Pandas Assigning to a Sliced Panel fails when slicing with boolean arrays

This is similar to Issue #1533 and I've adapted the example from there.

import numpy as np
import pandas as pd
a = pd.Panel(items=[1,2,3], major_axis=[11,22,33], minor_axis=[111,222,333])
b = pd.DataFrame(np.random.randn(2,3), index=[111,333], columns=[1,2,3])
c = pd.DataFrame(np.mat('1 2 3; 4 5 6'), index=[111,333], columns=[1,2,3])

# works now
print "\na.ix[...] before"
print a.ix[:, 22, [111, 333]]
a.ix[:, 22, [111, 333]] = b
print "\na.ix[...] after"
print a.ix[:, 22, [111, 333]]

tf = np.array([True, True, True])
tf2 = np.array([True, False, True])
# boolean indexing returns the same data
print "\na.ix[tf, ...] before"
print a.ix[tf, 22, tf2]
a.ix[tf, 22, tf2]
# but assigning to it fails
try:
    a.ix[tf, 22, tf2] = c
    print "\na.ix[tf, ...] after"
    a.ix[tf, 22, tf2]
except Exception, e:
    print e

# however if we first create view into the Panel
v = a.ix[:, 22, :]
print "\nv.ix[tf, ...] before"
print v.ix[tf2, tf]
# then assigning works
v.ix[tf2, tf] = c
print "\nv.ix[tf, ...] after"
print v.ix[tf2, tf]
print "\na.ix[tf, ...] after"
print a.ix[tf, 22, tf2]

produces output

a.ix[...] before
      1   2   3
111 NaN NaN NaN
333 NaN NaN NaN

a.ix[...] after
            1         2         3
111  0.329258  0.759436 -0.070773
333  0.528950 -1.276978  0.084436

a.ix[tf, ...] before
            1         2         3
111  0.329258  0.759436 -0.070773
333  0.528950 -1.276978  0.084436
shape mismatch: objects cannot be broadcast to a single shape

v.ix[tf, ...] before
            1         2         3
111  0.329258  0.759436 -0.070773
333  0.528950 -1.276978  0.084436

v.ix[tf, ...] after
     1  2  3
111  1  2  3
333  4  5  6

a.ix[tf, ...] after
     1  2  3
111  1  2  3
333  4  5  6

Comment From: jreback

not suprised....that section of code is very tricky

Comment From: jreback

closing as Panels are deprecated