Re: [Boost-users] Boost.Interprocess accessing an unordered_map in shared memory.

Subject: Re: [Boost-users] Boost.Interprocess accessing an unordered_map in shared memory.
From: lazylabs_at_[hidden]
Date: 2013-03-12 09:48:30

On Tue, Mar 12, 2013 at 6:49 PM, Nat Linden <nat_at_[hidden]> wrote:
> On Tue, Mar 12, 2013 at 12:43 AM, <lazylabs_at_[hidden]> wrote:
>
>> On Tue, Mar 12, 2013 at 2:21 AM, Ion Gaztañaga <igaztanaga_at_[hidden]>
>> wrote:
>
>> > El 11/03/2013 13:20, lazylabs_at_[hidden] escribió:
>>
>> >> However, I would prefer to do this with a std::string, which results
>> >> in compilation errors:
>> >>
>> >> persons->at(std::string("H")).print();
>> >>
>> >> Is it possible to write the above statement, i.e. accessing the map
>> >> allocated in shared memory with std::string?
>
>> > No, as the key_type is not std::string. It's a bit ugly but to fix this
>> > has no easy fix.
>
>> As I understand, the reason why we cannot lookup ShmemString with
>> std::string is because of the template argument, which I have given
>> while creating the unordered_map:
>>
>>                 std::equal_to<KeyType>,
>>
>> Just guessing, if something can be done here?
>
> It seems to me that within the last year or so another person on this
> mailing list (who had his own local idiosyncratic string type) was
> asking about passing std::string or const char* key arguments for map
> lookup. I don't even think it was specific to Interprocess. I believe
> he was proposing string comparators accepting range-of-characters
> rather than a specific string type.
>
> Sorry for vagueness; I don't have time to search for it myself...

Sorry if I was vague, I was more interested in "why" this cannot be
done. I already got a No (and that too from the library implementer
himself. Thanks) and based on that I have already implemented a
different solution in my code, to have integer lookups, rather than
string-based lookups.

Thanks.