underlying rdd to wrap over.
number of rows; if unspecified, we will compute with an inexpensive traversal.
number of columns; if unspecified, we will try to guess with an inexpensive traversal.
cache level to use. (Implementors usually want to override the default!)
unique partitioning tag. Used to detect identically partitioned operands.
true if the matrix is int-keyed, and if it also may have missing rows (will require a lazy fix for some physical operations.
cache level to use.
Action operator -- does not necessary means Spark action; but does mean running BLAS optimizer and writing down Spark graph lineage since last checkpointed DRM.
Collecting DRM to fron-end in-core Matrix.
Dump matrix as computed Mahout's DRM into specified (HD)FS path
output path to dump Matrix to
Explicit extraction of key class Tag
Changes the number of rows in the DRM without actually touching the underlying data.
new row dimension
unique partitioning tag.
if matrix was previously persisted into cache, delete cached representation