[
https://issues.apache.org/jira/browse/MAHOUT-1596?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14069023#comment-14069023
]
ASF GitHub Bot commented on MAHOUT-1596:
----------------------------------------
Github user dlyubimov commented on a diff in the pull request:
https://github.com/apache/mahout/pull/30#discussion_r15187303
--- Diff:
spark/src/main/scala/org/apache/mahout/sparkbindings/blas/RbindAB.scala ---
@@ -0,0 +1,77 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.mahout.sparkbindings.blas
+
+import org.apache.log4j.Logger
+import scala.reflect.ClassTag
+import org.apache.mahout.sparkbindings.drm.DrmRddInput
+import org.apache.mahout.math._
+import scalabindings._
+import RLikeOps._
+import org.apache.mahout.math.drm.logical.OpRbind
+import org.apache.spark.SparkContext._
+
+/** Physical Rbind */
+object RbindAB {
+
+ private val log = Logger.getLogger(RbindAB.getClass)
+
+ def rbindAB_int[K:ClassTag](op: OpRbind[K], srcA: DrmRddInput[K], srcB:
DrmRddInput[K]): DrmRddInput[K] = {
+
+ val a = srcA.asInstanceOf[DrmRddInput[Int]].toDrmRdd()
+ val b = srcB.asInstanceOf[DrmRddInput[Int]].toDrmRdd()
+ val n1 = op.A.nrow.asInstanceOf[Int]
+
+ val rdd = a union (b map({ case (key, vec) => ((key + n1), vec) }))
+
+ new DrmRddInput(rowWiseSrc = Some(op.ncol ->
rdd)).asInstanceOf[DrmRddInput[K]]
+ }
+
+ def rbindAB_long[K:ClassTag](op: OpRbind[K], srcA: DrmRddInput[K], srcB:
DrmRddInput[K]): DrmRddInput[K] = {
+
+ val a = srcA.asInstanceOf[DrmRddInput[Long]].toDrmRdd()
+ val b = srcB.asInstanceOf[DrmRddInput[Long]].toDrmRdd()
+ val n1 = op.A.nrow
+
+ val rdd = a union (b map({ case (key, vec) => ((key + n1), vec) }))
+
+ new DrmRddInput(rowWiseSrc = Some(op.ncol ->
rdd)).asInstanceOf[DrmRddInput[K]]
+ }
+
+ def rbindAB_string[K:ClassTag](op: OpRbind[K], srcA: DrmRddInput[K],
srcB: DrmRddInput[K]): DrmRddInput[K] = {
+
+ val a = srcA.asInstanceOf[DrmRddInput[String]].toDrmRdd()
+ val b = srcB.asInstanceOf[DrmRddInput[String]].toDrmRdd()
+ val n1 = op.A.nrow
+
+ val rdd = a union b
+
+ new DrmRddInput(rowWiseSrc = Some(op.ncol ->
rdd)).asInstanceOf[DrmRddInput[K]]
+ }
+
+ def rbindAB[K: ClassTag](op: OpRbind[K], srcA: DrmRddInput[K], srcB:
DrmRddInput[K]): DrmRddInput[K] = {
+ if (implicitly[ClassTag[K]] == implicitly[ClassTag[Int]])
+ rbindAB_int(op, srcA, srcB)
+ else if (implicitly[ClassTag[K]] == implicitly[ClassTag[Long]])
+ rbindAB_long(op, srcA, srcB)
+ else if (implicitly[ClassTag[K]] == implicitly[ClassTag[String]])
+ rbindAB_string(op, srcA, srcB)
+ else
+ throw new IllegalArgumentException("Unsupported Key type.")
--- End diff --
There really should be only two types of routines: (1) Int (2) everything
else.
Long does not have ordering semantics (unlike int) because ordering
algebraically is only important for transpositions, and since vector.set(i)
only accepts an Int, there is no special significance for Long keys thru-out
(i.e. log labels are bearing no special meaning compared to, say, 128 bit
hashes).
Other than that, it makes sense. One thing I am not sure if union() is the
best operation here, maybe there's a special opertaion that implies just
dumping all partitions together without doing reshuffling, need to check with
Spark API.
> support for rbind() operator on DRMs
> ------------------------------------
>
> Key: MAHOUT-1596
> URL: https://issues.apache.org/jira/browse/MAHOUT-1596
> Project: Mahout
> Issue Type: Bug
> Reporter: Anand Avati
>
> an R like operator, cbind(), for appending rows of second matrix below the
> rows of first matrix to create a new matrix.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
