Re: [boost] [proto] construct_: typos in docs and not able to get this to work

From: Eric Niebler (eric_at_[hidden])
Date: 2008-03-13 14:15:57

Hi Markus,

Markus Werle wrote:
> Having hard times to catch the ideas of the construct_ example.
> 
> First of all: There are 2 typos in the definition of construct:
> 
<snip>

Sorry for the typos! Thanks for catching them.

> To keep it off the other thread that I will continue later this day:
> 
> Could you please remark in the docs with big red letters that make_expr
> is a type in namespace proto::functional and a function in namespace 
> proto and why name clashes won't appear?
> 
> I do not like that naming convention, mainly because I cannot remember the 
> rules for the compiler for cases where both are visible.
> And yes, if you remember, I like to learn them again - now.

If you bring them both into scope, the names will clash. There shouldn't 
be a need to ever bring the proto::functional namespace into scope, so 
there shouldn't be a problem.

My approach to free functions in Proto is as follows: For each free 
function template proto::X, there is a class template 
proto::result_of::X which calculates the result type of proto::X. (Here 
I'm following existing practice established by Fusion.) There is also a 
class or class template proto::functional::X which is the function 
object equivalent of proto::X. These are provided as a convenience. I've 
had enough occasions to need them myself that I decided to provide them.

> Also for beginners like me it would be nice to have struct construct_ be 
> renamed to constructor. The '_' is easily overlooked at midnight 
> when you print 2 pages on one as I did. 
> And that text is not for the faint-hearted anyway.

Noted.

> Note that I'd prefer code repetition in this example: 
> please add definitions for _1 and _2, better: a complete, compilable
> example, see next section:

Yes, I'll build a complete example around this.

> Trouble
> --------
> 
> Below is code that does not compile. I get the impression that
> this led to the confusion in the other thread: I cannot see how
> proto::eval drops in in the std::transform call.
> Please correct the code in order to make it work.

In the construct() section in the docs I say, "Making it work with 
std::transform() takes a little more effort, but that's covered in the 
Calc2 example," and I provide a link to the example. Getting 
std::transform() to invoke proto::eval() involves proto::extends<>, 
which I haven't talked about yet at that point.

A better approach would be to provide a complete working construct() 
example and provide a link to that. In that spirit, I have added a 
Lambda example that shows a working construct() in action. HTH.

-- 
Eric Niebler
Boost Consulting
www.boost-consulting.com