Hacker News Reader

Comment by mhb | Hacker News Reader

Not the main point but if the if-else pattern gets tedious why is there no switch in Python?

Because a switch can usually be replaced by a dictionary?

    { 1: runOne, 2: runTwo, 3:runThree }[n]()

is equivalent to

    switch(n) { 
        case 1: runOne(); break; 
        case 2: runTwo(); break;
        case 3: runThree(); break;
    }

Also, on Smalltalk (which influenced python) there is no switch statement as well. On the other hand, Smalltalk has no if, while, for statements as well (they are all methods).

While Smalltalk doesn't have case statements, a) it's trivial to add them, and b) the way you get around it is totally different: rather than use a Dictionary, you instead generally make a pile of tiny classes, each with a fooBar: method, and then simply make sure you've got the right object type and call fooBar: on it indiscriminately--in other words, basically your classic visitor pattern.

That's totally possible in Python, too, but I don't really ever see it done.

I like to use defaultdict for this. A strategy dictionary, with a default strategy.

dispatch = defaultdict(default_stategy)

dispatch['one'] = new_strategy

print dispatch['two'] => default_strategy.

Beats case statements and elif chains.

I usually just provide a default argument to get:

dispatch.get('two', default_strategy)

This forces programmer to invent dummy names. And Python doesn't support nontrivial lambdas.

I've seen the argument about why that's more "pythonic", but it's not compelling. Allowing some syntactic sugar would make the meaning a lot more obvious.

Consing for control-flow is not normal, but on Python it is. Python - not even once.

In this case, a switch statement wouldn't help because you're dealing with a bunch of boolean functions rather than a single value. If you were using something like typeof and switched on the type, a switch statement would make more sense. However, even that would be pretty tedious for code like this, I think.

In general though, I have no idea why Python doesn't have a switch or case statement of some sort.

Now it's a single value:

    case [typeof(X) for x in X] of:
        [int, int]:
            ...
        [int, float]:
            ...

    { (int, int): methodWithTwoInts,
       (int, float): methodsWithIntAndFloat,
       ...
    }[  tuple([typeof(x) for x in X])  ]()

Actually this is better that switch (or whatever).

Some advanced dispatch code can notice that some tuples are dispatched more often that the rest and can check for them first.

Ditto for dict lookup.

But this would unexpected for switch code.

Python doesn't need a switch or case statement because you can write a suite of ad-hoc functions, stick them in a dictionary with your values as keys, and dispatch on the values:

    >>> def first():
    ...     print "a"
    ...
    >>> def second():
    ...     print "beta"
    ...
    >>> caseof = {1: first, 2: second}
    >>> caseof[1]()
    a
    >>> caseof[2]()
    beta
    >>>

It's not necessary. If there is long if/else chain in your code, something is probably wrong. In case of python (and other OO, weakly typed languages).

But my opinion is that it wouldn't hurt to have it there.

COND, man, COND.

Reply on news.ycombinator.com