/// <summary>
/// Gets WikiLinks from a section
/// </summary>
/// <param name="lines"></param>
/// <param name="source"></param>
/// <returns>List of links for section</returns>
public static List <WikiLink> GetWikiLinks(List <string> lines, Section source)
{
List <WikiLink> wikiLinks = new List <WikiLink>();
Regex wikiLinkRE = WikiMarkup.GetWikiLinkRE();
Regex subsectionRE = WikiMarkup.GetSubsectionRE();
Regex emphasisRE = WikiMarkup.GetEmphasisRE();
foreach (string line in lines)
{
MatchCollection mc = wikiLinkRE.Matches(line);
foreach (Match linkMatch in mc)
{
// create wiki link
string rawLink = linkMatch.Value;
string page = linkMatch.Groups["LinkDest"].ToString().Trim();
string section = linkMatch.Groups["Section"].ToString().Trim();
string displayText = linkMatch.Groups["DisplayText"].ToString().Trim();
WikiLink link = new WikiLink(rawLink, page, section, displayText, source);
wikiLinks.Add(link);
}
}
wikiLinks.Sort();
return(wikiLinks);
}
/// <summary>
/// Finds the end of a section.
/// </summary>
/// <param name="lines">Lines to search</param>
/// <param name="start">Where to start search. This must be a valid section line (e.g., ==Section X==)</param>
/// <returns>Number of last line in the section</returns>
public static int GetSectionEnd(List <string> lines, int start)
{
Regex r = WikiMarkup.GetSubsectionRE();
// get current depth
Match m = r.Match(lines[start]);
if (!m.Success)
{
throw new Exception("Invalid starting line for section");
}
int depth = m.Groups["SectionStart"].Length;
// find end of section
int subsectionEnd;
for (subsectionEnd = start + 1; subsectionEnd < lines.Count; ++subsectionEnd)
{
string line = lines[subsectionEnd];
m = r.Match(line);
if (m.Success)
{
int depthSub = m.Groups["SectionStart"].Length;
if (depthSub <= depth)
{
break;
}
}
}
return(subsectionEnd - 1);
}
/// <summary>
/// Removes markup from this page
/// </summary>
public void RemoveMarkup()
{
WikiMarkup.ProcessMarkup(_lines);
foreach (Section s in _sections)
{
s.RemoveMarkup();
}
}
/// <summary>
/// Writes the titleindex table
/// </summary>
/// <param name="ns">Namespace to read/write from/to</param>
/// <param name="titles">List of titles to write</param>
/// <returns>Last title written</returns>
public string WriteTitleIndex(Namespace ns, List <string> titles)
{
int nsVal = NamespaceValue(ns);
// process titles
for (int j = 0; j < titles.Count; ++j)
{
string title = titles[j];
if (title.Trim() == "")
{
continue;
}
string[] words = title.Split(new char[] { '_', '\\', '/' }, StringSplitOptions.RemoveEmptyEntries);
// process words in title
for (int k = 0; k < words.Length; ++k)
{
string word = words[k];
word = word.Trim().ToLower();
word = WikiMarkup.TrimPunctuation(word);
if (WikiMarkup.IsStopWord(word, true))
{
continue;
}
word = MySQLEscape(word);
title = MySQLEscape(title);
try
{
IDataReader reader = SubmitQuery("SELECT word FROM titleindex " +
"WHERE word=\"" + word + "\" AND page_namespace=" + nsVal + " AND page_title=\"" + title + "\"");
if (!reader.Read())
{
reader.Close();
ExecuteNonQuery("INSERT INTO titleindex (word, page_namespace, page_title) " +
"VALUES (\"" + word + "\", " + nsVal + ", \"" + title + "\")");
}
reader.Close();
_lastTitleWritten = title;
}
catch (Exception ex)
{
throw new Exception("Failed at title " + title + ", error: " + ex);
}
}
}
return(_lastTitleWritten);
}
/// <summary>
/// Constructor
/// </summary>
/// <param name="rawLink">Raw text for link</param>
/// <param name="pageTitle">Title of page being linked to</param>
/// <param name="pageSection">Section of page being linked to</param>
/// <param name="displayText">Display text for link</param>
/// <param name="sourceSection">Section of source page that contains this link</param>
public WikiLink(string rawLink, string pageTitle, string pageSection, string displayText, Section sourceSection)
{
_rawLink = rawLink;
_destPageTitle = pageTitle;
_destPageSection = pageSection;
_displayText = displayText != "" ? displayText : (pageTitle + (_destPageSection != "" ? "#" + _destPageSection : ""));
_displayText = WikiMarkup.ProcessMarkup(_displayText);
_destPageNamespace = WikiDB.Namespace.Main;
_sourcePage = sourceSection.ContainingPage;
_sourceSection = sourceSection;
_weight = 0;
}
/// <summary>
/// Gets the line number within a list of lines of the next section start
/// </summary>
/// <param name="lines">List of lines to search</param>
/// <param name="start">Where to start search</param>
/// <returns>Line number of section start</returns>
public static int GetNextSectionStart(List <string> lines, int start)
{
Regex sectionRE = WikiMarkup.GetSubsectionRE();
int sectionStart = start;
while (sectionStart < lines.Count &&
!sectionRE.Match(lines[sectionStart]).Success)
{
++sectionStart;
}
return(sectionStart);
}
/// <summary>
/// Constructor for marked up section
/// </summary>
/// <param name="name">Name of section</param>
/// <param name="wikiTextLines">List of lines for the page</param>
/// <param name="parentSection">Parent section</param>
/// <param name="containingPage">Page that contains this section</param>
public Section(string name, List <string> wikiTextLines, Section parentSection, Page containingPage)
: this()
{
if (wikiTextLines == null || wikiTextLines.Count == 0)
{
throw new Exception("Cannot create Section with blank text");
}
_lines = wikiTextLines;
_parentSection = parentSection;
_name = name;
_containingPage = containingPage;
// include the first line if it's not a section line (the case for headers)
Regex sectionRE = WikiMarkup.GetSubsectionRE();
int secLinesStart;
if (sectionRE.Match(_lines[0]).Success)
{
secLinesStart = 1;
}
else
{
secLinesStart = 0;
}
List <string> subsectionLines = ExtractLines(_lines, secLinesStart, _lines.Count - secLinesStart);
List <Section> subsections = ExtractSections(subsectionLines, this, _containingPage);
_subSections.AddRange(subsections);
// get links prior to first subsection
int subStart = GetNextSectionStart(_lines, secLinesStart);
List <string> preSubsectionLines = ExtractLines(_lines, 0, subStart - 1);
_wikiLinks = WikiLink.GetWikiLinks(preSubsectionLines, this);
// remove markup
WikiMarkup.ProcessMarkup(_lines);
}
/// <summary>
/// Extract a section and its subsections (and their subsections, etc.) from a list of lines
/// </summary>
/// <param name="lines">Lines to extract from</param>
/// <param name="parent">Parent section for top-level sections</param>
/// <param name="containingPage">Page that contains these sections</param>
/// <returns>List of Section instances</returns>
public static List <Section> ExtractSections(List <string> lines, Section parent, Page containingPage)
{
List <Section> sections = new List <Section>();
if (lines == null || lines.Count == 0)
{
return(sections);
}
// get sections
Regex subsectionRE = WikiMarkup.GetSubsectionRE();
int startLine = GetNextSectionStart(lines, 0);
while (startLine < lines.Count)
{
// check starting line
Match m = subsectionRE.Match(lines[startLine]);
if (!m.Success)
{
throw new Exception("Invalid starting line for section");
}
// get the end of this section
int endLine = GetSectionEnd(lines, startLine);
// get lines for this section
List <string> sectionLines = ExtractLines(lines, startLine, endLine);
// create section, which recursively extracts subsections
string secName = m.Groups["Name"].ToString();
secName = WikiMarkup.ProcessMarkup(secName);
Section s = new Section(secName, sectionLines, parent, containingPage);
sections.Add(s);
// get start of next section
startLine = GetNextSectionStart(lines, endLine + 1);
}
return(sections);
}
/// <summary>
/// Constructor
/// </summary>
/// <param name="title">Title of page</param>
/// <param name="ns">Namespace of this page</param>
/// <param name="ID">ID for this page</param>
/// <param name="wikiText">Wiki text for page</param>
/// <param name="database">Database containing this page</param>
/// <param name="followRedirection">Whether or not to follow redirection</param>
public Page(string title, WikiDB.Namespace ns, int ID, string wikiText, WikiDB database, bool followRedirection)
: this(title, ns, ID)
{
// remove irrelevant markup
wikiText = WikiMarkup.RemoveIrrelevantMarkup(wikiText);
// split page up into lines
_lines = Section.GetLines(wikiText);
int firstSectionStart = Section.GetNextSectionStart(_lines, 0);
List <string> headerLines = Section.ExtractLines(_lines, 0, firstSectionStart - 1);
if (headerLines.Count > 0)
{
Header h = new Header(headerLines, this);
_sections.Add(h);
}
// get sections
_sections.AddRange(Section.ExtractSections(_lines, null, this));
// check for redirect page
string firstLine = "";
if (_lines.Count > 0)
{
firstLine = _lines[0];
}
string redirect = "#redirect";
if (firstLine.Length >= redirect.Length &&
firstLine.Substring(0, redirect.Length).ToLower() == redirect &&
WikiLinks.Count == 1 &&
followRedirection)
{
// get redirect page
string redirectURL = WikiLinks[0].DestPageURL;
_redirectsTo = database.LookupPage(ns, redirectURL, followRedirection);
}
// process markup
WikiMarkup.ProcessMarkup(_lines);
// set line information for the page
SetLineInfo();
// get TF information
foreach (Section s in _sections)
{
foreach (string line in s.Lines)
{
string[] tokens = line.Split(' ');
foreach (string token in tokens)
{
// ignore case
string lowerToken = token.ToLower().Trim();
lowerToken = WikiMarkup.TrimPunctuation(lowerToken);
if (lowerToken == "" ||
WikiMarkup.IsStopWord(lowerToken, false))
{
continue;
}
if (!_termFrequencies.ContainsKey(lowerToken))
{
_termFrequencies[lowerToken] = 1.0F;
}
else
{
_termFrequencies[lowerToken] = _termFrequencies[lowerToken] + 1;
}
}
}
}
}
