Solr各单词前缀搜索实现：寻求更简洁高效的查询方案

Hey there! I get it—having to write +ne* +wo* works, but it's not the most elegant syntax for multi-term prefix searches. Let's go through some cleaner, more efficient approaches to get the same result where queries like "ne wo" pull up documents with "New World" in the target field.

1. Use the eDisMax Query Parser with Minimum Match #

The eDisMax parser lets you write more natural query syntax while enforcing that all prefix terms must match. Here's how:

Set the defType parameter to edismax, then use a simple query string with wildcards, and enforce that 100% of the terms must match using the mm (minimum match) parameter:

q=ne* wo*&defType=edismax&mm=100%

• mm=100% ensures every prefix term in your query has a match in the target field, just like the + operators in your original query.
• This syntax is much cleaner—no need for parentheses or explicit + signs, and it reads more like a natural search phrase.

2. Index-Time Prefixes with Edge N-Grams #

If you want to avoid having to add * to the end of each query term entirely, you can pre-generate prefixes at index time using an Edge N-Gram tokenizer. Here's how to set this up:

Step 1: Define a Custom Field Type #

Add this to your schema.xml (or managed-schema):

<fieldType name="text_edge_ngram" class="solr.TextField">
  <analyzer type="index">
    <tokenizer class="solr.StandardTokenizerFactory"/>
    <filter class="solr.LowerCaseFilterFactory"/>
    <filter class="solr.EdgeNGramFilterFactory" minGramSize="2" maxGramSize="15"/>
  </analyzer>
  <analyzer type="query">
    <tokenizer class="solr.StandardTokenizerFactory"/>
    <filter class="solr.LowerCaseFilterFactory"/>
  </analyzer>
</fieldType>

• The EdgeNGramFilterFactory generates all prefixes from 2 to 15 characters long for each token at index time (adjust minGramSize and maxGramSize based on your needs).

Step 2: Add a Copy Field #

Create a field using this type, and copy your target field into it:

<field name="target_prefix" type="text_edge_ngram" indexed="true" stored="false"/>
<copyField source="target" dest="target_prefix"/>

Step 3: Query Without Wildcards #

Now you can query using plain terms, no * needed:

q=target_prefix:ne wo&defType=edismax&mm=100%

• This makes the query even more user-friendly (users don't have to remember to add wildcards).
• Note: Edge N-Grams will increase your index size, so test this with your data volume to ensure performance stays acceptable.

3. Why Your Suggester Attempt Might Have Failed #

If you tried using a Suggester and it didn't work, it's likely because most Suggester implementations (like AnalyzingInfixSuggester) are designed for autocomplete suggestions, not full-document retrieval. They return candidate terms/phrases, not matching documents. If you want to use a Suggester to power prefix search, you'd need to:

• Configure the Suggester to generate multi-term suggestions.
• Take the suggested phrase and pass it to your main query (with wildcards or the Edge N-Gram field).
  But for direct document retrieval, the first two methods are more straightforward.

Final Recommendation #

• If you want a quick, clean syntax change without modifying your schema, go with the eDisMax + mm=100% approach—it's the simplest upgrade from your original query.
• If you want to eliminate wildcards entirely for end-users, the Edge N-Gram index-time approach is worth setting up, as long as you're okay with the slight index size increase.

内容的提问来源于stack exchange，提问作者Michał Osik