Revisiting naming conventions
Miller, James V (CRD)
millerjv at crd.ge.com
Tue Apr 18 08:46:22 EDT 2000
I guess the point is that BOTH stl and the proposed matrix library use the same convention which is
different from ours. Furthermore the stl convention is part of the ANSI standard.
However, I am not a big fan of the stl naming convention, but I can see a benefit to being
consistent.
-----Original Message-----
From: Ross Whitaker [mailto:whitaker at rolle.engr.utk.edu]
Sent: Monday, April 17, 2000 7:03 PM
To: Insight-Developers (E-mail)
Subject: Re: Revisiting naming conventions
******************
The question:
"The STL naming convention is different from ours. Should we adopt it instead?"
******************
My first thoughts:
Suppose we do. Then suppose we adopt yet another external library---such as a matrix library.
Suppose
it has another naming convention. Do we adopt it? What if they conflict? Then it's impossible.
It seems we should either:
1) Live with different naming conventions for borrowed libraries.
2) Subclass heavily used libraries/objects and make them use our naming convention.
The first one seems cleaner/easier.
Regards,
Ross
******************
"Miller, James V (CRD)" wrote:
> At the January meeting, we decided on a coding style where class names were prefixed with itk and
> used word capitalization, i.e. our classes are named itkImage, itkImageBase, etc.
>
> To support smart pointers, each class has an internal typedef named "Pointer". So to declare a
> smartpointer to an itkImage you would use itkImage::Pointer.
>
> To support generic programming and the user of iterators, the image class has typedefs
> itkImage::Iterator, itkImage::Index and itkImage::PixelType.
>
> We are using two external "packages", STL and a numerics package, that use a different
nomenclature.
> These packages use lower case class names and underscores. For instance, the stl vector class has
a
> typedef vector::interator and vector::value_type and vector::reference_type and vector::pointer.
>
> Therefore, itk's public interface uses one naming convention; however, under the hood, we reference
> systems that have a very different naming convention.
>
> Should we consider adopting the naming conventions of STL and the numerics package for consistency
> sake?
>
> There are two arguments on the table.
>
> 1. Using our original planned naming convention, it will be very clear what is itk code and what is
> STL/numerics code.
>
> 2. Using a consistent naming convention will allow people who already know STL to get up to speed
on
> itk quickly (and they will not have to remap STL/itk bindings).
>
> If we adopt STL conventions, how far should we take it? i.e. should our classnames take the form of
> itk_image instead of itkImage?
>
> We have a small enough code base now, that we can change our conventions easily. If we wait too
much
> longer, it will become VERY painful to switch.
>
> What are people's opinions on adopting the STL naming convention (which is part of the ANSI
> standard).
>
> Jim Miller
> _____________________________________
> Computer Graphics & Systems Program
> GE Corporate Research & Development
> Bldg. KW, Room C218B
> P.O. Box 8, Schenectady NY 12301
>
> millerjv at crd.ge.com <mailto:millerjv at crd.ge.com>
> (518) 387-4005, Dial Comm: 8*833-4005,
> Cell: (518) 505-7065, Fax: (518) 387-6981
>
> <<James Miller (E-mail).vcf>>
More information about the Insight-developers
mailing list