Package org.apache.sysds.runtime.io
Class ReaderDelta
java.lang.Object
org.apache.sysds.runtime.io.MatrixReader
org.apache.sysds.runtime.io.ReaderDelta
- Direct Known Subclasses:
ReaderDeltaParallel
Single-threaded native Delta Lake reader for matrices, built on the
Spark-free Delta Kernel library. It opens the latest snapshot of a Delta
table directory, reads its parquet data files through the kernel's default
engine (honoring deletion vectors), and materializes the numeric columns
into a dense
MatrixBlock.
Only numeric columns (double/float/long/int/short/byte/boolean) are supported, matching the all-double nature of a SystemDS matrix. Dimensions do not need to be known up front: the row count is discovered while scanning and the column count is taken from the table schema.
-
Constructor Summary
Constructors -
Method Summary
Modifier and TypeMethodDescriptionreadMatrixFromHDFS(String fname, long rlen, long clen, int blen, long estnnz) readMatrixFromInputStream(InputStream is, long rlen, long clen, int blen, long estnnz) Methods inherited from class org.apache.sysds.runtime.io.MatrixReader
checkValidInputFile
-
Constructor Details
-
ReaderDelta
public ReaderDelta()
-
-
Method Details
-
readMatrixFromHDFS
public MatrixBlock readMatrixFromHDFS(String fname, long rlen, long clen, int blen, long estnnz) throws IOException, DMLRuntimeException - Specified by:
readMatrixFromHDFSin classMatrixReader- Throws:
IOExceptionDMLRuntimeException
-
readMatrixFromInputStream
public MatrixBlock readMatrixFromInputStream(InputStream is, long rlen, long clen, int blen, long estnnz) throws IOException, DMLRuntimeException - Specified by:
readMatrixFromInputStreamin classMatrixReader- Throws:
IOExceptionDMLRuntimeException
-