For review: PEP 308 - If-then-else expression

[Erik Max Francis]

Andrew Dalke wrote:

> I like minor points.  They are things I can understand.
>
> So assuming you don't want to do internationalization, you could also
> do this example with
> 
>   print "I have %d %s%s." % (count, noun, "s"[:count!=1])
> or
>   print "I have %d %s%s." % (count, noun, ("", "s")[count!=1])
> or
>   print "I have %d %s%s." % (count, noun, "s" * (count != 1)

Of course you can.  But you wouldn't really suggest that these are
clearer, would you?  (And, of course, a true conditional expression
would be short-circuiting, although that isn't relevant to this example
here, but is one of the big payoffs to a _real_ conditional expression.)

In my opinion, the key issue about the benefit of a conditional operator
is about making code more readable.  If I see 's'[:count!=1] I certainly
have to sit and think about it a little bit; it's hardly a crystal clear
idiom.  ('', 's')[count != 1] is a tiny bit better since it's a slightly
more recognized idiom (and I've certainly used it on occasion), but it's
still not immediately clear.

Using a recognized idiom that's built into the language, whether it's x
if p else y or whatever might get adopted, makes it crystal clear what I
mean.  Same is with Booleans (which I also hoped for when first
encountering the language and supported the addition of one when it was
proposed); if I write return 1, I'm not sure whether that's going to be
used as a counting number, an integer, a degenerate
rational/float/complex number, a comparison result (-1, 0, +1), or a
Boolean.  When I see return True, I know right away.

If I see 

	's'[:count != 1]

I have to sit and think.

If I see

	's' if count != 1 else ''

(or whatever form gets adopted), provided I understand the form, the
intended meaning is much clearer.

Can conditional expressions be abused?  You bet; but so can any other
language feature.  Conditional expressions, if used properly, can make
code like your examples much clearer, because you don't have to pull a
trick to get the effect of a conditional operator, you have one already
built in.

> I was only somewhat joking above with the "minor points" comment.
> My feeling about the trinary operator is that, yes, there are a few
> places where it fits better.

My example was deliberately simple, to which there are (still in my
opinion) less clear alternatives.  For _real_ short circuiting, if you
want that compressed into an expression you have some real IOPyCC
contenders:

	(p and [x] or [y])[0]

or the incomparable

	(p and lambda: x or lambda: y)()

-- 
 Erik Max Francis / max at alcyone.com / http://www.alcyone.com/max/
 __ San Jose, CA, USA / 37 20 N 121 53 W / &tSftDotIotE
/  \ Ten lands are sooner known than one man.
\__/ (a Yiddish proverb)
    Bosskey.net: Aliens vs. Predator 2 / http://www.bosskey.net/avp2/
 A personal guide to Aliens vs. Predator 2.