search

bdaiinstitute / spatialmath-python

Create, manipulate and convert representations of position and orientation in 2D or 3D using Python
MIT License
519 stars 84 forks source link

Printing SE3 with symbolic equations #60

Open Jakubach opened 1 year ago

Jakubach commented 1 year ago

Hello, I noticed that matrices with long symbolic equations are printed very shrink. I found my own priting method and I suggest to align them by the longest term like this:

def print_SE3(SE3_matrix : SE3):
    SE3_numpy = SE3_matrix.A
    max_item_length = SE3_numpy.astype(np.string_).dtype.itemsize
    print(np.array2string(SE3_numpy,max_line_width=np.inf,precision=3,suppress_small=True,formatter = {'all':lambda x: f"{str(x):{max_item_length}}"}))

or:

def print_SE3(SE3_matrix : SE3):
    SE3_numpy = SE3_matrix.A
    max_item_length = SE3_numpy.astype(np.string_).dtype.itemsize
    with np.printoptions(precision=3, suppress=True, linewidth = np.inf, formatter = {'all':lambda x: f"{str(x):{max_item_length}}"}):
        print(SE3_numpy)

Results without print_SE3:

  -1.0*sin(theta2)*cos(theta1) 1.0*sin(theta1) 1.0*cos(theta1)*cos(theta2) (d3 + l2)*cos(theta1)*cos(theta2)  
  -1.0*sin(theta1)*sin(theta2) -1.0*cos(theta1) 1.0*sin(theta1)*cos(theta2) (d3 + l2)*sin(theta1)*cos(theta2)  
  1.0*cos(theta2) 0            1.0*sin(theta2) l1 + (d3 + l2)*sin(theta2)  
  0            0            0            1

Results with print_SE3:

[[-1.0*sin(theta2)*cos(theta1)      1.0*sin(theta1)                   1.0*cos(theta1)*cos(theta2)       (d3 + l2)*cos(theta1)*cos(theta2)]
 [-1.0*sin(theta1)*sin(theta2)      -1.0*cos(theta1)                  1.0*sin(theta1)*cos(theta2)       (d3 + l2)*sin(theta1)*cos(theta2)]
 [1.0*cos(theta2)                   0                                 1.0*sin(theta2)                   l1 + (d3 + l2)*sin(theta2)       ]
 [0                                 0                                 0                                 1                                ]]

I know that this function could be better so the inner columns should have less separation but I think it's good first step towards the good change.

petercorke commented 1 year ago

this is a good idea. I want the package to work well for numeric and symbolic values. I wonder what causes the big gaps between the columns? precision, suppress_small and suppress should have no effect here and there aren't really any other formatting options. It would be annoying to have to write the column layout code to fix this...