[
https://issues.apache.org/jira/browse/MAHOUT-1578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14028844#comment-14028844
]
ASF GitHub Bot commented on MAHOUT-1578:
----------------------------------------
Github user dlyubimov commented on a diff in the pull request:
https://github.com/apache/mahout/pull/16#discussion_r13688078
--- Diff: mrlegacy/src/main/java/org/apache/mahout/math/MatrixWritable.java
---
@@ -113,27 +113,28 @@ public static Matrix readMatrix(DataInput in) throws
IOException {
int rows = in.readInt();
int columns = in.readInt();
+ byte vectorFlags = in.readByte();
+
Matrix matrix;
+
if (dense) {
matrix = new DenseMatrix(rows, columns);
- } else {
- if (isSparseRowMatrix) {
- matrix = new SparseRowMatrix(rows, columns, sequential);
- } else {
- matrix = new SparseMatrix(rows, columns);
+ for (int row = 0; row < rows; row++) {
+ matrix.assignRow(row, VectorWritable.readVector(in, vectorFlags,
columns));
}
- }
-
- if (dense || isSparseRowMatrix) {
+ } else if (isSparseRowMatrix) {
+ Vector[] rowVectors = new Vector[rows];
for (int row = 0; row < rows; row++) {
- matrix.assignRow(row, VectorWritable.readVector(in));
--- End diff --
is that where the major problem was? is that because assignment of
sequential vector to sequential vector is that slow? or this is an assignment
of random vector to a sequential vector? (sequential to sequential actually
should be ok methinks).
Anyway I don't see any immediate problems, and the quality of your work
usually doesn't require any deep scrutiny, so i'd say ship it. Actually the
sooner the better, because i am very close to actually give it all a good
spanking
> Optimizations in matrix serialization
> -------------------------------------
>
> Key: MAHOUT-1578
> URL: https://issues.apache.org/jira/browse/MAHOUT-1578
> Project: Mahout
> Issue Type: Improvement
> Components: Math
> Reporter: Sebastian Schelter
> Fix For: 1.0
>
>
> MatrixWritable contains inefficient code in a few places:
>
> * type and size are stored with every vector, although they are the same for
> every vector
> * in some places vectors are added to the matrix via assign() in places
> where we could directly use the instance
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
