Re: Declaration design preference: type or name first

"James Harris" <james.harris.1@[EMAIL PROTECTED]
> wrote in message 
news:1193785291.791887.93360@[EMAIL PROTECTED]
> On 30 Oct, 22:29, "cr88192" <cr88...@[EMAIL PROTECTED]
> wrote:
> ...
>> one major factor though, is that of context independent parsing.
>> having the type first almost invariably leads to context dependent 
>> parsing.
>> with the name first, since you already know you are looking at a type, 
>> you
>> can be far more lax with the parsing (and maintain context
independence).
>>
>> example:
>> typedef int foo;
>>
>> ...
>> foo bar;    //only makes sense if the parser knows about foo
>
> Agreed.
>
>> var bar: foo;
>> the parser can easily figure out what is going on, regardless of
whether 
>> or
>> not it knows about foo.
>
> True. One thing I would sometimes like in C is to be able to search
> the source code for declarations but there is no common way to
> identify them. My current syntax suggests/requires a keyword so I am
> using "var" for any variable and "value" for a named or unnamed
> arbitrary constant (not "constant" but that's another topic!).
>

yeah.

I actually identify declarations in my case (for my header generation 
tools), by making it that everything that looks like a declaration, is 
assumed to be one...

lots of special checks are done to deal with the various cases.

for example:
foo *bar(baz a)
{
    ...
}

we know what it is, even if we don't recognize the type names, based
mostly 
on the configuration of the tokens.
my autoheader tools identify declarations in a similar manner (and work, 
most of the time...).


>> but, of course, there are many other strong considerations (such as,
for
>> example, nearly all c-family languages putting the type name first,
...).
>
> Didn't Dennis Ritchie of one of his colleagues design a language long
> after C where he put the name first, then the type, thus reversing the
> order used in C? If he changed to that form I wonder what the reason
> was. Maybe long experience suggested that was a better way to go.....
>

he probably wanted to avoid the context dependence and other 
parsing/syntactic issues associated with putting the type first...

when writing a C compiler, one finds that there are more than a few hidden

pitfals in the language (and it is not so simple of a piece of machinery
as 
it might at first seem...).


>