[SciPy-dev] records and new NOTSWAPPED changes

Travis Oliphant oliphant.travis at ieee.org
Wed Dec 14 17:16:48 CST 2005


Hey Chris,

The recent NOTSWAPPED changes were once again prompted by closely trying 
to understand records.  I need to modify the behavior of the byteorder 
argument to the records constructors and go back and make sure things 
are still working.  My thinking is that if the byteorder argument is 
given it will force a particular byteorder on all the descriptors 
(byte-order can also be indicated on the types themselves however):

e.g. '>i4,(2,3)>f8,a10'

so

formats = 'i4, (2,3)f8, a10' with byteorder = 'big'

would be equivalent to

formats = '>i4, (2,3)>f8, a10'

Sound OK?

But, there are still some outstanding issues regarding the record arrays 
and the new changes.  I should be able to work them out in the next few 
hours.  But, just to let you know in case you try things and see the 
errors on your new tests (thanks for those...)

Thanks for your help and encouragement.  I really like where we are 
ending up with this...

-Travis





More information about the Scipy-dev mailing list