问题
In our own RavenQueryableExtensions
class we have the following method:
public static IRavenQueryable<T> SearchMultiple<T>(this IRavenQueryable<T> self,
Expression<Func<T, object>> fieldSelector, string queries,
decimal boost = 1, SearchOptions options = SearchOptions.Or)
{
if (String.IsNullOrEmpty(queries)) throw new ArgumentNullException("queries");
// More than two spaces or tabs are replaced with a single space
var newQueries = Regex.Replace(queries, @"\s{2,}", " ");
// not important for this question:
//newQueries = SyncShared.ReplacePostcode(newQueries);
// Splits the search-string into separate search-terms
var searchValues = newQueries.Split(' ');
return self.SearchMultiple(fieldSelector, searchValues, boost, options);
}
public static IRavenQueryable<T> SearchMultiple<T>(this IRavenQueryable<T> self,
Expression<Func<T, object>> fieldSelector, IEnumerable<string> queries,
decimal boost = 1, SearchOptions options = SearchOptions.Or)
{
if (queries == null) throw new ArgumentNullException("queries");
return queries.Aggregate(self, (current, query) => current.Search(fieldSelector, query + "* ", boost, options, EscapeQueryOptions.AllowPostfixWildcard));
}
Which creates a search-query with all the loose search-terms in the searchValues-array. However, it seems to not recognize the special characters like &
or .
. For example:
- We have a list with Companies. One of the companies has the name "A & A something more".
- When I enter the search "A & A" or just "&" it doesn't find this company.
- It generated the following Query when I Debug:
{Query:(A*) AND Query:(\&*) AND Query:(A*)}
- The same goes when I enter any other special character like "." or "`".
Does anyone know how to alter the Search-method so it correctly formats these special characters?
Also, I don't know if it's relevant for our issue, but we also use a AsciiFoldingAnalyzer
class (see below). This class allows us to also search companies with characters like "é" or "ü" when we just enter "e" or "u".
using System.Collections.Generic;
using System.IO;
using Lucene.Net.Analysis;
using Lucene.Net.Analysis.Standard;
using Lucene.Net.Util;
namespace NatWa.MidOffice.RavenDb.ServerGoodies
{
public class AsciiFoldingAnalyser : StandardAnalyzer
{
public AsciiFoldingAnalyser(Version matchVersion)
: base(matchVersion)
{
}
public AsciiFoldingAnalyser(Version matchVersion, ISet<string> stopWords)
: base(matchVersion, stopWords)
{
}
public AsciiFoldingAnalyser(Version matchVersion, FileInfo stopwords)
: base(matchVersion, stopwords)
{
}
public AsciiFoldingAnalyser(Version matchVersion, TextReader stopwords)
: base(matchVersion, stopwords)
{
}
public override TokenStream TokenStream(string fieldName, TextReader reader)
{
return new LowerCaseFilter(new ASCIIFoldingFilter(base.TokenStream(fieldName, reader)));
}
}
}
And we use it in our Mappings like so:
public class UserLijst : AbstractIndexCreationTask<UserState, UserLijstResult>
{
public UserLijst()
{
Map = states => from state in states
select new UserLijstResult
{
Id = (UserId)state.AggregateId,
Naam = state.Naam,
Query = new object[]
{
state.Naam
}
};
Reduce = results => from result in results
group result by new { result.Id } into g
select new UserLijstResult
{
Id = g.Key.Id,
Naam = g.First().Naam,
Query = g.First().Query
};
Index("Query", FieldIndexing.Analyzed);
Analyze(result => result.Query, typeof(AsciiFoldingAnalyser).AssemblyQualifiedName);
}
}
回答1:
Ok, it turned out it was pretty easy. We were using a base Tokenizer in our Analyser, which filters out all special characters and characters with a length of 1. When we replaced
public override TokenStream TokenStream(string fieldName, TextReader reader)
{
return new LowerCaseFilter(new ASCIIFoldingFilter(base.TokenStream(fieldName, reader)));
}
in our AsciiFoldingAnalyser for:
public override TokenStream TokenStream(string fieldName, TextReader reader)
{
return new LowerCaseFilter(new ASCIIFoldingFilter(new WhitespaceTokenizer(reader)));
}
it works. We can search for the special characters.
We now get A LOT of results for a search like "A & A", since it finds all occurrences of the characters "a" and "&" in all indexed fields, so perhaps we need to alter a few more things to narrow this down a bit, but at least I got what I want when asking this question.
来源:https://stackoverflow.com/questions/28338943/iravenqueryablet-search-cant-find-sign-or-other-special-characters