On Friday, 30 January 2015 at 11:55:16 UTC, Laeeth Isharc wrote:
Hi.
The standard advice is not to worry about memory usage and
execution speed until profiling shows you where the problem is,
and I respect Knuth greatly as a thinker.
Still, one may learn from others' experience and cultivate good
habits early. To say that one should not prematurely optimize
is not to say that one should not try to avoid cases that tend
to be really bad, and I would rather learn from others what
these are then learn only the hard way.
For the time being I am still at early development stage and
have not yet tested things with the larger data sets I
anticipate eventually using. It's cheap to make slightly
different design decisions early, but much more painful further
down the line, particularly given my context.
As I understand it, foreach allocates when a simple C-style for
using an array index would not. I would like to learn more
about when this turns particularly expensive, and perhaps I
could put this up on the wiki if people think it is a good idea.
What exactly does it allocate, and how often, and how large is
this in relation to the size of the underlying data
(structs/classes/ranges)? Are there any cache effects to
consider? Happy to go to the source code if you can give me
some pointers.
Thanks in advice for any thoughts.
Laeeth.
foreach() is good for linked list because it uses a delegate so
it avoids to cross the items from 0 to i at each iteration.
for() is good for arrays because of the pointer arithmetic.
But the D way for foreach'es is to use ranges: popFront etc,
although i often feel too lazy to use them...
You'll probably get more technical answers...
