argument ordering (was: Re: Priority Queues, or lack thereof)

Okasaki, C. DR EECS Christopher.Okasaki at
Fri Aug 19 16:26:56 EDT 2005

Malcolm Wallace wrote:
>> This style supports multiple operations nicely, especially with
>> with the $ operator.
> But note that the version with $ is actually longer than the one
> with nested parens!  

Only because I put in the third $, which wasn't necessary, and which
most people using this style would probably omit.  Without that $,
the same length.  And of course I could make the first one shorter by
out some of the whitespace, but I would never recommend that.  Character
are mostly irrelevant anyway -- what matters is ease of reading.  To my
the version with the $ is slightly easier in this case, but opinions can
vary.  I think the $ style becomes much easier to read as you (1) add
more operations,
and (2) start mixing operations (eg, some inserts and some deletes).
For example,
   insert 1 $ delete 2 $ delete 3 $ insert 4 $ delete 5 someset
   insert (delete (delete (insert (delete someset 5) 4) 3) 2) 1
In the latter style, how easy is it to tell whether 3 is being inserted
or deleted?
In the former style, it is obvious that 3 is being deleted.

Of course, I will freely admit that such examples are fairly rare in
real programs.

> And where the operands are not statically known,
> the different orderings do simply come down to 
> foldl/foldr preference:
>     insert 1 $ insert 2 $ insert 3 $ someSet
>  ==  foldr insert someSet [1,2,3]
>     insert (insert (insert someSet 3) 2) 1
>  ==  foldl insert someSet [3,2,1]
> In some sense the foldl variant is more 'obvious' about the order in
> which elements are actually inserted.

Yes, I agree that the order of operations in the $ style is somewhat
counter-intuitive (in exactly the same way that "foo . bar . baz" is).
This rarely seems to be a problem in practice, however.  The most common
examples where I've seen this style used are things like initializing
dictionary of an interpreter with the "built-ins", in which case you
that none of the keys overlap so the order doesn't matter.

-- Chris

More information about the Libraries mailing list