Use reflection to determine source repository index

James Moger

2012-03-15 905d31391cb6a808605d9f5b58a8d30274c01e1a

Use reflection to determine source repository index

Unfortunately Lucene's MultiReader class does not expose a public method
for determining which index reader a document comes from.  There is a
private method which determines this information.

My new approach is to use reflection to call this private method and
determine the actual source index at runtime instead of storing the
repository name in the index. This gets me out of the "rename
repository" problem at the expense of a little runtime performance.

 src/com/gitblit/LuceneExecutor.java

@@ -21,6 +21,7 @@
import java.io.File;

import java.io.IOException;

import java.io.InputStream;

import java.lang.reflect.Method;

import java.text.MessageFormat;

import java.text.ParseException;

import java.util.ArrayList;

@@ -103,14 +104,13 @@
public class LuceneExecutor implements Runnable {

    

        

    private static final int INDEX_VERSION = 1;

    private static final int INDEX_VERSION = 2;



    private static final String FIELD_OBJECT_TYPE = "type";

    private static final String FIELD_ISSUE = "issue";

    private static final String FIELD_PATH = "path";

    private static final String FIELD_COMMIT = "commit";

    private static final String FIELD_BRANCH = "branch";

    private static final String FIELD_REPOSITORY = "repository";

    private static final String FIELD_SUMMARY = "summary";

    private static final String FIELD_CONTENT = "content";

    private static final String FIELD_AUTHOR = "author";

@@ -335,6 +335,24 @@
        }

        return name;

    }

	
    /**

     * Get the tree associated with the given commit.

     *

     * @param walk

     * @param commit

     * @return tree

     * @throws IOException

     */

    protected RevTree getTree(final RevWalk walk, final RevCommit commit)

            throws IOException {

        final RevTree tree = commit.getTree();

        if (tree != null) {

            return tree;

        }

        walk.parseHeaders(commit);

        return commit.getTree();

    }



    /**

     * Construct a keyname from the branch.

@@ -501,7 +519,6 @@
                        

                        Document doc = new Document();

                        doc.add(new Field(FIELD_OBJECT_TYPE, SearchObjectType.blob.name(), Store.YES, Index.NOT_ANALYZED_NO_NORMS));

                        doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.ANALYZED));

                        doc.add(new Field(FIELD_BRANCH, branchName, Store.YES, Index.ANALYZED));

                        doc.add(new Field(FIELD_COMMIT, commit.getName(), Store.YES, Index.ANALYZED));

                        doc.add(new Field(FIELD_PATH, path, Store.YES, Index.ANALYZED));

@@ -542,7 +559,6 @@
                // index the tip commit object

                if (indexedCommits.add(tipId)) {

                    Document doc = createDocument(tip, tags.get(tipId));

                    doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.ANALYZED));

                    doc.add(new Field(FIELD_BRANCH, branchName, Store.YES, Index.ANALYZED));

                    writer.addDocument(doc);

                    result.commitCount += 1;

@@ -557,7 +573,6 @@
                    String hash = rev.getId().getName();

                    if (indexedCommits.add(hash)) {

                        Document doc = createDocument(rev, tags.get(hash));

                        doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.ANALYZED));

                        doc.add(new Field(FIELD_BRANCH, branchName, Store.YES, Index.ANALYZED));

                        writer.addDocument(doc);

                        result.commitCount += 1;

@@ -577,7 +592,6 @@
                for (IssueModel issue : issues) {

                    result.issueCount++;

                    Document doc = createDocument(issue);

                    doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.ANALYZED));

                    writer.addDocument(doc);

                }

            }

@@ -594,24 +608,6 @@
        return result;

    }

    

    /**

     * Get the tree associated with the given commit.

     *

     * @param walk

     * @param commit

     * @return tree

     * @throws IOException

     */

    protected RevTree getTree(final RevWalk walk, final RevCommit commit)

            throws IOException {

        final RevTree tree = commit.getTree();

        if (tree != null) {

            return tree;

        }

        walk.parseHeaders(commit);

        return commit.getTree();

    }



    /**

     * Incrementally update the index with the specified commit for the

     * repository.

@@ -659,7 +655,6 @@
                    Document doc = new Document();

                    doc.add(new Field(FIELD_OBJECT_TYPE, SearchObjectType.blob.name(), Store.YES,

                            Index.NOT_ANALYZED));

                    doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.ANALYZED));

                    doc.add(new Field(FIELD_BRANCH, branch, Store.YES, Index.ANALYZED));

                    doc.add(new Field(FIELD_COMMIT, commit.getName(), Store.YES, Index.ANALYZED));

                    doc.add(new Field(FIELD_PATH, path.path, Store.YES, Index.ANALYZED));

@@ -844,7 +839,7 @@
        }

        return result;

    }



	
    /**

     * Creates a Lucene document from an issue.

     * 

@@ -906,7 +901,6 @@
     */

    private boolean index(String repositoryName, Document doc) {

        try {            

            doc.add(new Field(FIELD_REPOSITORY, repositoryName, Store.YES, Index.NOT_ANALYZED));

            IndexWriter writer = getIndexWriter(repositoryName);

            writer.addDocument(doc);

            resetIndexSearcher(repositoryName);

@@ -926,7 +920,6 @@
        result.author = doc.get(FIELD_AUTHOR);

        result.committer = doc.get(FIELD_COMMITTER);

        result.type = SearchObjectType.fromName(doc.get(FIELD_OBJECT_TYPE));

        result.repository = doc.get(FIELD_REPOSITORY);

        result.branch = doc.get(FIELD_BRANCH);

        result.commitId = doc.get(FIELD_COMMIT);

        result.issueId = doc.get(FIELD_ISSUE);

@@ -1057,7 +1050,7 @@
                    readers.add(repositoryIndex.getIndexReader());

                }

                IndexReader[] rdrs = readers.toArray(new IndexReader[readers.size()]);

                MultiReader reader = new MultiReader(rdrs);

                MultiSourceReader reader = new MultiSourceReader(rdrs);

                searcher = new IndexSearcher(reader);

            }

            Query rewrittenQuery = searcher.rewrite(query);

@@ -1067,8 +1060,17 @@
            for (int i = 0; i < hits.length; i++) {

                int docId = hits[i].doc;

                Document doc = searcher.doc(docId);

                // TODO identify the source index for the doc, then eliminate FIELD_REPOSITORY

                // TODO identify the source index for the doc, then eliminate FIELD_REPOSITORY				
                SearchResult result = createSearchResult(doc, hits[i].score);

                if (repositories.length == 1) {

                    // single repository search

                    result.repository = repositories[0];

                } else {

                    // multi-repository search

                    MultiSourceReader reader = (MultiSourceReader) searcher.getIndexReader();

                    int index = reader.getSourceIndex(docId);

                    result.repository = repositories[index];

                }

                String content = doc.get(FIELD_CONTENT);                

                result.fragment = getHighlightedFragment(analyzer, query, content, result);

                results.add(result);

@@ -1188,4 +1190,38 @@
            return (endTime - startTime)/1000f;

        }

    }

	
    /**

     * Custom subclass of MultiReader to identify the source index for a given

     * doc id.  This would not be necessary of there was a public method to

     * obtain this information.

     *  
     */

    private class MultiSourceReader extends MultiReader {

		
        final Method method;

		
        MultiSourceReader(IndexReader[] subReaders) {

            super(subReaders);

            Method m = null;

            try {

                m = MultiReader.class.getDeclaredMethod("readerIndex", int.class);

                m.setAccessible(true);

            } catch (Exception e) {

                logger.error("Error getting readerIndex method", e);

            }

            method = m;

        }

		
        int getSourceIndex(int docId) {

            int index = -1;

            try {

                Object o = method.invoke(this, docId);

                index = (Integer) o;

            } catch (Exception e) {

                logger.error("Error getting source index", e);

            }

            return index;

        }

    }

}