Re: [boost] Boost.Algorithm design question

Subject: Re: [boost] Boost.Algorithm design question
From: Phil Endecott (spam_from_boost_dev_at_[hidden])
Date: 2011-10-31 13:08:40

Peter Dimov wrote:
> Andrew Sutton wrote:
>> > None of this has anything to do with the fact that you want to require
>> > operator== to be an equivalence relation, and it doesn't have to be.
>>
>> Realistically though, when is == not (at least) an equivalence
>> relation?
>
> Realistically, and in this context, when the two sides are not of the same 
> type. 

In most of the cases that I care about, there is some underlying type 
like "string" or "integer", on which an equivalence relation exists in 
some formal sense.  Then there are some concrete C++ types like 
std::basic_string<char,allocator_1> or 
std::basic_string<char,allocator_2> or int32_t or int64_t.  operator== 
is defined on pairs of these concrete types in some way that 
approximates to the equivalence relation on the underlying formal type, 
but with some inevitable flakiness at the edges, such as comparison 
between two char*s or comparison between integers with different 
numbers of bits.

The difficulty, which is perhaps what Dave is grappling with, is how to 
specify an algorithm that does the "right thing" if-and-only-if the 
types do the "right thing", while remaining vague about what the "right 
thing" means to allow the implementation flexibility.

Regards,  Phil.