[Gambas-user] gb.data: Trie class

[Tobias Boege]

On Wed, 24 Sep 2014, Beno?t Minisini wrote:
> Le 23/09/2014 21:47, Tobias Boege a ?crit :
> > Also, Benoit: tell me what you think about the interface. I documented
> > everything in the source code (c_trie.c).
> >
> 
> The interface seems perfect. It's just the name "Trie" that I find ugly. :-)
> 

I pronounce it like "try". RadixTree or PrefixTree would be options for the
container -- nothing against that -- but what about the other class?
RadixTreePrefix or even PrefixTreePrefix? :-) Naming it Patricia (from the
acronym) would be fun but even more confusing, right?

> Another point on the implementation: you should not use malloc(), free() 
> and realloc(), but the functions provided by the interpreter API, i.e. 
> GB.Alloc(), GB.Free() and GB.Realloc().
> 
> They are usually faster, especially if you allocate small chunks of data.
> 
> Moreover, they can detect memory leaks by checking that every allocation 
> has been freed at the end of the program.
> 
> If you can compare with malloc() easily, don't hesitate not to trust me 
> and check. :-)
> 

The trie.c and trie.h are from another library I wrote. (That's why I
double-tested the projects with valgrind.) It should not be a problem
to convert the uses of malloc()/free() now.

Regards,
Tobi

-- 
"There's an old saying: Don't change anything... ever!" -- Mr. Monk