Re: [groff] hyphenation issues

[Ralph Corderoy]

Hi Keith,

> I was taught to:
>
>   - find the symbol name
>   - look right, for parentheses
>   - look left, token by token, to find the type

Yes, that's a simplification of the `right left' rule that I first saw
in Paul and Gail Anderson's _Advanced C: Tips and Techniques_,
https://amzn.to/2rn2LPv  And with it, signal(3) becomes intelligible.
:-)

    void (*signal(int sig, void (*func)(int)))(int);

>   char const *foo;
> would be read as:
>   - foo is a variable (not function)
>   - foo is a pointer (to data of type not yet established)
>   - foo is constant
>   - foo is a constant pointer to data of type char

I'd read it as

    foo    foo is a
    *      pointer to
    const  const
    char   char

Note how `const' and `char' stand for themselves.

>   const char *foo;
> *unambiguously* reads as:

Well, they're both unambiguous!

>   - foo is a variable (not function)
>   - foo is a pointer (to data of type not yet established
>   - foo is a pointer to data of type char
>   - foo is a pointer to data of type char, which is constant

This is a more clumsy read:

    foo    foo is a
    *      pointer to
    char   a char
    const  that is const

I used your syntax for decades and only switched in the last year after
chatting to a programmer that's always done it the other way since
introducing the Small C compiler into his company to replace assembler;
so, quite a while.  He'd had no external influences so worked out what
he thought was logical.

`const char * const foo' has const come before its thing one time,
`char', and after another, `*', and that seems poor.  Plus the key
thing, `char', is buried a bit.  `char const * const foo' has `const'
consistently come after its thing, and the first thing I read is `char'.

It works consistently.

    $ awk '{s=""; for (i=NF; i>=1; i--) s=s" "$i; print substr(s, 2)}' <<\E |
    > char               foo
    > char const         foo
    > char       *       foo
    > char const *       foo
    > char       * const foo
    > char const * const foo
    > E
    > sed 's/foo/& is a/; s/\*/pointer to/g'
    foo is a char
    foo is a const char
    foo is a pointer to char
    foo is a pointer to const char
    foo is a const pointer to char
    foo is a const pointer to const char
    $

I accept it isn't groff's style;  it wasn't mine until recently.  But it
does seem logical, consistent, and if it wasn't for historic reasons,
probably the sole grammar that C would allow.  (I don't care about C++.)

-- 
Cheers, Ralph.
https://plus.google.com/+RalphCorderoy