read me .txt

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Polynomial Number class for Matlab
m-files
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

fields:

.mant:
column vector

.exponent:
int64

The length(.mant) after operation is adjusted to max length() of arguments, so it is *very important* to set proper length of the first of argument in sequence of calculation.


PN/NORMALIZED r = normalized( p )
Normalization:
First digit of mantissa, i.e. .mant(1) ~= 0
exeption for (~0~): .mant == [ 0 0 ... 0 ]
 (~0~) can have any .exponent ( so it is possible to create  (~0~)*(~1~0~)^a, like in mpower.m)

%DONE: AlwaysNormalized conception:
After operations like PN([...]), +, -, ln() result is normalized, so PNs ale always normalized (so comparing PNs to (~0~) means comparing .mant(1) to 0)


%TODO: Alternative conception DoNotTruncation
----------------------------------------------------------------
After operations like +, -, * result could have longer mantissa. Now mantissa is truncated to the longest of the arguments length. This behaviour is applicable to real time calculation.
But for some test ("off line"), when time and memory consuming is not important truncation could not be performed.


% PN/DOUBLE conversion PN -> vector, 
----------------------------------------------------------------
PN digits on the left of ~...~ are neglected, because in many cases this digits means abstractions (negative time samples, abstract function coefficients) 

%TODO: In eq.m averaging eps will be more usable:  

%TODO: In mrdivide.m optimization (inv. using conv)
