[Insight-developers] Dashboard Cleanup: mini0.nlm MacOSX-cross-rosetta : 50 tests failing

[Luis Ibanez]

I would certainly prefer to use "while" loops
instead of "for" loops, whenever possible.


A "for" loop compacts three statements in a
single line, and becomes a nightmare for debugging.


      Luis


---------------------------------------------------------------------
On Thu, Jan 28, 2010 at 2:16 PM, Michel Audette
<michel.audette at kitware.com> wrote:
> Hi Kevin,
>
> I added a pair of brackets that makes d local. I only alerted Luis, because
> I thought that would almost spam the list. In any event, I wanted to
> preserve the functionality as much as possible, although Luis is steering me
> to while loops these days, as they tend to debug well. I'll keep an eye on
> the dashboard.
>
> Cheers,
>
> Michel
>
> On Thu, Jan 28, 2010 at 1:59 PM, Kevin H. Hobbs <hobbsk at ohiou.edu> wrote:
>>
>> On 01/28/2010 01:31 PM, Michel Audette wrote:
>> > Hi Luis,
>> >
>> > your value *d* in the bolded part of your suggested code is undefined
>> > when used as an index. It has to be as
>> > d = dim - 1;
>> > l = coord[d];
>> > d--;
>> > for coord to have a valid index, doesn't it?
>> >
>> > I propose... (extra check on validity of index d)
>> > #define NRRD_INDEX_GEN(I, coord, size, dim)   \
>> > int d;                                                                 \
>> > d = (dim) - 1;                                                      \
>> > if ( (d) >= 0 )                                                      \
>> > {
>> > \
>> >   (I) = (coord)[d];                                                \
>> >   d--;                                                                 \
>> >   while( d >= 0 )                                                 \
>> >   {
>> > \
>> >   (I) = (coord)[d] + (size)[d] * (I);                          \
>> >   d--;                                                                 \
>> >   }
>> > \
>> > }
>> > \
>> > and to take out the pragmas.
>> >
>> > Seem reasonable?
>> >
>> > Michel
>> >
>>
>> I like your check for d >= 0. Nothing will happen if dim = 1 where I'm
>> guessing the present code would do something undefined if dim == 1. Does
>> NRRD even deal with 1-d things?
>>
>> The only other thing that jumps to mind is that since this is a macro in
>> C we might want to follow the lead of the original header and wrap the
>> whole thing up in a do{...}while(0) I guess this is to avoid redefining
>> d or declaring d in the middle of a block of code. This will probably
>> make the code easier to digest for older compilers but I'm by no means
>> an expert here.
>>
>> Instead of do{...}while(0) could it be if ( dim >= 1 ){...} else {...} ?
>>
>
>
>
> --
> Michel Audette, Ph.D.
> R & D Engineer,
> Kitware Inc.,
> Chapel Hill, N.C.
>
>