Towards a more Pythonic "Ternary Operator"

Dennis Reinhardt DennisR at dair.com
Wed Mar 5 03:14:06 CET 2003


>  If we can all agree that
> this is a weakness of Python, then we can look of various ways
> to solve the problem.

The weakness is that Python has a "dangling else" ambiguity which requires
indentation to disambiguate (i.e. the if-else must occupy two lines
minimum).

One would like to write something like

    A = if <cond> then B else C endif  (not Python!)

Ignoring the assignment for awhile, can get close to a one-liner if with

    if <cond>: B; else: C   (not valid Python)

Arguably (and incorrectly argued), the ";" allows this to be written as a
one-liner  To see why this can never be valid Python, consider nesting "if
<cond2>: D" for B to yield:

    if <cond>: if <cond2>: D; else: C

It is truly ambiguous whether this parses as:

    if <cond>:
        if <cond2>: D;
        else: C
or

    if <cond>:
        if <cond2>: D;
    else: C

The solution I see for resolving "dangling else" is to put a token at the
end, just as I have in the "endif" in the first example (this example is
from PLD2, a language I designed).  I think tacking something onto if-end is
impractical without something new tacked on at the beginning to notify the
interpreter to parse for the something on the end.

>From where I sit, resolving "dangling else" in Python without indentation
just introduces cruft into the language.  There are things my language will
not do well.  A single-line "if" does not look like a good fit in Python.
I would be inclined here to not change anything.

>     <variable> = ( if <condition>:
>                        <value-a>
>                    else:
>                        <value-b>
>                  )

I presume that

     <variable> = ( if <condition>: value-s>)

is also allowed.  Does this extend to

     <variable> = subroutine_call(( if <condition>: value-s>))

Where the outer parens show that subroutine_call is being invoked and inner
parens show if expression is being evaluated.
--

Dennis Reinhardt

http://www.spamai.com






More information about the Python-list mailing list