Why input fraction to eigenvector and then evalf output are all the same when input different ?

i would like to see more decimal numbers

digits := 36 

command can not show more decimal numbers

why can not see the difference?

if start from fraction

After set round screen display to 36 digits in options

If start from floating value

the result different from start from fraction

start from which is the most accurate and correct?

if fraction is correct, why all result are the same even if input are different?

Real part + complex part

but 

sometimes it display

complex part + real part 

how to consistent display real part + complex part?

I got decimal places problem,

not known correct or not

the value are different but difference is constant 

i do not know how many places needed to get exact result

i do not believe the difference is constant

because the matrix are different

but even if using 36 decimal places still constant, 

i notice increasing decimal places , the constant difference is changed

is it possible  to output fraction when calculate eigenvector?

if there a series of matrix,

current matrix and previous matrix has same value , then defined as pattern

and

then define x and y are the position of pattern , if Matrix([[777,x,x],[77,7,7],[2,y,y]]

then code auto exchange x and y become Matrix([[777,y,y],[77,7,7],[2,x,x]]

but this recognization need to observe manually, 

can this be automated for any kind of pattern , such as exchange or other operations of pattern?

can neural network or deep learning in maple code recognize the operations such as exchange part of row or flip or symmetric about axis or other need imaginable operations ?

can maple code saved into database for self modifying easily?

self modifying , metaprogramming , database, snapshot, blockchain for code integrity etc