Re: [Boost-users] multi_index_container

From: Joaquín Mª López Muñoz (joaquin_at_[hidden])
Date: 2006-02-17 02:47:51

Aashit Soni ha escrito:

> Some correction!! in my previous mail.
>
>  Thank you Mr Lopez,
>
> > Slower in which aspect? Insertion, deletion, lookup?
>  I have observed it is slower in insertion. I am roughly creating some
>  200,000 entries into it. One entry is arround 10 Bytes of information. Each
>  creation also creates some external objects to that are being cached
>  in this container.

Well, it's hard to say. Note that if you're comparing a preexistent hash
table with a multi_index_container composed of a hashed plus
an ordered index, the comparison is not fair --the multi-index
container is doing more work. On the other hand, it is perfectly possible
that your previous hash container is just faster, but I don't think it
can be *much* faster. If you can provide me (through the list or privately)
with some test code I can take a deeper look.

>  - I am having iteration problem in hash -look up also. My hash index is
>  non_unique. I am not sure how the collisions are handled?

non_unique means equivalent elements (elements with the same key) are
allowed into the container.
In the particular case of hashed indices, equivalent elements are guaranteed
to be adjacent.

> When i call
>  find and if the collisions are there i get the last inserted node,
> how shall i get previously inserted nodes.

Use equal_range() to get all of them. As explained above, equivalent
elements are kept adacent.

>  In some case, i need to reserve  the space for the elements, So when i
>  create my multi_index container, i want to tell it saying that i need
>  x amount of space so please do the allocation and keep it ready. Is
>  such facility available.

No such facility, sorry. This preallocation stuff is not in the spirit
of node-based STL containers. You might gain some performance
by using custom allocators specialized for small node requests:
take a look for instance at boost::fast_pool_allocator:

http://boost.org/libs/pool/doc/interfaces/pool_alloc.html

>  I will just incorporate- your sugguesion and check how it is working.
>
>  Thank you once again, I really appreciate your time.
>  regards,
>  aashit

Good luck,

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo