[multi-index] Key extractor for 1:n relationship? - Boost-users - lists.preview.boost.org

newer
[Multi-Index] has_index<BMI,TAG>...

[multi-index] Key extractor for 1:n relationship?

older
Using vectors of ptrs

Boris Schaeling

17 Jul 2009 17 Jul '09

3:06 a.m.

All the key extractors in the documentation of Boost.MultiIndex return one value for an element in a container. Is it somehow possible to use a 1:n relationship between element and value? The documentation uses this container definition (among others): typedef multi_index_container< employee, indexed_by< hashed_unique< const_mem_fun >

employee_set;

Let's assume an employee can have more than one name (std::string becomes std::vectorstd::string). Now if someone searches for Tom or John the same employee should be returned by the container. Is this possible? Boris

Reply

Sign in to reply online Use email software

Show replies by date

plarroy

17 Jul 17 Jul

5:12 a.m.

Boris Schaeling wrote:

All the key extractors in the documentation of Boost.MultiIndex return one value for an element in a container. Is it somehow possible to use a 1:n relationship between element and value?

The documentation uses this container definition (among others):

typedef multi_index_container< employee, indexed_by< hashed_unique< const_mem_fun >

...
employee_set;

Let's assume an employee can have more than one name (std::string becomes std::vectorstd::string). Now if someone searches for Tom or John the same employee should be returned by the container. Is this possible?

A vector of strings is still an object, and hashed_unique indexes are unique. For indexes that allow duplicate items you want "equal_range" like: typedef pmt_t::index<pid>::type pid_idx_t; pair pid_match = pid_idx.equal_range(50); while(pid_match.first != pid_match.second ) { } Regards. Pedro.

Boris

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Reply

Sign in to reply online Use email software

Boris Schaeling

12:15 p.m.

On Fri, 17 Jul 2009 07:12:00 +0200, plarroy wrote:

Boris Schaeling wrote:

...
All the key extractors in the documentation of Boost.MultiIndex return one value for an element in a container. Is it somehow possible to use a 1:n relationship between element and value?

The documentation uses this container definition (among others):

typedef multi_index_container< employee, indexed_by< hashed_unique< const_mem_fun >

...
employee_set;

Let's assume an employee can have more than one name (std::string becomes std::vectorstd::string). Now if someone searches for Tom or John the same employee should be returned by the container. Is this possible?

A vector of strings is still an object, and hashed_unique indexes are unique.

Either you misunderstand me or I misunderstand you. :) The hash value would be created for the entire vector? I want to use a string in the vector as a key, not the entire vector itself. If an employee has the names Tom and John I want to find the employee also if I only search for Tom. There would be one employee object in the container but two hash values would need to be generated. Not sure if this is possible at all? Boris

Reply

Sign in to reply online Use email software

Ovanes Markarian

1:54 p.m.

I remember to ask a similar question. I was told that it is not supported. Finally I reconsidered my design, which was more optimal at all. On the other hand, it is not said that hash works optimal in that case. You might better do a lexicographical_compare to the keys and find first one of the keys from the set, as to hash all and look-up for hash values. Therefore a multi-map or multi-set as part of your multi-index might be a better choice. Otherwise, you need to hash all possible tuple combinations, which is an NP-complete problem. There is a nice article in Dr.Dobb's on what is more efficient hash based container or stl::set or map. http://www.ddj.com/cpp/198800559?pgno=1 Hope that helps, Regards, Ovanes On Fri, Jul 17, 2009 at 2:15 PM, Boris Schaeling wrote:

On Fri, 17 Jul 2009 07:12:00 +0200, plarroy wrote:

Either you misunderstand me or I misunderstand you. :) The hash value would be created for the entire vector? I want to use a string in the vector as a key, not the entire vector itself. If an employee has the names Tom and John I want to find the employee also if I only search for Tom. There would be one employee object in the container but two hash values would need to be generated. Not sure if this is possible at all?

Boris

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Reply

Sign in to reply online Use email software

Joaquin M Lopez Munoz

2:14 p.m.

Boris Schaeling writes:

All the key extractors in the documentation of Boost.MultiIndex return one value for an element in a container. Is it somehow possible to use a 1:n relationship between element and value?

The documentation uses this container definition (among others):

typedef multi_index_container< employee, indexed_by< hashed_unique< const_mem_fun >

...
employee_set;

Let's assume an employee can have more than one name (std::string becomes std::vectorstd::string). Now if someone searches for Tom or John the same employee should be returned by the container. Is this possible?

No, I don't think it's possible in general. Take into account that when you insert an element x into a multi_index_container m, then x will appear exactly once when traversing any index of m. So if you want something like multiple retrieval within one index you'd have to insert x multiple times. I'm afraid you'll need a separate container to hold this 1:n relationship. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

Reply

Sign in to reply online Use email software

5600

Age (days ago)

5600

Last active (days ago)

Download

4 comments

4 participants

tags

participants (4)

Boris Schaeling
Joaquin M Lopez Munoz
Ovanes Markarian
plarroy