[GitHub] [lucene] rmuir commented on a change in pull request #166: LUCENE-9905: Move HNSW build parameters to codec

[GitBox <gi...@apache.org>]

rmuir commented on a change in pull request #166:
URL: https://github.com/apache/lucene/pull/166#discussion_r645249079



##########
File path: lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90HnswVectorFormat.java
##########
@@ -76,14 +79,55 @@
   static final int VERSION_START = 0;
   static final int VERSION_CURRENT = VERSION_START;
 
-  /** Sole constructor */
+  static final String BEAM_WIDTH_KEY =
+      Lucene90HnswVectorFormat.class.getSimpleName() + ".beam_width";
+  static final String MAX_CONN_KEY = Lucene90HnswVectorFormat.class.getSimpleName() + ".max_conn";
+
+  /**
+   * Controls how many of the nearest neighbor candidates are connected to the new node. See {@link
+   * HnswGraph} for details.
+   */
+  private final int maxConn;
+
+  /**
+   * The number of candidate neighbors to track while searching the graph for each newly inserted
+   * node. See {@link HnswGraph} for details.
+   */
+  private final int beamWidth;
+
   public Lucene90HnswVectorFormat() {
     super("Lucene90HnswVectorFormat");
+    this.maxConn = HnswGraphBuilder.DEFAULT_MAX_CONN;
+    this.beamWidth = HnswGraphBuilder.DEFAULT_BEAM_WIDTH;
+  }
+
+  public Lucene90HnswVectorFormat(int maxConn, int beamWidth) {
+    super("Lucene90HnswVectorFormat");
+    this.maxConn = maxConn;
+    this.beamWidth = beamWidth;
   }
 
   @Override
   public VectorWriter fieldsWriter(SegmentWriteState state) throws IOException {
-    return new Lucene90HnswVectorWriter(state);
+    SegmentInfo segmentInfo = state.segmentInfo;
+    putFormatAttribute(segmentInfo, MAX_CONN_KEY, String.valueOf(maxConn));
+    putFormatAttribute(segmentInfo, BEAM_WIDTH_KEY, String.valueOf(beamWidth));
+    return new Lucene90HnswVectorWriter(state, maxConn, beamWidth);
+  }
+
+  private void putFormatAttribute(SegmentInfo si, String key, String value) {
+    String previousValue = si.putAttribute(key, value);
+    if (previousValue != null && previousValue.equals(value) == false) {

Review comment:
       I don't think these should be written. If someone is using the per-field impl, and has different fields with different values, then they'd trample each other.




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org