Type of lambda function returning a lambda function...
 

.... I'm trying to declare a lambda function returning a lambda function, and, unfortunately, I could not use auto key word since it would be the return type of my function. The lambda function looks like this:

typedef std::function<void(string)> Func;
vector<pair> set;
auto lambda = [&set] (string cmd, Func func) {
set.push_back(pair(cmd, func));
return lambda;
};
return lambda;

I try to use std::function template to express that, but.. As you see, an infinite recursion occurs...

So how could I do that? Any one have any idea?

BTW, the above code compiles without the last returning line. So I CAN have a lambda function like that.

---------------------
Posted by Haochen Xie

Re: Type of lambda function returning a lambda function...
 

Haochen Xie於 2013年3月17日星期日UTC+8上�1時45分42秒 寫�：
> ... I'm trying to declare a lambda function returning a lambda function, and, unfortunately, I could not use auto key word since it would be the return type of my function. The lambda function looks like this:
>
>
>
> typedef std::function Func;
>
> vector set;
>
> auto lambda = [&set] (string cmd, Func func) {
>
> set.push_back(pair(cmd, func));
>
> return lambda;
>
> };
>
> return lambda;
>
>
>
> I try to use std::function template to express that, but.. As you see, aninfinite recursion occurs...
>
>
>
> So how could I do that? Any one have any idea?
>
>
>
> BTW, the above code compiles without the last returning line. So I CAN have a lambda function like that.
>
>
>
> ---------------------
>
> Posted by Haochen Xie

In C/C++ to implement a function which could map an input
function into another in a 32/64 bit OS is different but similar.

What really matters is the type of the functions in the domain and
the co-domain under implementations.

input output
1. static fixed compiled codes 1,2,3
2. JIT compiled byte codes of a VM 1,2,3
3. JIT compiled machine codes 1,2,3

Now there are 9 possibilities in different cases.

I'll suggest you learn the decorator part in python first.

Re: Type of lambda function returning a lambda function...
 

On Saturday, March 16, 2013 11:17:43 AM UTC-7, 88888 Dihedral wrote:
>
> In C/C++ to implement a function which could map an input
> function into another in a 32/64 bit OS is different but similar.
>
> What really matters is the type of the functions in the domain and
> the co-domain under implementations.
>

Thank you for your reply, but actually I have no idea what you are talking about...

> I'll suggest you learn the decorator part in python first.

May I assume that you tried to answer in python? If my post is confusing, I apologize. But I'm talking about C++, not python. Please point out if I'm wrong.

---------------------
Posted by Haochen Xie

Re: Type of lambda function returning a lambda function...
 

On Saturday, 16 March 2013 19:45:42 UTC+2, Haochen Xie wrote:
> ... I'm trying to declare a lambda function returning a lambda function, and,
> unfortunately, I could not use auto key word since it would be the return
> type of my function. The lambda function looks like this:
>
> typedef std::function<void(string)> Func;
> vector<pair> set;
> auto lambda = [&set] (string cmd, Func func) {
> set.push_back(pair(cmd, func));
> return lambda;
> };
> return lambda;

You are trying to make infinite recursion that does terminate itself somehow?
No way. Maybe in some functional language it is fine but in C++ the world
is finite thing. Different religion ... you see. All unterminated recursions and
infinite cycles are undefined behavior and belong to world of inefficient
dreamers and such.

Re: Type of lambda function returning a lambda function...
 

On Saturday, March 16, 2013 12:26:34 PM UTC-7, Öö Tiib wrote:

> You are trying to make infinite recursion that does terminate itself somehow?
> No way. Maybe in some functional language it is fine but in C++ the world
> is finite thing. Different religion ... you see. All unterminated recursions and
> infinite cycles are undefined behavior and belong to world of inefficient
> dreamers and such.

Nope. That is not a loop at all. All I want to do is to simulate some code like this without defining a new type (code also available at http://pastebin.com/5NBAiwCZ):

typedef std::function<void(std::string)> Func;
class Menu
{
public:
struct Entry
{
std::string cmd;
Func func;
};
const class AppendHelper
^^^^^^^^^^^^ <- the helper type
{
public:
AppendHelper(Menu &menu) : menu(menu) {}
const AppendHelper &operator()(std::string cmd, Func func) const
{ menu._command_set.push_back({cmd, func}); return *this; }
private:
Menu &menu;
} append_menu = {*this};

// For debugging
void list_commands()
{
for(Entry e : _command_set) {
printf("%s\t: %X\n", e.cmd.c_str(), &e.func);
}
}

private:
std::vector<Entry> _command_set;
};

So that I can have the following code:

Menu menu;
menu.append_menu
("hehe", [] (std::string input) {printf("hehe");})
("haha", [] (std::string input) {printf("haha");});

I want to try to use lambda function to do this, but that seems to be impossible... Well, anyway, that should be very hard (if possible) in a strong typed language like c++.

On Saturday, March 16, 2013 2:15:54 PM UTC-7, Andy Champ wrote:
>
> Back to your code... your lambda won't get instantiated until the second
> return statement. Most likely the compiler doesn't see the problem until
> then.

Yeah, that's probably why the code compiled for me when I tried yesterday. But, even without the returning line, it doesn't compile now... I should have got something wrong.

Thank you guys anyway. It's always fun to explore how dirty you can get yourself with c++ ;)