matrix_ops.sql_in File Reference

Implementation of matrix operations in SQL. More...

Functions
text	matrix_info (text message)
text	matrix_info ()
matrix_result	matrix_mult (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out, text out_args)
	Multiplies two matrices. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column index. More...
matrix_result	matrix_mult (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out)
text	matrix_mult (text message)
text	matrix_mult ()
matrix_result	matrix_add (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out, text out_args)
	Adds two matrices. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column index. More...
matrix_result	matrix_add (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out)
text	matrix_add (text message)
text	matrix_add ()
matrix_result	matrix_sub (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out, text out_args)
	Subs two matrices. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
matrix_result	matrix_sub (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out)
text	matrix_sub (text message)
text	matrix_sub ()
float8[]	matrix_extract_row (text matrix_in, text in_args, integer index)
	Extract row from matrix given row index. It requires that all the values are NON-NULL. This is the dense or sparse representation of the matrix where the matrix elements are indexed by row id for dense and by the row and column id for sparse. More...
text	matrix_extract_row (text message)
text	matrix_extract_row ()
float8[]	matrix_extract_col (text matrix_in, text in_args, integer index)
	Extract column from matrix given column index. It requires that all the values are NON-NULL. This is the dense or sparse representation of the matrix where the matrix elements are indexed by row id for dense and by the row and column id for sparse. More...
text	matrix_extract_col (text message)
text	matrix_extract_col ()
matrix_result	matrix_max (text matrix_in, text in_args, integer dim, text matrix_out, boolean fetch_index)
	Get max value along dimension from matrix. Also returns corresponding index if set fetch_index. It requires that all the values are NON-NULL. This is the dense or sparse representation of the matrix where the matrix elements are indexed by row id for dense and by the row and column id for sparse. More...
text	matrix_max (text message)
text	matrix_max ()
matrix_result	matrix_min (text matrix_in, text in_args, integer dim, text matrix_out, boolean fetch_index)
	Get min value along dimension from matrix. Also returns corresponding index if set fetch_index. It requires that all the values are NON-NULL. This is the dense or sparse representation of the matrix where the matrix elements are indexed by row id for dense and by the row and column id for sparse. More...
text	matrix_min (text message)
text	matrix_min ()
float8[]	matrix_sum (text matrix_in, text in_args, integer dim)
	Calculate sum along dimension for matrix. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
text	matrix_sum (text message)
text	matrix_sum ()
float8[]	matrix_mean (text matrix_in, text in_args, integer dim)
	Calculate mean along dimension for matrix. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
text	matrix_mean (text message)
text	matrix_mean ()
matrix_result	matrix_scalar_mult (text matrix_in, text in_args, float8 scalar, text matrix_out, text out_args)
	Multiplies one matrix with scalar. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
text	matrix_scalar_mult (text message)
text	matrix_scalar_mult ()
float8[]	matrix_vec_mult (text matrix_in, text in_args, float8[] vector)
	Multiplies one matrix with vector. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
text	matrix_vec_mult (text message)
text	matrix_vec_mult ()
matrix_result	matrix_elem_mult (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out, text out_args)
	Multiplies two matrices by element. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column id. More...
matrix_result	matrix_elem_mult (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out)
text	matrix_elem_mult (text message)
text	matrix_elem_mult ()
matrix_result	matrix_trans (text matrix_in, text in_args, text matrix_out, text out_args, boolean use_temp_table)
	Transpose matrix. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row and column index. More...
matrix_result	matrix_trans (text matrix_in, text in_args, text matrix_out, text out_args)
matrix_result	matrix_trans (text matrix_in, text in_args, text matrix_out)
text	matrix_trans (text message)
text	matrix_trans ()
matrix_result	matrix_sparsify (text matrix_in, text in_args, text matrix_out, text out_args, boolean use_temp_table)
	Converts matrix to sparse representation. It requires that all the values are NON-NULL. This is the dense representation of the matrix where the matrix elements are indexed by the row and column index. More...
matrix_result	matrix_sparsify (text matrix_in, text in_args, text matrix_out, text out_args)
matrix_result	matrix_sparsify (text matrix_in, text in_args, text matrix_out)
text	matrix_sparsify (text message)
text	matrix_sparsify ()
matrix_result	matrix_densify (text matrix_in, text in_args, text matrix_out, text out_args, boolean use_temp_table)
	Converts matrix to dense representation. It requires that all the values are NON-NULL. This is the sparse representation of the matrix where the matrix elements are indexed by the row index. More...
matrix_result	matrix_densify (text matrix_in, text in_args, text matrix_out, text out_args)
matrix_result	matrix_densify (text matrix_in, text in_args, text matrix_out)
text	matrix_densify (text message)
text	matrix_densify ()
void	matrix_scale_and_add (text matrix_a, text a_args, text matrix_b, text b_args, float8 scale, text matrix_out, text out_args)
	Scale and add matrix operation ( R <- A + sB) More...
float8	matrix_norm (text matrix_in, text in_args, text norm_type)
	Compute matrix norm (of various types) More...
float8	matrix_norm (text matrix_in, text in_args)
	Compute matrix norm of fro. More...
matrix_result	matrix_block_mult (text matrix_a, text a_args, text matrix_b, text b_args, text matrix_out, text out_args)
matrix_result	matrix_block_square (text matrix_in, text in_args, text matrix_out, text out_args)
matrix_result	matrix_block_trans (text matrix_in, text in_args, text matrix_out, text out_args, boolean use_temp_table)
matrix_result	matrix_block_trans (text matrix_in, text in_args, text matrix_out, text out_args)
matrix_result	matrix_square (text matrix_in, text in_args, text matrix_out, text out_args)
	Calculate square of matrix. It requires that all the values are NON-NULL. This is the dense representation of the matrix where the matrix elements are indexed by the row index. More...
matrix_result	matrix_square (text matrix_in, text in_args, text matrix_out)
integer[]	__rand_vector (integer dim)
integer[]	__rand_block (integer row_dim, integer col_dim)
integer[]	__rand_block (integer dim)
set< float8[]>	__matrix_row_split (float8[] row_in, integer size)
float8[]	__matrix_densify_sfunc (float8[] state, integer col_dim, integer col_id, float8 val)
aggregate float8[]	__matrix_densify_agg (integer, integer, float8)
float8[]	__matrix_blockize_sfunc (float8[] state, integer row_id, float8[] row_vec, integer rsize)
aggregate float8[]	__matrix_blockize_agg (integer, float8[], integer)
float8[]	__matrix_unblockize_sfunc (float8[] state, integer total_col_dim, integer col_id, float8[] row_vec)
aggregate float8[]	__matrix_unblockize_agg (integer, integer, float8[])
float8[]	matrix_mem_mult (float[] matrix_a, float[] matrix_b, boolean trans_b)
float8[]	matrix_mem_mult (float[] matrix_a, float[] matrix_b)
float8[]	matrix_mem_trans (float[] matrix)
float8[]	__matrix_mem_sum_sfunc (float[] state, float[] matrix)
aggregate float8[]	__matrix_mem_sum (float8[])
matrix_result	__matrix_column_to_array_format (text matrix_in, text row_id, text matrix_out, boolean istemp)
matrix_result	matrix_blockize (text matrix_in, text in_args, integer rsize, integer csize, text matrix_out, text out_args)
matrix_result	matrix_blockize (text matrix_in, text in_args, integer rsize, integer csize, text matrix_out)
matrix_result	matrix_unblockize (text matrix_in, text in_args, text matrix_out, text out_args)
set< float8[]>	__matrix_unnest_block (float8[] block)

Detailed Description

Date

April 2011

Function Documentation

aggregate float8 [] __matrix_blockize_agg	(	integer	,
		float8	[],
		integer
	)

float8 [] __matrix_blockize_sfunc	(	float8[]	state,
		integer	row_id,
		float8[]	row_vec,
		integer	rsize
	)

matrix_result __matrix_column_to_array_format	(	text	matrix_in,
		text	row_id,
		text	matrix_out,
		boolean	istemp
	)

aggregate float8 [] __matrix_densify_agg	(	integer	,
		integer	,
		float8
	)

float8 [] __matrix_densify_sfunc	(	float8[]	state,
		integer	col_dim,
		integer	col_id,
		float8	val
	)

aggregate float8 [] __matrix_mem_sum

(

float8

[]

)

float8 [] __matrix_mem_sum_sfunc	(	float[]	state,
		float[]	matrix
	)

set<float8[]> __matrix_row_split	(	float8[]	row_in,
		integer	size
	)

aggregate float8 [] __matrix_unblockize_agg	(	integer	,
		integer	,
		float8	[]
	)

float8 [] __matrix_unblockize_sfunc	(	float8[]	state,
		integer	total_col_dim,
		integer	col_id,
		float8[]	row_vec
	)

set<float8[]> __matrix_unnest_block

(

float8[]

block

)

integer [] __rand_block	(	integer	row_dim,
		integer	col_dim
	)

integer [] __rand_block

(

integer

dim

)

integer [] __rand_vector

(

integer

dim

)

matrix_result matrix_add	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_a	Name of the table containing the first matrix
a_row	Name of the column containing the row index for the first matrix
a_col	Name of the column containing the column index for the first matrix
a_val	Name of the column containing the matrix values for the first matrix
matrix_b	Name of the table containing the second matrix
b_row	Name of the column containing the row index for the second matrix
b_col	Name of the column containing the column index for the second matrix
b_val	Name of the column containing the matrix values for the second matrix
matrix_out	Name of the table where to output the result matrix

Returns

Name of the table containing the result matrix

matrix_result matrix_add	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out
	)

text matrix_add

(

text

message

)

text matrix_add

(

)

matrix_result matrix_block_mult	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_block_square	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_block_trans	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args,
		boolean	use_temp_table
	)

matrix_result matrix_block_trans	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_blockize	(	text	matrix_in,
		text	in_args,
		integer	rsize,
		integer	csize,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_blockize	(	text	matrix_in,
		text	in_args,
		integer	rsize,
		integer	csize,
		text	matrix_out
	)

matrix_result matrix_densify	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args,
		boolean	use_temp_table
	)

Parameters

matrix_in	Name of the table containing the input matrix
in_row	Name of the column containing the row index for the input matrix
in_col	Name of the column containing the column index for the input matrix
in_val	Name of the column containing the matrix values for the input matrix
matrix_out	Name of the table where to output the result matrix
use_temp_table	Specify if use temp table for result

Returns

Name of the table containing the result matrix

matrix_result matrix_densify	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_densify	(	text	matrix_in,
		text	in_args,
		text	matrix_out
	)

text matrix_densify

(

text

message

)

text matrix_densify

(

)

matrix_result matrix_elem_mult	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_a	Name of the table containing the first matrix
a_row	Name of the column containing the row id for the first matrix
a_col	Name of the column containing the column id for the first matrix
a_val	Name of the column containing the matrix values for the first matrix
matrix_b	Name of the table containing the second matrix
b_row	Name of the column containing the row id for the second matrix
b_col	Name of the column containing the column id for the second matrix
b_val	Name of the column containing the matrix values for the second matrix
matrix_out	Name of the table where to output the result matrix

Returns

Name of the table containing the result matrix

matrix_result matrix_elem_mult	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out
	)

text matrix_elem_mult

(

text

message

)

text matrix_elem_mult

(

)

float8 [] matrix_extract_col	(	text	matrix_in,
		text	in_args,
		integer	index
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	A string containing multiple named arguments of the form "name=value". This argument is used as a container for multiple parameters related to a single matrix.
index	Column index for wanted.

Returns

Vector containing given column elements.

text matrix_extract_col

(

text

message

)

text matrix_extract_col

(

)

float8 [] matrix_extract_row	(	text	matrix_in,
		text	in_args,
		integer	index
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	A string containing multiple named arguments of the form "name=value". This argument is used as a container for multiple parameters related to a single matrix.
index	Row index for wanted.

Returns

Vector containing given row elements.

text matrix_extract_row

(

text

message

)

text matrix_extract_row

(

)

text matrix_info

(

text

message

)

text matrix_info

(

)

matrix_result matrix_max	(	text	matrix_in,
		text	in_args,
		integer	dim,
		text	matrix_out,
		boolean	fetch_index
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	A string containing multiple named arguments of the form "name=value". This argument is used as a container for multiple parameters related to a single matrix.
dim	Which dimension to collapse. 1 means row wise, and 2 means column wise
matrix_out	Result matrix which contains one column: max, and another column: index if fetch_index is set. The type of column is array
fetch_index	True when want to get corresponding index of max value. Default is false

Returns

matrix_out.

text matrix_max

(

text

message

)

text matrix_max

(

)

float8 [] matrix_mean	(	text	matrix_in,
		text	in_args,
		integer	dim
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	Name-value pair string containing options for matrix_in
dim	Which dimension to collapse. 1 means row wise, and 2 means column wise

text matrix_mean

(

text

message

)

text matrix_mean

(

)

float8 [] matrix_mem_mult	(	float[]	matrix_a,
		float[]	matrix_b,
		boolean	trans_b
	)

float8 [] matrix_mem_mult	(	float[]	matrix_a,
		float[]	matrix_b
	)

float8 [] matrix_mem_trans

(

float[]

matrix

)

matrix_result matrix_min	(	text	matrix_in,
		text	in_args,
		integer	dim,
		text	matrix_out,
		boolean	fetch_index
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	A string containing multiple named arguments of the form "name=value". This argument is used as a container for multiple parameters related to a single matrix.
dim	Which dimension to collapse. 1 means row wise, and 2 means column wise
matrix_out	Result matrix which contains one column: min, and another column: index if fetch_index is set. The type of column is array
fetch_index	True when want to get corresponding index of min value. Default is false

Returns

matrix_out.

text matrix_min

(

text

message

)

text matrix_min

(

)

matrix_result matrix_mult	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_a	Name of the table containing the first matrix
a_row	Name of the column containing the row index for the first matrix
a_col	Name of the column containing the column index for the first matrix
a_val	Name of the column containing the matrix values for the first matrix
a_trans	Boolean to indicate if first matrix should be transposed before multiplication
matrix_b	Name of the table containing the second matrix
b_row	Name of the column containing the row index for the second matrix
b_col	Name of the column containing the column index for the second matrix
b_val	Name of the column containing the matrix values for the second matrix
b_trans	Boolean to indicate if second matrix should be transposed before multiplication
matrix_out	Name of the table where to output the result matrix

Returns

Name of the table containing the result matrix

matrix_result matrix_mult	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out
	)

text matrix_mult

(

text

message

)

text matrix_mult

(

)

float8 matrix_norm	(	text	matrix_in,
		text	in_args,
		text	norm_type
	)

Parameters

matrix_a	Source matrix table (dense matrix format)
norm_type	Type of norm used (default: fro)

Returns

DOUBLE PRECISION

Usage Currenty, we support the following norms

(a) 'fro': Computes the frobenius norm

SELECT matrix_norm('matrix_table_name', 'fro');

float8 matrix_norm	(	text	matrix_in,
		text	in_args
	)

Parameters

matrix_a

Source matrix table (dense matrix format)

Returns

DOUBLE PRECISION

Usage Currenty, we support the following norms

(a) 'fro': Computes the frobenius norm

SELECT matrix_norm('matrix_table_name', 'fro');

matrix_result matrix_scalar_mult	(	text	matrix_in,
		text	in_args,
		float8	scalar,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	Name-value pair string containing options for matrix_in
scalar	Scalar value used to be multiplied with matrix_in
matrix_out	Name of the table where to output the result matrix
out_args	Name-value pair string containing options for matrix_in

Returns

Name of the table containing the result matrix

text matrix_scalar_mult

(

text

message

)

text matrix_scalar_mult

(

)

void matrix_scale_and_add	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		float8	scale,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_a	Source matrix table A (dense matrix format)
matrix_b	Source matrix table B (dense matrix format)
scale	Scale for matrix B
matrix_out	Output table for matrix R <- A + sB

Returns

VOID

Usage

For matrix addition and subtration, you can use the same function with different values of scalar

SELECT matrix_scale_and_add('matrix_a', 'matrix_b', 1, 'matrix_out');
SELECT matrix_scale_and_add('matrix_a', 'matrix_b', -1, 'matrix_out');

matrix_result matrix_sparsify	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args,
		boolean	use_temp_table
	)

Parameters

matrix_in	Name of the table containing the input matrix
matrix_out	Name of the table where to output the result matrix
use_temp_table	Specify if use temp table for result

Returns

Name of the table containing the result matrix

matrix_result matrix_sparsify	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_sparsify	(	text	matrix_in,
		text	in_args,
		text	matrix_out
	)

text matrix_sparsify

(

text

message

)

text matrix_sparsify

(

)

matrix_result matrix_square	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_in	Name of the table containing the input matrix
matrix_out	Name of the table where to output the result matrix

Returns

Name of the table containing the result matrix

matrix_result matrix_square	(	text	matrix_in,
		text	in_args,
		text	matrix_out
	)

matrix_result matrix_sub	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out,
		text	out_args
	)

Parameters

matrix_a	Name of the table containing the first matrix
a_row	Name of the column containing the row id for the first matrix
a_col	Name of the column containing the column id for the first matrix
a_val	Name of the column containing the matrix values for the first matrix
matrix_b	Name of the table containing the second matrix
b_row	Name of the column containing the row id for the second matrix
b_col	Name of the column containing the column id for the second matrix
b_val	Name of the column containing the matrix values for the second matrix
matrix_out	Name of the table where to output the result matrix

Returns

Name of the table containing the result matrix

matrix_result matrix_sub	(	text	matrix_a,
		text	a_args,
		text	matrix_b,
		text	b_args,
		text	matrix_out
	)

text matrix_sub

(

text

message

)

text matrix_sub

(

)

float8 [] matrix_sum	(	text	matrix_in,
		text	in_args,
		integer	dim
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	Name-value pair string containing options for matrix_in
dim	Which dimension to collapse. 1 means row wise, and 2 means column wise

text matrix_sum

(

text

message

)

text matrix_sum

(

)

matrix_result matrix_trans	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args,
		boolean	use_temp_table
	)

Parameters

matrix_in	Name of the table containing the input matrix
in_row	Name of the column containing the row index for the input matrix
in_col	Name of the column containing the column index for the input matrix
in_val	Name of the column containing the matrix values for the input matrix
matrix_out	Name of the table where to output the result matrix
use_temp_table	Specify if use temp table for result

Returns

Name of the table containing the result matrix

matrix_result matrix_trans	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

matrix_result matrix_trans	(	text	matrix_in,
		text	in_args,
		text	matrix_out
	)

text matrix_trans

(

text

message

)

text matrix_trans

(

)

matrix_result matrix_unblockize	(	text	matrix_in,
		text	in_args,
		text	matrix_out,
		text	out_args
	)

float8 [] matrix_vec_mult	(	text	matrix_in,
		text	in_args,
		float8[]	vector
	)

Parameters

matrix_in	Name of the table containing the matrix
in_args	Name-value pair string containing options for matrix_in
vector	Vector value used to be multiplied with matrix_in

Returns

Column vector which is the result of matrix_in * vector

text matrix_vec_mult

(

text

message

)

text matrix_vec_mult

(

)