Files
raven/server/AyaNova/biz/Search.cs
2018-09-19 14:59:15 +00:00

458 lines
19 KiB
C#

using System;
using System.Globalization;
using System.Text;
using System.Collections.Generic;
using System.IO;
using Newtonsoft.Json.Linq;
using Microsoft.Extensions.Logging;
using Microsoft.EntityFrameworkCore;
using AyaNova.Util;
using AyaNova.Models;
namespace AyaNova.Biz
{
//This class handles word breaking, processing keywords and searching for results
public static class Search
{
//Initial keyword indexing consists of
//WordBreaker - break down into words
//ProcessKeywords into database
#region ProcessKeywords into Database
public static void ProcessNewObjectKeywords(AyContext ct, long localeId, long objectID, AyaType objectType, string name, params string[] text)
{
ProcessKeywords(ct, localeId, objectID, objectType, true, name, text);
}
public static void ProcessUpdatedObjectKeywords(AyContext ct, long localeId, long objectID, AyaType objectType, string name, params string[] text)
{
ProcessKeywords(ct, localeId, objectID, objectType, false, name, text);
}
public static void ProcessDeletedObjectKeywords(AyContext ct, long objectID, AyaType objectType)
{
//Be careful in future, if you put ToString at the end of each object in the string interpolation
//npgsql driver will assume it's a string and put quotes around it triggering an error that a string can't be compared to an int
ct.Database.ExecuteSqlCommand($"delete from asearchkey where objectid={objectID} and objecttype={(int)objectType}");
}
/// <summary>
/// Process the keywords into the dictionary
/// </summary>
private static void ProcessKeywords(AyContext ct, long localeId, long objectID, AyaType objectType, bool newRecord, string name, params string[] text)
{
//TODO: Code this, using method idea from v7 code and adding the handling of the new InName flag and Name separately.
//Note that as initially coded in widget test class the context will be saved by the controller as it is also done with the event log
//So theoretically I don't save here, but it may turn out that the code requires a save so in that case need to re-do the WidgetController calls to here to
//account for the save
//IF NOT NEW, DELETE ALL EXISTING ENTRIES FOR OBJECT TYPE AND ID
if (!newRecord)
{
ProcessDeletedObjectKeywords(ct, objectID, objectType);
}
//BREAK STRING ARRAY INTO KEYWORD LIST
List<string> KeyWordList = Break(localeId, text);
//EARLY EXIT IF NO KEYWORDS OR NAME RECORD TO PROCESS
if (KeyWordList.Count == 0 && string.IsNullOrWhiteSpace(name))
{
return;
}
//ITERATE ALL THE KEYWORDS, SEARCH IN THE SEARCHDICTIONARY TABLE AND COLLECT ID'S OF ANY PRE-EXISTING IN DB KEYWORDS
//ITERATE THROUGH THE KEYWORDS THAT DO *NOT* HAVE MATCHES IN THE SEARCHDICTIONARY AND ADD THEM TO THE SEARCH DICTIONARY, COLLECTING THEIR ID'S
//CREATE THE SEARCHKEY RECORDS FOR ALL THE KEYWORDS
}
#endregion
#region Breaker
//Class to hold relevant locale data for breaking text
public class LocaleWordBreakingData
{
public bool CJKIndex { get; set; }
public List<string> StopWords { get; set; }
public LocaleWordBreakingData()
{
CJKIndex = false;
StopWords = new List<string>();
}
}
//Get the current stopwords for the user's locale
private static LocaleWordBreakingData GetLocaleSearchData(long localeId, AyContext ct = null)
{
LocaleWordBreakingData LSD = new LocaleWordBreakingData();
if (ct == null)
ct = ServiceProviderProvider.DBContext;
//Get stopwords
//Validate locale id, if not right then use default instead
var Param = new Api.Controllers.LocaleController.LocaleSubsetParam();
Param.LocaleId = LocaleBiz.EnsuredLocaleIdStatic(localeId, ct);
Param.Keys.Add("StopWords1");
Param.Keys.Add("StopWords2");
Param.Keys.Add("StopWords3");
Param.Keys.Add("StopWords4");
Param.Keys.Add("StopWords5");
Param.Keys.Add("StopWords6");
Param.Keys.Add("StopWords7");
var Stops = LocaleBiz.GetSubsetStatic(Param).Result;
foreach (KeyValuePair<string, string> kvp in Stops)
{
//Each stopwords locale key is a space delimited list of words and in the case of an empty local string (i.e. StopWords7) it's value is a single question mark
if (kvp.Value != "?")
{
LSD.StopWords.AddRange(kvp.Value.Split(" "));
}
}
LSD.CJKIndex = LocaleBiz.GetCJKIndex(localeId, ct).Result;
return LSD;
}
public enum TokenTypes
{ Nothing, Separator, CJK, Latin };
/// <summary>
/// Take an array of strings and
/// return a single string
/// containing unique only, lowercase comma delimited
/// keywords suitable for passing to a
/// stored procedure or other function
///
/// Use Locale setting CJKIndex=true to handle Chinese, Japanese, Korean etc
/// (languages with no easily identifiable word boundaries as in english)
/// </summary>
///
/// <param name="localeId"></param>
/// <param name="text">An array of 0 to * strings of text</param>
/// <returns>List of strings</returns>
internal static List<string> Break(long localeId, params string[] text)
{
List<string> KeyWordList = new List<string>(BreakCore(localeId, false, text).Split(','));
return KeyWordList;
}
/// <summary>
/// Used to Process users search phrase and preserve wild
/// cards entered
/// </summary>
/// <param name="localeId"></param>
/// <param name="text"></param>
/// <returns></returns>
internal static string BreakSearchPhrase(long localeId, params string[] text)
{
return BreakCore(localeId, true, text);
}
/// <summary>
/// Stop words list reset upon login or editing of localized text
/// used for eliminating noise words from search dictionary
/// </summary>
public static System.Collections.Generic.List<string> StopList = null;
internal static string BreakCore(long localeId, bool KeepWildCards, params string[] text)
{
//Get stopwords and CJKIndex flag value
LocaleWordBreakingData LSD = GetLocaleSearchData(localeId);
int MAXWORDLENGTH = 255;
StringBuilder sbResults = new StringBuilder();
//List to temporarily hold parsed words
//used to easily ensure unique words only
List<string> tempParsedWords = new List<string>();
StringBuilder sb = new StringBuilder();
StringBuilder sbWord = new StringBuilder();
// System.IO.StringWriter sr = new System.IO.StringWriter(sb);
// System.Xml.XmlTextWriter w = new System.Xml.XmlTextWriter(sr);
// w.Formatting = System.Xml.Formatting.Indented;
// w.WriteStartElement("Items");
//Loop through each of the passed in strings
foreach (string s in text)
{
if (s == null || s == "") continue;
//get all the characters in a unicode compliant manner...
TextElementEnumerator t = StringInfo.GetTextElementEnumerator(s);
//start at the top
t.Reset();
TokenTypes LastToken = TokenTypes.Nothing;
//Used by CJK
bool BasicLatinBlock = true;
//Process each "character" (text element,glyph whatever) in the
//current string
while (t.MoveNext())
{
//get it as a character
char c = t.GetTextElement()[0];
if (!LSD.CJKIndex)
{
#region regular tokenizer
//Is it a token we want to include?
//Or a wildcard character
if (char.IsLetterOrDigit(c) || (KeepWildCards && c == '%'))
{
#region Include token
//All latin text is converted to lower case
c = char.ToLower(c);
//Do we already have a word?
if (sbWord.Length > 0)
{
//Maybe we need to flush this word into the word list
//if we're over the word length limit
if (sbWord.Length >= MAXWORDLENGTH)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
sbWord.Append(c);
LastToken = TokenTypes.Latin;
continue;
}
}
//append character and go on to next one
sbWord.Append(c);
LastToken = TokenTypes.Latin;
continue;
#endregion
}
else
{
#region Word Boundary token
LastToken = TokenTypes.Separator;
if (sbWord.Length > 0)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
continue;
}
#endregion
}
#endregion
}
else
{
#region CJK Tokenizer
//Is it a basic latin charater? (ascii basically)
//see: http://www.unicode.org/charts/index.html
//and here for a funky online viewer:
//http://www.fileformat.info/info/unicode/block/index.htm
//we need to know this so that regular english text
//within cjk text gets properly indexed as whole words
BasicLatinBlock = false;
if ((int)c < 256) BasicLatinBlock = true;
if (BasicLatinBlock)
{
//Is it a token we want to include?
if (char.IsLetterOrDigit(c) || (KeepWildCards && c == '%'))
{
#region Latin Include token
//All latin text is converted to lower case
c = char.ToLower(c);
//Do we already have a word?
if (sbWord.Length > 0)
{
//Maybe we need to flush this word into the word list
//if we're over the word length limit or we are going from
//CJK to latin
if (LastToken == TokenTypes.CJK || sbWord.Length >= MAXWORDLENGTH)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
sbWord.Append(c);
LastToken = TokenTypes.Latin;
continue;
}
}
//append character and go on to next one
sbWord.Append(c);
LastToken = TokenTypes.Latin;
continue;
#endregion
}
else
{
#region Latin Word Boundary token
LastToken = TokenTypes.Separator;
if (sbWord.Length > 0)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
continue;
}
#endregion
}
}
else//CJK character
{
if (char.IsLetter(c) || (KeepWildCards && c == '%'))
{
#region CJK Include token
//Do we already have a word?
if (sbWord.Length > 0)
{
//Maybe we need to flush this word into the word list
//if we're over the word length limit or we are going from
//latin TO CJK
if (LastToken == TokenTypes.Latin || sbWord.Length >= MAXWORDLENGTH)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
sbWord.Append(c);
LastToken = TokenTypes.CJK;
continue;
}
if (LastToken == TokenTypes.CJK)
{
//we're here because there is more than zero characters already stored
//and the last was CJK so we need append current character
//and flush the resultant 2 character n-gram
sbWord.Append(c);
System.Diagnostics.Debug.Assert(sbWord.Length == 2);
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
sbWord.Append(c);
LastToken = TokenTypes.CJK;
continue;
}
}
//append character and go on to next one
sbWord.Append(c);
LastToken = TokenTypes.CJK;
continue;
#endregion
}
else
{
#region CJK Word Boundary token
LastToken = TokenTypes.Separator;
if (sbWord.Length > 0)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
continue;
}
#endregion
}
}
#endregion
}
}
//Flush out the last word
if (sbWord.Length > 0)
{
//flush away...
if (!tempParsedWords.Contains(sbWord.ToString()))
{
tempParsedWords.Add(sbWord.ToString());
}
sbWord.Length = 0;
}
}
//bail early if there is nothing indexed
if (tempParsedWords.Count == 0) return "";
//Make a return string array
//from the word list
foreach (string s in tempParsedWords)
{
//Add only non stopwords
if (!StopList.Contains(s))
{
sbResults.Append(s);
sbResults.Append(",");
}
}
//sometimes all the results are stop words so you end up
//here with nothing in sbResults.
return sbResults.ToString().TrimEnd(',');
}
#endregion
}//eoc
}//eons