Lucene技术问询：如何让首次检索返回并评分全部文档？

Great question—this is a common gotcha when working with Lucene's query execution model. Let's break this down into two parts: how to force Lucene to score all documents for your target query, and how Lucene decides which documents get scored in the first place.

How to Force Lucene to Score All Documents for a Query #

If you want Lucene to evaluate and score every document in your index against your query (even if none match the original query criteria), here are a few reliable approaches:

1. Combine Your Query with MatchAllDocsQuery #

The simplest way is to wrap your target query in a BooleanQuery alongside a MatchAllDocsQuery, using SHOULD clauses. This ensures every document is included in the result set, and your original query's scoring logic is applied where applicable:

// Your original query that's returning no scored documents
Query targetQuery = new TermQuery(new Term("content", "your-query-term"));

// Query that matches every document in the index
Query matchAllQuery = new MatchAllDocsQuery();

// Combine them: all documents are included, and targetQuery matches get boosted
BooleanQuery combinedQuery = new BooleanQuery.Builder()
    .add(targetQuery, BooleanClause.Occur.SHOULD)
    .add(matchAllQuery, BooleanClause.Occur.SHOULD)
    .build();

Documents that match targetQuery will get their full query score plus the base score from MatchAllDocsQuery (usually 1.0), while non-matching documents will just get the base score.

2. Adjust Minimum Match Threshold (For Boolean Queries) #

If your original query is a BooleanQuery with SHOULD clauses, check if you've set a minimumNumberShouldMatch value that's too high. Setting this to 0 ensures even documents that match zero of your SHOULD clauses are included and scored:

BooleanQuery.Builder queryBuilder = new BooleanQuery.Builder();
queryBuilder.add(new TermQuery(new Term("content", "your-query-term")), BooleanClause.Occur.SHOULD);
// Allow documents with zero matching clauses to be included
queryBuilder.setMinimumNumberShouldMatch(0);
Query query = queryBuilder.build();

Non-matching documents will receive a score of 0, but they'll still be part of the result set and processed by the scoring pipeline.

3. Custom Query Rewriting #

For more advanced scenarios, you can implement a custom QueryRewriter that automatically wraps any query in a combination with MatchAllDocsQuery. This is useful if you want this behavior to apply globally across your application.

How Lucene Selects Documents to Score #

Lucene's query execution pipeline follows these key steps to determine which documents get scored:

• Query Rewriting: First, your query is rewritten via Query.rewrite(IndexReader) to a more efficient execution form. For example, a complex BooleanQuery might be simplified or optimized here.
• Weight & Scorer Creation: The rewritten query creates a Weight object, which generates a Scorer via Weight.scorer(LeafReaderContext). The Scorer is responsible for iterating over documents that match the query criteria and calculating their scores.
• Matching Documents: By default, the Scorer only iterates over documents that match the query's conditions. If no documents match, the Scorer returns null, and no scoring occurs—this is exactly what you're seeing.

Key classes to reference in Lucene's documentation:

• Query: The base class for all queries; its rewrite method dictates how the query is transformed for execution.
• Weight: Bridges the query and index, responsible for scoring logic and creating the Scorer.
• Scorer: Iterates over matching documents and computes scores for each.
• MatchAllDocsQuery: A special query whose Scorer iterates over every document in the index, forcing full index scoring.

内容的提问来源于stack exchange，提问作者robcarney