Some supporting functions that are internal to fields top level methods. These are used to convert between the efficient but opaque format used by spam and more easily checked format based directly on the row and column indices of non zero elements.
spind2full(obj)spam2full(obj)
spind2spam(obj, add.zero.rows=TRUE)
spam2spind(obj)
Either a list with the sparse index components (spind) or an obj of class spam.
If TRUE an entire row is zero add a hard zero value to the element in the first column for each zero row. The spam format requires at least one element in each row to have an explicit value. It is OK if this value is zero but one must be specified.
The differencee in formats is best illustarted by an example:
A 4X5 sparse matrix:
[,1] [,2] [,3] [,4] [,5]
[1,] 1 9 0 0 33
[2,] 0 0 0 26 34
[3,] 3 11 0 27 35
[4,] 0 12 20 0 36
spind format is a list with components "ind", "ra" and "da" here is how the matrix above would be encoded:
ind
I
[1,] 1 1
[2,] 1 2
[3,] 1 5
[4,] 2 4
[5,] 2 5
[6,] 3 1
[7,] 3 2
[8,] 3 4
[9,] 3 5
[10,] 4 2
[11,] 4 3
[12,] 4 5da
[1] 4 5
ra
[1] 1 9 33 26 34 3 11 27 35 12 20 36
spam format is an S4 class with slot names "entries", "colindices", "rowpointers" and "dimension".
entries
[1] 1 9 33 26 34 3 11 27 35 12 20 36
colindices
[1] 1 2 5 4 5 1 2 4 5 2 3 5
rowpointers
[1] 1 4 6 10 13
dimension
[1] 4 5
The row pointers are the position in the array of entries where the next row starts.
NOTE: It is possible for the spind format to have a missing row of all zeroes but this not allowed in spam format and produces an error message.
as.spam