Repository navigation
coercion rules for boolean arrays are broken (Trac #453) #1051
Description
Original ticket http://projects.scipy.org/numpy/ticket/453 on 2007-02-16 by @chanley, assigned to unknown.
In Python boolean multiplication results in the following:
In [1]: True * 4096
Out[1]: 4096
NUMARRAY is consistent with what you get from Python:
In [2]: import numarray
In [3]: a = numarray.ones(10,type=numarray.Bool)
In [4]: print a
[1 1 1 1 1 1 1 1 1 1]
In [5]: a * 4096
Out[5]: array([4096, 4096, 4096, 4096, 4096, 4096, 4096, 4096, 4096, 4096])
However, numpy gives the following results:
In [6]: import numpy
In [7]: b = numpy.ones(10,dtype=numpy.bool_)
In [8]: print b
[ True True True True True True True True True True]
In [9]: b * 4096
Out[9]: array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=int8)
The NUMPY coercion rules are not even self consistent. If you extract an element from array b to get a rank zero array you get the same behavior as Python:
In [10]: print b[0]
True
In [11]: b[0] * 4096
Out[11]: 4096
Activity
numpy-gitbot commentedon Oct 19, 2012
@teoliphant wrote on 2007-03-26
This should be fixed in SVN, r3601
The scalar/array coercion rule is that the type of the scalar will be ignored if the scalar is of the same kind or a lesser kind than the array.
The recognized kinds are (in order)
BOOL
POSINT
NEGINT
FLOAT
COMPLEX
OBJECT