Re: [boost] [lambda] better switch syntax

From: Joel de Guzman (joel_at_[hidden])
Date: 2003-10-17 22:57:44

Powell, Gary <powellg_at_[hidden]> wrote:
> Actually,
> 
>     switch_(condition)
>     [
>         case_(label1)[expr1],
>         case_(label2)[expr2],
>         case_(label3)[expr3]
>         default_[expr4]
>     ]
> ----------------------------------
> has problems with "default_[expr4]"

Sorry, typo--missing comma :o"

> but I think
> 
>     switch_(condition)
>     [
>         case_(label1)[expr1],
>         case_(label2)[expr2],
>         case_(label3)[expr3]
>         .default_[expr4]
>     ]
> 
> or
>     switch_(condition)
>     [
>         case_(label1)[expr1],
>         case_(label2)[expr2],
>         case_(label3)[expr3],
>         default_(expr4)
>     ]
> 
> could be made to work.
> 
> The nice thing about having .default is that it could end the sequence of
> case_()[]'s by generating a unique expression template which would not allow
> either a operator,(expression) or operator,(case_expression).  

I prefer the comma, for consistency. Also, I think this should be
legal (as it is in plain C++):

    switch_(condition)
    [
        default_(expr4),
        case_(label1)[expr1],
        case_(label2)[expr2],
        case_(label3)[expr3]
    ]

> what you also want to avoid is allowing:
> 
>     switch_(condition)
>     [
>         case_(label1)[expr1],
>         expr1a,
>         case_(label2)[expr2],
>         case_(label3)[expr3]
>         .default_[expr4]
>     ]
> 
> or
> 
>     switch_(condition)
>     [
>         case_(label1),
>         expr1,
>         case_(label2)[expr2],
>         case_(label3)[expr3]
>         .default_[expr4]
>     ]
> 
> or
>     switch_(condition)
>     [
>         case_(label1),
>         expr1,
>         case_(label2)[expr2]
>        .default_[expr4],
>         case_(label3)[expr3]
> 
>     ]
> 
> where a useless expression is inserted into the middle of the case statements.

Right. Those should be illegal.

> BTW, Jaakko and I did most of this work a while ago, playing around, but if
> Phoenix-2 the merger of LL and Phoenix is due out, there is no point in
> pushing updates to LL.  

Cool! I'll reply to this in another email and let's discuss this off list....

Cheers,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net