In the javadoc of *computeImportance()*, it shows the importance of node is Sum{*computeImportanceOfChild()*}, but in the implementation it shows Max{*computeImportanceOfChild()*}, this is inconsistent. The Javadoc has some errors, it will cause some confusion.

// /** * Computes the <dfn>importance</dfn> of a node. Importance is defined as * follows: * * <ul> * <li>the root {@link RelSubset} has an importance of 1</li> * <li>the importance of any other subset is the sum of its importance to * its parents</li> * <li>The importance of children is pro-rated according to the cost of the * children. Consider a node which has a cost of 3, and children with costs * of 2 and 5. The total cost is 10. If the node has an importance of .5, * then the children will have importance of .1 and .25. The retains .15 * importance points, to reflect the fact that work needs to be done on the * node's algorithm.</li> * </ul> * * <p>The formula for the importance <i>I</i> of node n is: * * <blockquote>I<sub>n</sub> = Sum<sub>parents p of n</sub>{I<sub>p</sub> . * W <sub>n, p</sub>}</blockquote> * * <p>where W<sub>n, p</sub>, the weight of n within its parent p, is * * <blockquote>W<sub>n, p</sub> = Cost<sub>n</sub> / (SelfCost<sub>p</sub> + * Cost<sub>n0</sub> + ... + Cost<sub>nk</sub>) * </blockquote> */ double computeImportance(RelSubset subset) { double importance; if (subset == planner.root) { // The root always has importance = 1 importance = 1.0; } else { final RelMetadataQuery mq = subset.getCluster().getMetadataQuery(); // The importance of a subset is the max of its importance to its // parents importance = 0.0; for (RelSubset parent : subset.getParentSubsets(planner)) { final double childImportance = computeImportanceOfChild(mq, subset, parent); importance = Math.max(importance, childImportance); } } LOGGER.trace("Importance of [{}] is {}", subset, importance); return importance; }

- links to