protected spiderModuleBase(string __name, string __desc, ISpiderEvaluatorBase __parent)
{
name = __name;
code = name[0].ToString().ToUpper();
description = __desc;
_parent = __parent;
}
public void eventDLCInitiated(ISpiderEvaluatorBase __spider, crawlerDomainTask __task, modelSpiderSiteRecord __wRecord)
{
foreach (IPlugInCommonBase <crawlReportingStageEnum, directReporterBase> plug in allPlugins)
{
plug.eventDLCInitiated(null, __task, __wRecord); /// aJob, __machine, __tRecord);
}
}
public languageModule(ISpiderEvaluatorBase __parent, basicLanguage __langA, basicLanguage __langB)
: base("Language Module", "The Targets are distributed into layers by the Passive rules and Active rules testing the tokens of the Target.", __parent)
{
languageA = __langA;
languageB = __langB;
// setup();
}
public layerDistributionRuleBase(string __name, string __description, int __layerID, ISpiderEvaluatorBase __parent, int __layer2ID = 0) : base()
{
parent = __parent;
name = __name;
description = __description;
layerID = __layerID;
layer2ID = __layer2ID;
}
public spiderEvalRuleForLinkBase(string __name, string __description, int __scoreUnit, int __penaltyUnit, ISpiderEvaluatorBase __parent)
{
name = __name;
description = __description;
parent = __parent;
penaltyUnit = __penaltyUnit;
scoreUnit = __scoreUnit;
}
public layerLanguageTFIDF_ARule(basicLanguage __language, int __layerID, ISpiderEvaluatorBase __parent, int __layerID2 = -1)
: base("Language TF-IDF Test ({0})", "Tests Target tokens against the specified language [{0}], sets layerID [{1}] and calculates layer weight score as sum of matched Target token TF-IDFs minus sum of unmatched."
+ "If resulting weight score is more than 0 the layerID [{1}] is assigned, if it's less than 0 then the layer2ID [{2}] is assigned", __layerID, __parent, __layerID2)
{
language = __language;
name = string.Format(name, language.languageEnglishName);
description = string.Format(description, language.languageEnglishName, layerID, layer2ID);
}
public static String GetCrawlFolderName(ISpiderEvaluatorBase spider, crawlerDomainTaskMachineSettings crawlerJobEngineSettings, String templateString)
{
stringTemplate template = new stringTemplate(templateString);
PropertyCollection data = GetData(crawlerJobEngineSettings, spider);
return(template.applyToContent(data));
}
public static String GetCrawlFolderName(ISpiderEvaluatorBase spider, ICrawlJobContext state, String templateString)
{
stringTemplate template = new stringTemplate(templateString);
PropertyCollection data = GetData(state, spider);
return(template.applyToContent(data));
}
public spiderEvalRuleForPageBase(string __name, string __description, int __scoreUnit, int __penaltyUnit, ISpiderEvaluatorBase __parent)
{
name = __name;
description = __description;
parent = __parent;
scoreUnit = __scoreUnit;
penaltyUnit = __penaltyUnit;
mode = spiderEvalRuleResultEnum.active;
}
public rankPageRank(ISpiderEvaluatorBase __parent, double __alpha = 0.85, double __convergence = 0.0001, int __checkSteps = 10)
: base("PageRankRule", "Integrates PageRank algorithm included from C# open source project [https://github.com/jeffersonhwang/pagerank]. Dumping value (d) is set to:" + __alpha.ToString() + ", convergence (c): "
+ __convergence + ", check steps: " + __checkSteps + ". Targets are ranked by associated page rank multiplied to integer factor: {0}. When no pages were loaded it assigns maximum score."
, 1000, 10, __parent)
{
description = string.Format(description, scoreUnit);
alpha = __alpha;
convergence = __convergence;
checkSteps = __checkSteps;
}
public diversityModule(double ttd, double ptd, ISpiderEvaluatorBase __parent, int expansionSteps) :
base("Diversity Module", "Use inversed semantic similarity between a target and Target link url/title corpus and Target page content corpus", __parent)
{
tt_diversityFactor = ttd;
pt_diversityFactor = ptd;
termExpansionSteps = expansionSteps;
// setup();
}
public templateModule(bool isRankingAlternative, ISpiderEvaluatorBase __parent)
: base("Template Module", "The Targets are distributed into layers according to the semantic role of the host content block. ", __parent)
{
alternative = isRankingAlternative;
if (alternative)
{
description = description + "This is alternative implementation with XPath depth based ranking";
}
// setup();
}
public structureModule(bool isRankingAlternative, ISpiderEvaluatorBase __parent)
: base("Structure Module", "The Targets are distributed into layers according to the URL-structure graph tree", __parent)
{
alternative = isRankingAlternative;
if (alternative)
{
description = description + "This is alternative implementation with URL structure graph based ranking";
}
//setup();
}
public void EnterStage(modelSpiderSiteRecord wRecord, ISpiderEvaluatorBase sEvaluator)
{
wRecord.logBuilder.log("-- entering stage [" + name + "] with " + objectives.Count() + " objectives.");
wRecord.logBuilder.log("> " + description + " (codename:" + codename + ")");
wRecord.logBuilder.log("> stage iteration limit: " + stageIterationLimit + " (global limit:" + GLOBAL_stageIterationLimit + ")");
foreach (spiderObjective objective in objectives)
{
objective.prepare();
wRecord.logBuilder.log("> Objective [" + objective.name + "] t:" + objective.supervisor + " ");
}
}
public rankHITS(ISpiderEvaluatorBase __parent, double __convergence = 0.0001, int __checkSteps = 10)
: base("rankTargetsHITS", "Adaptation of HITS algorithm for inner crawl frontier ranking. Convergence (c): "
+ __convergence + ", iteration steps: " + __checkSteps + ". Targets are ranked by associated Authority + Hub score multiplied by integer factor: {0}."
, 1000, 0, __parent)
{
description = string.Format(description, scoreUnit);
convergence = __convergence;
checkSteps = __checkSteps;
hits = new HITSRank();
}
public rankDiversityALink(ISpiderEvaluatorBase __parent, double __target_sd = 0.5, double __page_sd = 0.5, int __expansionSteps = 3)
: base("TokenDiversity", "Ranks links by semantic diversity (inverted similarity) against Target tokens (" + __target_sd.ToString("P") + ") and crawled pages tokens (" + __page_sd.ToString("P") + ")."
+ "Target tokens are semantically expanded in [" + __expansionSteps + "] steps using Semantic Lexicon"
, 100000, 0, __parent)
{
expansionSteps = __expansionSteps;
target_sd = __target_sd;
page_sd = __page_sd;
//subject = spiderEvalRuleSubjectEnum.targets;
mode = spiderEvalRuleResultEnum.active;
doAdjustScoreByLanguageDetection = imbWEMManager.settings.crawlAdHok.FLAG_doAdjustDiversityScore;
__parent.settings.doEnableDLC_TFIDF = true;
}
public static PropertyCollection GetData(ICrawlJobContext state, ISpiderEvaluatorBase crawler)
{
PropertyCollection data = new PropertyCollection();
data[nameComposerFields.crawlerClassName] = crawler.GetType().Name;
data[nameComposerFields.crawlerTitleName] = crawler.name;
data[nameComposerFields.crawlerFileFriendlyName] = crawler.name.getCleanFilepath().Replace("-", "");
data[nameComposerFields.variablePLmax] = crawler.settings.limitTotalPageLoad;
data[nameComposerFields.variableLT] = crawler.settings.limitIterationNewLinks;
//data[nameComposerFields.variableTCmax] = state.crawlerJobEngineSettings.TC_max;
data[nameComposerFields.sampleSize] = state.sampleList.Count();
//data[nameComposerFields.sampleFileSource] = state.sampleFile;
//data[nameComposerFields.sampleName] = state.sampleTags;
return(data);
}
public static void SetTestIDAndSignature(this IReportBenchmark target, ISpiderEvaluatorBase evaluator, ICrawlJobContext state, modelSpiderTestRecord tRecord)
{
target.TestSignature = evaluator.name + "|DC:" + state.sampleList.Count + "|PL:" + evaluator.settings.limitTotalPageLoad + "|LT:" + evaluator.settings.limitIterationNewLinks + "|IID:" + imbWEMManager.index.current_indexID + "|SID:" + imbWEMManager.index.indexSessionEntry.SessionID.add(evaluator.settings.SignatureSufix);
target.Crawler = tRecord.instance.name;
if (evaluator != null)
{
target.TestID = md5.GetMd5Hash(objectSerialization.ObjectToXML(imbWEMManager.settings)) + "-" + evaluator.crawlerHash + "-" + md5.GetMd5Hash(target.TestSignature).toWidthMaximum(3, "");
}
else if (state != null)
{
target.TestID = state.setupHash_global + "-" + state.setupHash_crawler;
}
else
{
target.TestID = md5.GetMd5Hash(target.GetHashCode().ToString());
}
}
public void eventIteration(ISpiderEvaluatorBase __spider, crawlerDomainTask __task, modelSpiderSiteRecord __wRecord)
{
if (!IsEnabled)
{
return;
}
foreach (reportPlugIn_base plug in allPlugins)
{
try
{
plug.eventIteration(__spider, __task, __wRecord);
}
catch (Exception ex)
{
aceLog.log("Reporting Plugin [" + plug.name + "]:" + plug.GetType().Name + " at status report execution crashed: " + ex.Message);
crawlerErrorLog cel = new crawlerErrorLog(ex, null, null, crawlerErrorEnum.indexPlugin);
cel.SaveXML();
}
}
}
public static void Describe(this ISpiderEvaluatorBase evaluator, ITextRender output)
{
output.AppendHeading("Crawler [" + evaluator.name + "]");
output.AppendPair("Class name", evaluator.GetType().Name, true, ": ");
output.AppendPair("Description", evaluator.description, true, ": ");
if (evaluator is spiderModularEvaluatorBase)
{
spiderModularEvaluatorBase evaluator_spiderModularEvaluatorBase = (spiderModularEvaluatorBase)evaluator;
foreach (var md in evaluator_spiderModularEvaluatorBase.modules)
{
md.DescribeModule(output);
}
}
if (evaluator is spiderEvaluatorSimpleBase)
{
spiderEvaluatorSimpleBase evaluator_spiderEvaluatorSimpleBase = (spiderEvaluatorSimpleBase)evaluator;
evaluator_spiderEvaluatorSimpleBase.linkActiveRules.ToList().ForEach(x => x.DescribeRule(output));
}
output.AppendHorizontalLine();
output.open("div", "General configuration", "Crawler configuration properties declared in common settings class");
evaluator.settings.GetUserManual(output, "", true, true);
output.close();
}
public abstract void eventIteration(ISpiderEvaluatorBase __spider, crawlerDomainTask __task, modelSpiderSiteRecord __wRecord);
public layerBlockRolePRule(nodeBlockSemanticRoleEnum __semanticRole, int __layerID, ISpiderEvaluatorBase __parent, int __layer2ID = -1)
: base("Block Role ({0})", "Links found in a block with semantic role [{0}] are set to layerID [{1}], otherwise to [{2}].", __layerID, __parent, __layer2ID)
//: base(__name, __description, __layerID, __parent, __layer2ID)
{
semanticRole = __semanticRole;
name = string.Format(name, semanticRole);
description = string.Format(description, semanticRole, layerID, layer2ID);
__parent.settings.doEnableDLC_BlockTree = true;
}
/// <summary> /// Just when new DLC thread was prepared to run /// </summary> /// <param name="__spider">The spider.</param> /// <param name="__task">The task.</param> /// <param name="__wRecord">The w record.</param> public abstract void eventDLCInitiated(ISpiderEvaluatorBase __spider, crawlerDomainTask __task, modelSpiderSiteRecord __wRecord);
public abstract void eventUniversal <TFirst, TSecond>(crawlerDomainTaskIterationPhase stage, ISpiderEvaluatorBase __parent, TFirst __task, TSecond __resource);
/// <summary> /// Events the plugin installed --- when new spider instance being constructed /// </summary> /// <param name="__spider">The spider.</param> /// <param name="__task">The task.</param> public abstract void eventPluginInstalled(ISpiderEvaluatorBase __spider);
public rankTargetUrlGraph(ISpiderEvaluatorBase __parent, int __scoreUnit = 100)
: base("Target URL Graph", "In the Learning phase creates link-path graph out of all links sent for evaluation, according to normalized score of matched node it will assign proportion of [{0}].", __scoreUnit, 0, __parent)
{
description = string.Format(description, __scoreUnit);
}
protected spiderLayerModuleBase(string __name, string __desc, ISpiderEvaluatorBase __parent) : base(__name, __desc, __parent)
{
}
public spiderRankingModuleBase(string __name, string __desc, ISpiderEvaluatorBase __parent) : base(__name, __desc, __parent)
{
}
public ruleRandomNoiseScore(int maximum, int minimum, ISpiderEvaluatorBase __parent)
: base("NoiseScore", "Passive rule adding random score between {0} and {1} points - used to create initial noise in the frontier. ", maximum, minimum, __parent)
{
description = string.Format(description, maximum, minimum);
rnd = new Random();
}
public void reportIteration(dataUnitSpiderIteration dataUnit, modelSpiderSiteRecord wRecord, ISpiderEvaluatorBase evaluator)
{
iterationPerformanceRecord ip_record = new iterationPerformanceRecord(wRecord);
wRecord.iterationTableRecord.Add(ip_record);
folderNode fn; //siteRecords[wRecord].Add(dataUnit.iteration.ToString("D3"), wRecord.domainInfo.domainRootName + dataUnit.iteration.ToString("D3"), "Iteration " + dataUnit.iteration + " on domain: " + wRecord.domainInfo.domainName);
if (imbWEMManager.settings.directReportEngine.doIterationReport)
{
if (imbWEMManager.settings.directReportEngine.doDomainReport)
{
fn = getIterationFolder(dataUnit.iteration, wRecord);
if (REPORT_WRECORD_LOG)
{
wRecord.logBuilder.getLastLine().saveStringToFile(fn.pathFor("wrecord.txt"));
}
string fileprefix = wRecord.domainInfo.domainRootName.getCleanFilepath();
textByIteration url_loaded = urlsLoaded[wRecord]; //.GetOrAdd(wRecord, new textByIteration());
textByIteration url_detected = urlsDetected[wRecord]; //, new textByIteration());
//textByIteration terms_ext = termsExtracted[wRecord];
//textByIteration sentence_ext = sentencesExtracted[wRecord];
if (REPORT_MODULES)
{
if (imbWEMManager.settings.directReportEngine.DR_ReportModules_XMLIteration)
{
if (wRecord.tRecord.instance is spiderModularEvaluatorBase)
{
wRecord.frontierDLC.reportIterationOut(wRecord, fn);
}
}
}
string its = dataUnit.iteration.ToString("D3");
//DataTable dt = wRecord.context.targets.GetDataTable();
//dt.SetTitle(fileprefix + "_targets");
//dt.serializeDataTable(aceCommonTypes.enums.dataTableExportEnum.csv, "", fn, notation);
//sentence_ext[dataUnit.iteration].AddRangeUnique(wRecord.context.targets.blocks.GetHashList());
//if (REPORT_ITERATION_TERMS)
//{
// fileunit blocks = new fileunit(fn.pathFor(its + "_blc.txt"), false);
// blocks.setContentLines(sentence_ext[dataUnit.iteration]);
// blocks.Save();
//}
if (REPORT_TIMELINE)
{
objectSerialization.saveObjectToXML(ip_record, fn.pathFor("performance.xml"));
}
if (REPORT_ITERATION_URLS)
{
if (wRecord.iteration > 0)
{
builderForMarkdown now_loaded = new builderForMarkdown();
//fileunit now_loaded = new fileunit(fn.pathFor(its + "_loadedNow.txt"), false);
List <spiderTarget> targets_loaded = wRecord.context.targets.GetLoadedInIteration(wRecord.iteration - 1);
int tc = 0;
foreach (spiderTarget t in targets_loaded)
{
reportTarget(t, fn, tc);
now_loaded.AppendLine(t.url);
now_loaded.AppendHorizontalLine();
now_loaded.Append(t.marks.GetActiveResults());
now_loaded.AppendHorizontalLine();
now_loaded.Append(t.marks.GetPassiveResults());
now_loaded.AppendHorizontalLine();
var dt = t.marks.getHistory(t.url, wRecord.tRecord.instance.name);
dt.Save(fn, imbWEMManager.authorNotation, its + "_loadedNow");
now_loaded.AppendTable(dt, false);
tc++;
}
now_loaded.ToString().saveStringToFile(fn.pathFor(its + "_loadedNow.txt"));
spiderTaskResult loadResults = wRecord.spiderTaskResults[wRecord.iteration - 1];
loadResults.getDataTable().GetReportAndSave(fn, notation, "loadResults", true); // .serializeDataTable(aceCommonTypes.enums.dataTableExportEnum.excel, "loadResults", fn, notation);
}
fileunit detected = new fileunit(fn.pathFor(its + "_dt.txt"), false);
fileunit loaded = new fileunit(fn.pathFor(its + "_ld.txt"), false);
fileunit relp = new fileunit(fn.pathFor(its + "_srb_ld.txt"), false);
relp.Append(wRecord.relevantPages, true);
foreach (spiderTarget t in wRecord.context.targets)
{
if (t.page != null)
{
//t.contentBlocks.ForEach(x => sentence_ext[dataUnit.iteration].AddUnique(x.textHash));
loaded.Append(t.url);
url_loaded[dataUnit.iteration].Add(t.url);
}
else
{
detected.Append(t.url);
url_detected[dataUnit.iteration].Add(t.url);
}
}
string lineFormat = "{0,5} {1,30} [s:{1,6}]" + Environment.NewLine;
fileunit active = new fileunit(fn.pathFor(its + "_act.txt"), false);
int c = 1;
foreach (var lnk in wRecord.web.webActiveLinks)
{
active.Append(string.Format(lineFormat, c, lnk.url, lnk.marks.score));
active.Append(lnk.marks.GetLayerAssociation());
c++;
}
detected.Save();
loaded.Save();
active.Save();
}
}
}
wRecord.tRecord.instance.reportIteration(this, wRecord);
}