I'm considering changing the semantics of operator[] in stlsoft::frequency_map to return 0, rather than throwing std::out_of_range, if a given key does not exist in the map.

stlsoft::frequency_map<std::string>  fm;

fm.push("abc");
fm.push("abc");
fm.push("def");

assert(2 == fm["abc"]);
assert(1 == fm["def"]);
assert(0 == fm["ghi"]); // **

Currently, statement ** will throw an instance of std::bad_alloc

Does anyone have a problem with / an opinion about this projected change?

Cheers

Matt

Matthew Wilson Wrote:

> I'm considering changing the semantics of operator[] in stlsoft::frequency_map to return 0, rather than throwing std::out_of_range, if a given key does not exist in the map.
> 
> stlsoft::frequency_map<std::string>  fm;
> 
> fm.push("abc");
> fm.push("abc");
> fm.push("def");
> 
> assert(2 == fm["abc"]);
> assert(1 == fm["def"]);
> assert(0 == fm["ghi"]); // **
> 
> Currently, statement ** will throw an instance of std::bad_alloc
> 
> Does anyone have a problem with / an opinion about this projected change?

imo it is an improvement to the semantics. 0 is never going to be valid for an element in the fm, so there's no problem.

Jim