Braces are not Compiler-Fluff.

[T. William Wells]

In article <849 at unh.UUCP> jeff at unh.UUCP (Jeffrey E. F. Friedl) writes:
:
: In article <272 at twwells.uucp> bill at twwells.uucp (T. William Wells) writes:
: >  I believe ...
: >  that braces exist only to make the compiler happy and should
: >  therefore be as unobtrusive as possible.

:     It's been said that "indentation is enough to show the
: block structure". Sure, but taking this idea to an extreme,
: one could say that having all the pages in a book is sufficient
: to determine what page you're on -- one doesn't need page numbers.

Drop using analogies. They are never proper means of demonstration.
Not only that, but yours does not even succeed as an illustration.
It's completely wrong, there being no relationship between the
purpose of page numbers in a book and the purpose of indentation or
braces in a program. The page numbers are in the book because one
wants absolute information, to be used without reference to the
surrounding pages; the braces or indentation are there to indicate
relationships. The two are not comparable.

: >  [braces glued to a control structure are] sometimes necessary for
: >  the compiler (but I always use the braces), but contributing little or
:                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^      ^^^^^^^^^^^^^^^^^^^^^^^^^
: >  nothing to the understandability of the program.  The indentation of
:   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
: >  the program is what tells you the control relationships; the braces
: >  are redundant information, at best they can augment what the
: >  indentation is already telling you.
:
:     If they contribute little or nothing, why put them when you don't
: need them (as you also said you do)?

They are there for the *writer* of the code. They help prevent
mistakes.  Also, closing braces provide continuity in the indentation
so that determining what the indentation actually is becomes an easier
perceptual task.

:                     It's just easier to see it right there (by
: the page numbers, or the braces -- both of which augment what you're
: already able to find determine).

But you have forgotten or don't know an important fact: to understand
the code, using braces, one must *conceptually* identify the
relationship between the parts.  This takes time. This also takes
concentration and part of one's short term memory.  Indentation, on
the other hand, relies on one's perceptual systems; this is
automatic, takes essentially no time, and costs nothing in terms of
concentration and short term memory.

:     The only reason I say any of this is that if style is going
: to be discussed in a rational fashion (i.e. not a Religious FlameFest)
: then I think it's better to discuss braces from the point of view of
: "how can I use these braces to make things more readable" rather than
: from the point of view of "where can I stash these things".

But that is an incorrect point of view: before asking "how can I
use..." one *must* ask "ought I use...". And in this case, the answer
is clear: braces, given the existence of indentation, are the wrong
solution and should not be used for that function.  It is therefore
purposeless to ask how they should be used for that function.

To summarize: there are (at least) two possible ways to indicate
control relationships in a program, indentation and braces.  They are
equivalent, in the sense that the braces can always be used to
generate the proper indentation and the indentation can always be
used to generate the proper braces.  Thus, the choice between them is
between which conveys the information most readily.  Since
indentation does so with little cost to the reader, but braces with a
large cost, the choice is clear: use indentation to convey the
information. Now, for understandability, braces have little use other
than to convey this control information (the exception being that
closing braces provide continuity of indentation, an argument for
always using braces with multiple levels of indentation) this implies
that braces should be used for their only essential function: telling
the compiler what the control relationships are, and should otherwise
be made as unobtrusive as possible.

This, among style issues, is one of the few that is actually
decidable on objective grounds. Arguing it, without reference to
those objective grounds is pointless.

---
Bill
{ uunet!proxftl | novavax } !twwells!bill