I sometimes find it helpful to think of the query parts as applying 'filtering' logic, helping to understand how query components play together in determining the acceptable set of results (mostly ignoring scoring here, which would usually sort the candidate results).
Consider a set of 10 douments with a single field having the following simplistic content: doc 0 content: a z 0 doc 1 content: a z 1 doc 2 content: a z 2 doc 3 content: a z 3 doc 4 content: a z 4 doc 5 content: b z 5 doc 6 content: b z 6 doc 7 content: b z 7 doc 8 content: b z 8 doc 9 content: b z 9 And consider the following queries on that single field: query = 3 Only 3 is a valid result. query = +3 Only docs containing 3 are valid candidates (docs not containing 3 are not valid candidates). So, again, only 3 is a valid result. query = 2 3 Only docs containing any of 2 3 are valid candidates. So both 2 and 3 are valid results. query = 2 +3 Only docs containing 3 are valid candidates (docs not containing 3 are not valid). Docs that contain 2, are valid candidates (but only if they also contain 3). So only 3 is a valid result. query = 2 6 b Only docs containing at least one of 2 6 b are valid candidates. So valid results are: 2 5 6 7 8 9 query = 2 6 +b Only docs containing b are valid candidates (docs not containing b are not valid). Docs containing 2, 6, or both are valid candidates (but ony if they also contain b). So valid results are: 5 6 7 8 9 As you can now see, for any query with <should> as well as <must/not> parts: query = <should-part> +<must-part> query = <should-part> -<must-not-part> The <should part> only affects the scoring of the valid documents. It does not define the set of valid documents. In other words, documents that do not meet the <should-part> can find their way to the query result set. On the other hand, documents that do not meet the <must-part> or the <must-not-part> can never find their way to the result set. To complete the picture, maybe worth noting is that (1) the <must-part> also contributes to the score of result documents; (2) filters can be used for having a <must-logic> without affecting scoring; and (3) <must-not-part> does not affect the scoring of result documents (obviously). The examples in your test case would be: query = 3 4 5 6 query = (3 4 5 6) Docs containing any of 3 4 5 6 are valid. So valid results are 3 4 5 6. query = +(3 4 5 6) Docs containing any of 3 4 5 6 are valid (and only these docs are valid). So again valid results are 3 4 5 6. query = (3 4 5 6) +x Docs containing any of 3 4 5 6 are valid. Only docs containing x are valid candidates (docs not containing x are not valid). So valid results are all docs: 0 1 2 3 4 5 6 7 8 9 query = +(3 4 5 6) +x Only Docs containing any of 3 4 5 6 are valid (docs not containing any of 3 4 5 6 are not valid). Only docs containing x are valid candidates (docs not containing x are not valid). So valid results are: 3 4 5 6 query = +(3 4 5 6) +b Only Docs containing any of 3 4 5 6 are valid (docs not containing any of 3 4 5 6 are not valid). Only docs containing b are valid candidates (docs not containing b are not valid). So valid results are: 5 6 - Doron Erik Hatcher <[EMAIL PROTECTED]> wrote on 05/10/2006 02:58:02: > > On Oct 5, 2006, at 4:59 AM, Tom Hill wrote: > > It's clear that my problem here comes from a lack of understanding > > of the semantics of SHOULD, MUST, and MUST_NOT. > > > > I haven't found a clear description of this (except for a brief > > comment here http://mail-archives.apache.org/mod_mbox/lucene-java- > > dev/200408.mbox/[EMAIL PROTECTED]). Most of the > > other descriptions I have found discuss it as if it they were > > boolean operators, which they aren't, quite. > > > > Can someone explain to me what the semantics of SHOULD and MUST > > are, to the point that I could see from the explanation that > > Query: +(name:[A TO C] name:[Q TO Z]) +PHB:NO > > would produce the results that I wanted, and > > Query: (name:[A TO C] name:[Q TO Z]) +PHB:NO > > does not? > > With the latter query, since you later said PHB is "NO" for all > documents, you're getting all documents back regardless of the name > field, because you didn't require (MUST) the first clause. The > first query is requiring that PHB:NO __AND__ the name is in those > ranges. The second query is requiring PHB:NO __OR__ the name is in > those ranges. > > > Or why > > Query: (name:[A TO C] name:[Q TO Z]) +PHB:NO > > Again, this one selects all documents regardless of the first SHOULD > clause because PHB is "NO" for all documents and the first clause is > a SHOULD, so it does not matter what it matches other than for > ranking purposes. > > > returns different results than > > Query: (name:[A TO C] name:[Q TO Z]) > > even though PHB is NO for all docs > > This is only selecting documents that match either of those ranges. > > Thanks for submitting the code example to show exactly what you're > doing. > > > Obviously, nesting makes it much harder (for me) to understand. > > What does it mean to nest a SHOULD nested in a MUST (nested in a ...). > > +(name:A OR name:B) +PHB:NO > > means that name must be A __OR__ B (has to be one or the other, or > potentially both even), and PHB has to be "NO". > > Erik > > > > > > Thanks again, > > > > Tom > > > > > > > > > > -----------------------------------------------------------Sample > > import java.io.IOException; > > > > import org.apache.lucene.analysis.Analyzer; > > import org.apache.lucene.analysis.standard.StandardAnalyzer; > > import org.apache.lucene.document.*; > > import org.apache.lucene.index.*; > > import org.apache.lucene.search.*; > > import org.apache.lucene.store.*; > > > > public class Should > > { > > public static final String NAME_FIELD = "name"; > > > > public static final String MANAGER_FIELD = "PHB"; > > > > public static Document createDoc(String name, String manager) > > { > > Document d = new Document(); > > d.add(new Field(NAME_FIELD, name, Field.Store.YES, > > Field.Index.UN_TOKENIZED)); > > d.add(new Field(MANAGER_FIELD, manager, > > Field.Store.YES, Field.Index.UN_TOKENIZED)); > > return d; > > } > > > > static BooleanQuery buildQuery(BooleanClause.Occur occur) > > { > > BooleanQuery bq = new BooleanQuery(); > > Term lower; > > Term upper; > > lower = new Term(NAME_FIELD, "A"); > > upper = new Term(NAME_FIELD, "C"); > > bq.add(new RangeQuery(lower, upper, true), occur); > > lower = new Term(NAME_FIELD, "Q"); > > upper = new Term(NAME_FIELD, "Z"); > > bq.add(new RangeQuery(lower, upper, true), occur); > > return bq; > > } > > > > static BooleanQuery buildQuery2() > > { > > BooleanQuery mainq = buildQuery > > (BooleanClause.Occur.SHOULD); > > mainq.add(new TermQuery(new Term(MANAGER_FIELD, > > "NO")), BooleanClause.Occur.MUST); > > return mainq; > > } > > > > static BooleanQuery buildQuery3(BooleanClause.Occur occur) > > { > > BooleanQuery mainq = new BooleanQuery(); > > BooleanQuery subq = new BooleanQuery(); > > Term lower; > > Term upper; > > lower = new Term(NAME_FIELD, "A"); > > upper = new Term(NAME_FIELD, "C"); > > subq.add(new RangeQuery(lower, upper, true), > > BooleanClause.Occur.SHOULD); > > lower = new Term(NAME_FIELD, "Q"); > > upper = new Term(NAME_FIELD, "Z"); > > subq.add(new RangeQuery(lower, upper, true), > > BooleanClause.Occur.SHOULD); > > mainq.add(subq, occur); > > mainq.add(new TermQuery(new Term(MANAGER_FIELD, > > "NO")), BooleanClause.Occur.MUST); > > return mainq; > > } > > > > private static void runQuery(IndexSearcher is, Query q) > > throws IOException > > { > > Hits hits; > > System.out.println("Query: " + q); > > hits = is.search(q); > > System.out.println("Hits: " + hits.length()); > > for (int docNo = 0; docNo < hits.length(); docNo++) > > { > > System.out.println("DOC: " + hits.doc(docNo)); > > } > > System.out.println(); > > } > > > > public static void main(String[] args) throws Exception > > { > > Directory dir = new RAMDirectory(); > > Analyzer analyzer = new StandardAnalyzer(); > > IndexWriter iw = new IndexWriter(dir, analyzer, true); > > iw.addDocument(createDoc("Asok", "NO")); > > iw.addDocument(createDoc("Dogbert", "NO")); > > iw.addDocument(createDoc("Ratbert", "NO")); > > iw.close(); > > > > IndexSearcher is = new IndexSearcher(dir); > > Query q; > > q = buildQuery(BooleanClause.Occur.SHOULD); > > runQuery(is, q); > > q = buildQuery(BooleanClause.Occur.MUST); > > runQuery(is, q); > > q = buildQuery2(); > > runQuery(is, q); > > q = buildQuery3(BooleanClause.Occur.SHOULD); > > runQuery(is, q); > > q = buildQuery3(BooleanClause.Occur.MUST); > > runQuery(is, q); > > is.close(); > > } > > } > > ----------------Output (I trying to get two results back) > > Query: name:[A TO C] name:[Q TO Z] > > Hits: 2 > > DOC: Document<stored/uncompressed,indexed<name:Asok> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Ratbert> stored/ > > uncompressed,indexed<PHB:NO>> > > > > Query: +name:[A TO C] +name:[Q TO Z] > > Hits: 0 > > > > Query: name:[A TO C] name:[Q TO Z] +PHB:NO > > Hits: 3 > > DOC: Document<stored/uncompressed,indexed<name:Asok> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Ratbert> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Dogbert> stored/ > > uncompressed,indexed<PHB:NO>> > > > > Query: (name:[A TO C] name:[Q TO Z]) +PHB:NO > > Hits: 3 > > DOC: Document<stored/uncompressed,indexed<name:Asok> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Ratbert> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Dogbert> stored/ > > uncompressed,indexed<PHB:NO>> > > > > Query: +(name:[A TO C] name:[Q TO Z]) +PHB:NO > > Hits: 2 > > DOC: Document<stored/uncompressed,indexed<name:Asok> stored/ > > uncompressed,indexed<PHB:NO>> > > DOC: Document<stored/uncompressed,indexed<name:Ratbert> stored/ > > uncompressed,indexed<PHB:NO>> > > > > ------------------ > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
