Re: complex number API

[Mark Galassi <rosalia@cygnus.com>]

    >> >One shortcoming I see is that gsl_complex_t is longer.

    Gerard> Now, there is reasonable response. I like to emphasize the
    Gerard> point by saying that, when people start talking about
    Gerard> things being too long to type, I get very worried. I hope
    Gerard> Mark was joking.

Don't worry: I don't care about typing length.  My point was about
screen real estate in the editor :-)

    Gerard> The only qualm I have about '_t' is that (for whatever
    Gerard> perverse reason), I generally associate it with a "simple"
    Gerard> type, the canonical examples being things like size_t,
    Gerard> off_t, etc., which exist as abstractions over built-in
    Gerard> types, essentially for platform-independence. Maybe there
    Gerard> should be another convention for non-simple types
    Gerard> (structs), which makes it clear that they are
    Gerard> objects. Nitpicking, I know.  Does this ever bother
    Gerard> anybody else?

I think that there are many examples of complex types in POSIX that
have _t.  Let me see...  Hmm: I'm drawing a blank :-).  Still, I think
that part of the motivation behind the _t habit in POSIX is to make
the types opaque so that you don't care how they are defined.  I'll
look into it some more.