FIX: linalg._qz String formatter syntax error

[ilayn]

In the _qz function the warning has the string formatting command

warnings.warn('.... %d ...' % info -1)

which leads to a TypeError since the argument is not inside brackets. Thus, switched to the "modern" version with '...{}...'.format(info-1).

In the meantime also changed the warning type and set the stacklevel and some housekeeping.

To trigger the problem I've used

a = np.array([[-0.0566335083203545, 0.7618601631045951, 0.1380799881600429,  -0.2496547271520516],
              [-0.6953017686620624, 0.1721825836377841, 0.3524311236692528,  -0.0976480305217913],
              [-0.2778527965349485, 0.0822406719680793, 0.3642622234574927,  0.4324046824057498],
              [0.1630969211638234, -0.1925663248287017, 0.4859689326072161,  0.5655695646632785]])

b = np.array([[-0.6282702999111819, 0.2566040837956725, -0.0590286999428144,  -0.4484746121151296],
              [ 2.993120574942286,  -3.0140376750215672, -2.4376036127193954,  -0.1083956777260289],
              [-0.3395903200140198, -0.0297149078005973,  0.5974019693416047,   0.3581544256691085],
              [ 0.19152708930348,    2.3546781693565211, -2.1809506731995891,  -3.6531715901566919]])

aa, bb, *_ = sp.linalg.qz(a, b, output='complex')

[person142]

Comparing to e.g. https://github.com/scipy/scipy/blob/v0.18.1/scipy/linalg/decomp_schur.py#L90 it seems that that linalg doesn't always do this. Why do it here?

[ilayn]

It's just a habit I guess. I can revert it back I really don't mind. Just picked up from some tutorial a while back which stayed.

[person142]

Not like it's a big deal. But maybe if it is to be changed it would be better to them all at once.

[ilayn]

I've cancelled the build and if you don't have other comments, I'm leaning towards merging this.