/// <summary>
/// Stores a word in the catalog.
/// </summary>
/// <param name="wordText">The word to store.</param>
/// <param name="document">The document the word occurs in.</param>
/// <param name="firstCharIndex">The index of the first character of the word in the document the word occurs at.</param>
/// <param name="wordIndex">The index of the word in the document.</param>
/// <param name="location">The location of the word.</param>
/// <param name="newWord">The new word, or <c>null</c>.</param>
/// <param name="dumpedWord">The dumped word data, or <c>null</c>.</param>
/// <returns>The dumped word mapping data.</returns>
/// <remarks>Storing a word in the index is <b>O(n log n)</b>,
/// where <b>n</b> is the number of words already in the index.</remarks>
protected DumpedWordMapping StoreWord(string wordText, IDocument document, ushort firstCharIndex, ushort wordIndex,
WordLocation location, out Word newWord, out DumpedWord dumpedWord)
{
wordText = wordText.ToLower(CultureInfo.InvariantCulture);
lock (this) {
Word word = null;
if (!catalog.TryGetValue(wordText, out word))
{
// Use ZERO as initial ID, update when IndexStorer has stored the word
// A reference to this newly-created word must be passed outside this method
word = new Word(0, wordText);
catalog.Add(wordText, word);
newWord = word;
dumpedWord = new DumpedWord(word);
}
else
{
newWord = null;
dumpedWord = null;
}
word.AddOccurrence(document, firstCharIndex, wordIndex, location);
return(new DumpedWordMapping(word.ID, document.ID, firstCharIndex, wordIndex, location.Location));
}
}
/// <summary>
/// Determines whether a <see cref="DumpedWord" /> is contained in a list.
/// </summary>
/// <param name="word">The word.</param>
/// <param name="list">The list.</param>
/// <returns><c>true</c> if the word is contained in the list, <c>false</c> orherwise.</returns>
protected static bool Find(DumpedWord word, IEnumerable <DumpedWord> list)
{
foreach (DumpedWord w in list)
{
if (w.ID == word.ID && w.Text == word.Text)
{
return(true);
}
}
return(false);
}
/// <summary>
/// Writes a <see cref="DumpedWord" /> to a <see cref="BinaryWriter" />.
/// </summary>
/// <param name="writer">The <see cref="BinaryWriter" />.</param>
/// <param name="word">The <see cref="DumpedWord" />.</param>
private static void WriteDumpedWord(BinaryWriter writer, DumpedWord word)
{
//if(word.Text.Length == 0) throw new InvalidOperationException();
writer.Write(word.ID);
writer.Write(word.Text);
}
/// <summary>
/// Loads the index from the data store the first time.
/// </summary>
/// <param name="documents">The dumped documents.</param>
/// <param name="words">The dumped words.</param>
/// <param name="mappings">The dumped word mappings.</param>
protected override void LoadIndexInternal(out DumpedDocument[] documents, out DumpedWord[] words, out DumpedWordMapping[] mappings)
{
uint maxDocumentId = 0;
uint maxWordId = 0;
// 1. Load Documents
using(FileStream fs = new FileStream(documentsFile, FileMode.Open, FileAccess.Read, FileShare.None)) {
int count = ReadCount(fs);
BinaryReader reader = new BinaryReader(fs, Encoding.UTF8);
documents = new DumpedDocument[count];
for(int i = 0; i < count; i++) {
documents[i] = ReadDumpedDocument(reader);
if(documents[i].ID > maxDocumentId) maxDocumentId = documents[i].ID;
}
firstFreeDocumentId = maxDocumentId + 1;
}
// 2. Load Words
using(FileStream fs = new FileStream(wordsFile, FileMode.Open, FileAccess.Read, FileShare.None)) {
int count = ReadCount(fs);
BinaryReader reader = new BinaryReader(fs, Encoding.UTF8);
words = new DumpedWord[count];
for(int i = 0; i < count; i++) {
words[i] = ReadDumpedWord(reader);
if(words[i].ID > maxWordId) maxWordId = words[i].ID;
}
firstFreeWordId = maxWordId + 1;
}
// 3. Load Mappings
using(FileStream fs = new FileStream(mappingsFile, FileMode.Open, FileAccess.Read, FileShare.None)) {
int count = ReadCount(fs);
BinaryReader reader = new BinaryReader(fs, Encoding.UTF8);
mappings = new DumpedWordMapping[count];
for(int i = 0; i < count; i++) {
mappings[i] = ReadDumpedWordMapping(reader);
}
}
}
/// <summary>
/// Stores a word in the catalog.
/// </summary>
/// <param name="wordText">The word to store.</param>
/// <param name="document">The document the word occurs in.</param>
/// <param name="firstCharIndex">The index of the first character of the word in the document the word occurs at.</param>
/// <param name="wordIndex">The index of the word in the document.</param>
/// <param name="location">The location of the word.</param>
/// <param name="newWord">The new word, or <c>null</c>.</param>
/// <param name="dumpedWord">The dumped word data, or <c>null</c>.</param>
/// <returns>The dumped word mapping data.</returns>
/// <remarks>Storing a word in the index is <b>O(n log n)</b>,
/// where <b>n</b> is the number of words already in the index.</remarks>
protected DumpedWordMapping StoreWord(string wordText, IDocument document, ushort firstCharIndex, ushort wordIndex,
WordLocation location, out Word newWord, out DumpedWord dumpedWord)
{
wordText = wordText.ToLower(CultureInfo.InvariantCulture);
lock(this) {
Word word = null;
if(!catalog.TryGetValue(wordText, out word)) {
// Use ZERO as initial ID, update when IndexStorer has stored the word
// A reference to this newly-created word must be passed outside this method
word = new Word(0, wordText);
catalog.Add(wordText, word);
newWord = word;
dumpedWord = new DumpedWord(word);
}
else {
newWord = null;
dumpedWord = null;
}
word.AddOccurrence(document, firstCharIndex, wordIndex, location);
return new DumpedWordMapping(word.ID, document.ID, firstCharIndex, wordIndex, location.Location);
}
}
/// <summary>
/// Initializes index data by completely emptying the index catalog and storing the specified data.
/// </summary>
/// <param name="documents">The documents.</param>
/// <param name="words">The words.</param>
/// <param name="mappings">The mappings.</param>
/// <remarks>The method <b>does not</b> check the consistency of the data passed as arguments.</remarks>
/// <exception cref="ArgumentNullException">If <paramref name="documents"/>, <paramref name="words"/> or <paramref name="mappings"/> are <c>null</c>.</exception>
/// <exception cref="InvalidOperationException">If <see cref="M:SetBuildDocumentDelegate"/> was not called.</exception>
public void InitializeData(DumpedDocument[] documents, DumpedWord[] words, DumpedWordMapping[] mappings)
{
if(documents == null) throw new ArgumentNullException("documents");
if(words == null) throw new ArgumentNullException("words");
if(mappings == null) throw new ArgumentNullException("mappings");
if(buildDocument == null) throw new InvalidOperationException("InitializeData can be invoked only when the BuildDocument delegate is set");
lock(this) {
catalog.Clear();
catalog = new Dictionary<string, Word>(words.Length);
// Contains the IDs of documents that are missing
List<uint> missingDocuments = new List<uint>(50);
// 1. Prepare a dictionary with all documents for use in the last step
Dictionary<uint, IDocument> tempDocuments = new Dictionary<uint, IDocument>(documents.Length);
foreach(DumpedDocument doc in documents) {
IDocument builtDoc = buildDocument(doc);
// Null means that the document no longer exists - silently skip it
if(builtDoc != null) {
tempDocuments.Add(doc.ID, builtDoc);
}
else {
missingDocuments.Add(doc.ID);
}
}
// 2. Load words into the catalog, keeping track of them by ID in a dictionary for the next step
Dictionary<ulong, Word> tempWords = new Dictionary<ulong, Word>(words.Length);
// Test for hashing algorithm -- no more used since sequential IDs
//if(words.Length > 0 && words[0].ID != Tools.HashString(words[0].Text)) {
// throw new InvalidOperationException("The search engine index seems to use an outdated hashing algorithm");
//}
foreach(DumpedWord w in words) {
Word word = new Word(w.ID, w.Text);
/*if(tempWords.ContainsKey(w.ID)) {
string t = string.Format("CURRENT: {0}, {1} --- EXISTING: {2}", w.ID, word, tempWords[w.ID]);
Console.WriteLine(t);
}*/
tempWords.Add(w.ID, word);
/*if(catalog.ContainsKey(w.Text)) {
string t = string.Format("CURRENT: {0}, {1} --- EXISTING: {2}", w.ID, word, catalog[w.Text]);
Console.WriteLine(t);
}*/
catalog.Add(w.Text, word);
}
// 3. Add mappings and documents
foreach(DumpedWordMapping map in mappings) {
// HACK: Skip mappings that refer to missing documents and gracefully skip unknown words
if(!missingDocuments.Contains(map.DocumentID)) {
try {
tempWords[map.WordID].AddOccurrence(tempDocuments[map.DocumentID],
map.FirstCharIndex, map.WordIndex, WordLocation.GetInstance(map.Location));
}
catch(KeyNotFoundException) { }
}
}
}
}
/// <summary> /// Loads the index from the data store the first time. /// </summary> /// <param name="documents">The dumped documents.</param> /// <param name="words">The dumped words.</param> /// <param name="mappings">The dumped word mappings.</param> protected abstract void LoadIndexInternal(out DumpedDocument[] documents, out DumpedWord[] words, out DumpedWordMapping[] mappings);
/// <summary>
/// Determines whether a <see cref="DumpedWord" /> is contained in a list.
/// </summary>
/// <param name="word">The word.</param>
/// <param name="list">The list.</param>
/// <returns><c>true</c> if the word is contained in the list, <c>false</c> orherwise.</returns>
protected static bool Find(DumpedWord word, IEnumerable<DumpedWord> list)
{
foreach(DumpedWord w in list) {
if(w.ID == word.ID && w.Text == word.Text) return true;
}
return false;
}
/// <summary>
/// Stores a document in the index.
/// </summary>
/// <param name="document">The document.</param>
/// <param name="keywords">The document keywords, if any, an empty array or <c>null</c> otherwise.</param>
/// <param name="content">The content of the document.</param>
/// <param name="state">A state object that is passed to the IndexStorer SaveDate/DeleteData function.</param>
/// <returns>The number of indexed words (including duplicates) in the document title and content.</returns>
/// <remarks>Indexing the content of the document is <b>O(n)</b>,
/// where <b>n</b> is the total number of words in the document.
/// If the specified document was already in the index, all the old occurrences
/// are deleted from the index.</remarks>
/// <exception cref="ArgumentNullException">If <paramref name="document"/> or <paramref name="content"/> are <c>null</c>.</exception>
public int StoreDocument(IDocument document, string[] keywords, string content, object state)
{
if (document == null)
{
throw new ArgumentNullException("document");
}
if (keywords == null)
{
keywords = new string[0];
}
if (content == null)
{
throw new ArgumentNullException("content");
}
lock (this) {
DumpedChange removeChange = RemoveDocumentInternal(document);
if (removeChange != null)
{
OnIndexChange(document, IndexChangeType.DocumentRemoved, removeChange, state);
}
}
keywords = Tools.CleanupKeywords(keywords);
// When the IndexStorer handles the IndexChanged event and a document is added, the storer generates a new ID and returns it
// via the event handler, then the in-memory index is updated (the document instance is shared across all words) - the final ID
// is generated by the actual IndexStorer implementation (SaveData properly populates the Result field in the args)
List <DumpedWord> dw = new List <DumpedWord>(content.Length / 5);
List <DumpedWordMapping> dm = new List <DumpedWordMapping>(content.Length / 5);
Word tempWord = null;
List <Word> newWords = new List <Word>(50);
DumpedWord tempDumpedWord = null;
int count = 0;
uint sequentialWordId = uint.MaxValue;
// Store content words
WordInfo[] words = document.Tokenize(content);
words = Tools.RemoveStopWords(words, stopWords);
foreach (WordInfo info in words)
{
dm.Add(StoreWord(info.Text, document, info.FirstCharIndex, info.WordIndex, WordLocation.Content, out tempWord, out tempDumpedWord));
if (tempDumpedWord != null && tempWord != null)
{
dm[dm.Count - 1].WordID = sequentialWordId;
tempDumpedWord.ID = sequentialWordId;
dw.Add(tempDumpedWord);
tempWord.ID = sequentialWordId;
newWords.Add(tempWord);
sequentialWordId--;
}
}
count += words.Length;
// Store title words
words = document.Tokenize(document.Title);
words = Tools.RemoveStopWords(words, stopWords);
foreach (WordInfo info in words)
{
dm.Add(StoreWord(info.Text, document, info.FirstCharIndex, info.WordIndex, WordLocation.Title, out tempWord, out tempDumpedWord));
if (tempDumpedWord != null && tempWord != null)
{
dm[dm.Count - 1].WordID = sequentialWordId;
tempDumpedWord.ID = sequentialWordId;
dw.Add(tempDumpedWord);
tempWord.ID = sequentialWordId;
newWords.Add(tempWord);
sequentialWordId--;
}
}
count += words.Length;
ushort tempCount = 0;
// Store keywords
for (ushort i = 0; i < (ushort)keywords.Length; i++)
{
dm.Add(StoreWord(keywords[i], document, tempCount, i, WordLocation.Keywords, out tempWord, out tempDumpedWord));
if (tempDumpedWord != null && tempWord != null)
{
dm[dm.Count - 1].WordID = sequentialWordId;
tempDumpedWord.ID = sequentialWordId;
dw.Add(tempDumpedWord);
tempWord.ID = sequentialWordId;
newWords.Add(tempWord);
sequentialWordId--;
}
tempCount += (ushort)(1 + keywords[i].Length);
}
count += keywords.Length;
IndexStorerResult result = OnIndexChange(document, IndexChangeType.DocumentAdded,
new DumpedChange(new DumpedDocument(document), dw, dm), state);
// Update document ID
if (result != null && result.DocumentID.HasValue)
{
document.ID = result.DocumentID.Value;
}
else
{
// HACK: result is null -> index is corrupted, silently return
return(0);
}
// Update word IDs in newWords
bool wordIdUpdated = false;
foreach (Word word in newWords)
{
wordIdUpdated = false;
foreach (WordId id in result.WordIDs)
{
if (id.Text == word.Text)
{
word.ID = id.ID;
wordIdUpdated = true;
break;
}
}
if (!wordIdUpdated)
{
throw new InvalidOperationException("No ID for new word");
}
}
return(count);
}