if (isTarget)
{
break;
}
}
}
if (!isTarget)
{
return(null);
}
if (_globalValues != null && _globalValues.Count > 0)
{
foreach (var enviromentValue in _globalValues)
{
string name = enviromentValue.Name;
var value = page.Selectable.Select(SelectorUtil.Parse(enviromentValue)).GetValue();
page.Request.PutExtra(name, value);
}
}
ISelector selector = SelectorUtil.Parse(EntityMetadata.Entity.Selector);
if (selector != null && EntityMetadata.Entity.Multi)
{
var list = page.Selectable.SelectList(selector).Nodes();
if (list == null || list.Count == 0)
{
result = null;
}
else
{
var countToken = EntityMetadata.Limit;
if (countToken != null)
{
list = list.Take(countToken.Value).ToList();
}
int index = 0;
foreach (var item in list)
{
JObject obj = ProcessSingle(page, item, index);
if (obj != null)
{
result.Add(obj);
}
index++;
}
}
}
else
{
ISelectable select = selector == null ? page.Selectable : page.Selectable.Select(selector);
if (select != null)
{
var singleResult = ProcessSingle(page, select, 0);
result = new List <JObject> {
singleResult
};
}
else
{
result = null;
}
}
//if (EntityMetadata.TargetUrlsCreators != null && EntityMetadata.TargetUrlExtractors.Count > 0)
//{
// foreach (var targetUrlsCreator in EntityMetadata.TargetUrlsCreators)
// {
// page.AddTargetRequests(targetUrlsCreator.Handle(page));
// }
//}
if (!page.MissExtractTargetUrls)
{
ExtractLinks(page, EntityMetadata.TargetUrlExtractors);
}
return(result);
}
public dynamic Process(Page page)
{
if (_enviromentValues != null && _enviromentValues.Count > 0)
{
foreach (var enviromentValue in _enviromentValues)
{
string name = enviromentValue.Name;
var value = page.Selectable.Select(SelectorUtil.Parse(enviromentValue.Selector)).GetValue();
page.Request.PutExtra(name, value);
}
}
bool isMulti = false;
ISelector selector = SelectorUtil.Parse(_entityDefine.Selector);
if (selector == null)
{
isMulti = false;
}
else
{
isMulti = _entityDefine.Multi;
}
if (isMulti)
{
var list = page.Selectable.SelectList(selector).Nodes();
if (list == null || list.Count == 0)
{
return(null);
}
var countToken = _entityDefine.Limit;
if (countToken != null)
{
list = list.Take(countToken.Value).ToList();
}
List <JObject> result = new List <JObject>();
int index = 0;
foreach (var item in list)
{
JObject obj = ProcessSingle(page, item, _entityDefine, index);
if (obj != null)
{
result.Add(obj);
}
index++;
}
return(result);
}
else
{
ISelectable select;
if (selector == null)
{
select = page.Selectable;
}
else
{
select = page.Selectable.Select(selector);
if (select == null)
{
return(null);
}
}
return(ProcessSingle(page, select, _entityDefine, 0));
}
}
