std::transform to append to a string

[kvnil]

Cheers,

I've been struggling all day today with this, maybe you could give me some
useful pointers.
I would like to (elegantly) concatenate a string out of all node values in
a map. I have to do it without using boost (as it is forbidden to use it
in our project).

(Forgive me for giving a non-compiling code.)

---- begin ----
map<int,stringm ;
string s;

transform(m.beg in(),m.end(),?? ??, select2nd<int,s tring>());
---- end ----

What would you recommend placing instead of ???? ?
I've an ugly solution where I declare (something like):

vector<stringv;
transform(m.beg in(),m.end(), back_inserterer (v), select2nd<int,s tring>());
and then of course I could build the string out of the vector.

Why can't I do something like:
transform(m.beg in(),m.end(),ba ck_inserter(s.b egin()), select2nd<int,s tring>());

Regards
kev

[Kai-Uwe Bux]

kvnil wrote:

Cheers,

I've been struggling all day today with this, maybe you could give me some
useful pointers.
I would like to (elegantly) concatenate a string out of all node values in
a map. I have to do it without using boost (as it is forbidden to use it
in our project).

(Forgive me for giving a non-compiling code.)

---- begin ----
map<int,stringm ;
string s;

transform(m.beg in(),m.end(),?? ??, select2nd<int,s tring>());
---- end ----

What's wrong with a for-loop?

for ( map<int,string> ::const_iterato r iter = m.begin();
iter != m.end(); ++iter ) {
s.append( iter->second );
}

What would you recommend placing instead of ???? ?

An append_iterator . You can easily write one. Look at some code for
back_inserter and modify it to use append() instead of push_back().

In fact, you could go all the way and write a generic command_iterato r, that
takes a function object upon construction and calls then translates

*out_iter = value;

into

command( value );

I've an ugly solution where I declare (something like):

vector<stringv;
transform(m.beg in(),m.end(), back_inserterer (v), select2nd<int,s tring>());
and then of course I could build the string out of the vector.

Why can't I do something like:
transform(m.beg in(),m.end(),ba ck_inserter(s.b egin()),
select2nd<int,s tring>());

Because the elements in s are characters not strings?
Best

Kai-Uwe Bux

[kvnil]

>... <snip...

>(Forgive me for giving a non-compiling code.)

---- begin ----
map<int,string m;
string s;
transform(m.be gin(),m.end(),? ???, select2nd<int,s tring>()); ---- end
----

What's wrong with a for-loop?

for ( map<int,string> ::const_iterato r iter = m.begin();
iter != m.end(); ++iter ) {
s.append( iter->second );
}

>What would you recommend placing instead of ???? ?

An append_iterator . You can easily write one. Look at some code for
back_inserter and modify it to use append() instead of push_back().

thanks! I'll try and keep everybody posted :)

In fact, you could go all the way and write a generic
command_iterato r, that takes a function object upon construction and
calls then translates

*out_iter = value;

into

command( value );

Nice - but I do not see the merit (no offence in any way - you've tremendously
helped with your previous suggestion). Indeed transform would perform the
*outer=value call,
but won't I be able to achieve the same with std::foreach? It already calls
a functor for each object.
So I would need to implement an adaptable binary function (since the binary
function from <functionrequir est const on functor arguments), bind1st it's
first argument to my appendee (the target string), and then std::foreach

Maybe it would be even possible to somehow adapt string::append (with mem_fun),
the only problem is that during std::foreach on a map I would get back a
pair<int,string and not a string. Don't see yet how to overcome this. Maybe
with select2nd.

Cheers and thanks!