[Numpy-tickets] [NumPy] #228: Backward compatibility plan

NumPy numpy-tickets at scipy.net
Fri Aug 4 14:16:18 CDT 2006


#228: Backward compatibility plan
----------------------+-----------------------------------------------------
 Reporter:  oliphant  |        Owner:  oliphant   
     Type:  task      |       Status:  assigned   
 Priority:  normal    |    Milestone:  1.0 Release
Component:  Other     |      Version:             
 Severity:  normal    |   Resolution:             
 Keywords:            |  
----------------------+-----------------------------------------------------
Old description:

> The plan for compatibility with Numeric/Numarray is
>
> 1) Create functions in oldnumeric and numarray that create and return
> NumPy arrays but do it using the syntax of the old packages (i.e.
> typecode=, savespace=).
> Also the name-spaces when doing from numpy.oldnumeric import * should
> only include names that were in Numeric or Numarray already:
>
> 2) Create 2 scripts for Numeric conversion and 2 for Numarray conversion
>
>    a) The first conversion script just changes the imports to the
> compatibility
>       layer and alters attribute and method access to be compatible with
>       NumPy arrays. This script should work most of the time.
>
>    b) The second conversion script tries to actually make the changes to
> the code
>       so it will work without the compatibility layer.  This script will
> be
>       harder to get to work perfectly and will probably require some
>       manual help from the user.

New description:

 The plan for compatibility with Numeric/Numarray is

 1) Create functions in oldnumeric and numarray that create and return
 NumPy arrays but do it using the syntax of the old packages (i.e.
 typecode=, savespace=). Also the name-spaces when doing from
 numpy.oldnumeric import * should only include names that were in Numeric
 or Numarray already:

 2) Create 2 scripts for Numeric conversion and 2 for Numarray conversion

    a) The first conversion script just changes the imports to the
 compatibility layer and alters attribute and method access to be
 compatible with NumPy arrays. This script should work most of the time.

    b) The second conversion script tries to actually make the changes to
 the code so it will work without the compatibility layer.  This script
 will be harder to get to work perfectly and will probably require some
 manual help from the user.

-- 
Ticket URL: <http://projects.scipy.org/scipy/numpy/ticket/228>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.


More information about the Numpy-tickets mailing list