concatenate Numeric

[Robert Kern]

Sheldon wrote:
> Hi,
> 
> I am trying to build a large array using concatenate function in
> python.
> So as I loop over the number of arrays, of which there are 12 (4 down
> and 3 across), I create 3 long arrays by concatenating them at the
> bottom and then concatenating them side by side:

[snip]

> print 'Max lat', max(max(ppslat_all)), '\t','Min lat',
> min(min(ppslat_all))
> print 'Max lon', max(max(ppslon_all)), '\t','Min lon',
> min(min(ppslon_all))
> 
> *****************
> Now this works, the array size is correct but the longitude values
> given for max and min are wrong. What is even stranger to me is that
> when I write the array in binary format to a file and read it with
> Matlab, the max and min are correct but when I read it back with python
> the max and min are again incorrect for only the longitude data. I
> saved the max and min for the longitude for each array and then check
> it in the python program and they are correct at the end but the
> max(max(ppslon)) values is incorrect.  Does anyone knows why this is
> so?
> If I was doing something wrong then Matlab would not have returned
> correct values.

Don't use min() and max() on multidimensional arrays. They won't give sensible 
answers.


In [11]: a = RA.random([3,5])

In [12]: a
Out[12]:
array([[ 0.01721657,  0.64291363,  0.33210659,  0.89887972,  0.24437849],
        [ 0.88205348,  0.00839329,  0.35999039,  0.9966411 ,  0.54957126],
        [ 0.59983864,  0.18983323,  0.13727718,  0.8987289 ,  0.05425076]])

In [13]: min(a)
Out[13]: array([ 0.59983864,  0.18983323,  0.13727718,  0.8987289 ,  0.05425076])


The builtin min() and max() compare the values in the sequence. In this case, 
those values are the rows of the arrays. Numeric uses rich comparisons, so the 
result of a comparison is a boolean array. Numeric also has the convention that 
if any of the elements of an array are considered to be True, then the whole 
array is.


In [16]: a[1] < a[2]
Out[16]: array([0, 1, 0, 0, 0])

In [17]: bool(_)
Out[17]: True

In [18]: a[2] < a[1]
Out[18]: array([1, 0, 1, 1, 1])

In [19]: bool(_)
Out[19]: True


This makes min(a) incorrect when len(a.shape) > 1. Instead, use the minimum and 
maximum ufuncs provided with Numeric:


In [21]: N.minimum.reduce(a.flat)
Out[21]: 0.0083932917161983426

In [22]: N.maximum.reduce(a.flat)
Out[22]: 0.99664110397663608


-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
  that is made terrible by our own mad attempt to interpret it as though it had
  an underlying truth."
   -- Umberto Eco