The protected part of VMatrix must be accessible to the
matrix stack. The signature functions as in the vmm and
vdoub. It provides an indication of how corrupt or active
the matrix has become.
The function mindex( int i, int j)
calculates a vdoub vector index.![[*]](/file/16906/SOURCE.mdf/programm/msdos/c/yamp/vmatrix.tex/footnote.png)
Stripped of the casts, the function is
simply (i-1)*c + (j-1) , where i and j are matrix
indices, and c is the number of columns of the matrix.
Matrices are indexed from 1 to r, and 1 to c, so there are
r*c elements in the corresponding vdoub. However the vdoub
indices range from 0 to r*c - 1, so the i-1 and j-1 force
the minimum matrix indices to be 1. This calculation forces
the matrix to be in row major order. The
curvecind is a long integer
containing the vdoub index of the current matrix element.
The name is a string type which can be
manipulated in the matrix functions. The functions are
found in the public part, but are Nameit(), Getname(),
Showname(), and StringAddress().
The next three functions control the allocation of the
vdoub vectors. Replace() calls
PurgeVectors() and
SetupVectors() if the VMatrix
being replaced has different dimensions than the replacing
VMatrix. Replace() is called by the
operator=(). PurgeVectors()
calls vfree() to free the hdr in the protected part of
vdoub. SetupVectors() calls vmalloc() to create a new hdr
for vdoub.
For the in-ram version, SetupVectors() allocates r*c+1 elements
on the heap using calloc(). The first element is set to
SIGNATURE, then the base is shifted by 1. Thus, m[-1] is a check
value. If the value is not equal to SIGNATURE, then the matrix
has been deleted or clobbered.
The function NewReference()
is used by the matrix stack to
set up a reference to a base vector instead of a copy. It is
part of the deep copy technique used by push.