On Mon, May 6, 2013 at 12:28 AM, Lindsey Kuper <[email protected]>wrote:
> On Sun, May 5, 2013 at 6:17 PM, Andreas Rossberg <[email protected]> > wrote: > > On May 5, 2013, at 23:54 , Lindsey Kuper <[email protected]> wrote: > >> On Sun, May 5, 2013 at 4:19 PM, Noam Yorav-Raphael <[email protected]> > wrote: > >>> I have a simple suggestion: the current implementation of zip() > returns an > >>> iterator which stops whenever one of the two iterators it gets stop. > >>> I use zip() in python quite a bit. I always have a few lists, where > the i'th > >>> value in each corresponds to the same thing. I use zip in python to > iterate > >>> over a few of those lists in parallel. > >>> > >>> I think this is the usual use case. In this use case, when the two > lists > >>> have a different length it means that I have a bug. it seems to me that > >>> Python's behavior, and current Rust behavior, is contrary to "Errors > should > >>> never pass silently" from the zen of Python. > >>> > >>> What do you think of changing this, so that zip() will fail in such a > case? > >>> Another iterator, say, "zipcut" can implement the current behavior if > >>> needed. > >> > >> For what it's worth, in Wikipedia's comparison of implementations of > >> zip for various languages [0], none of them raise an error when the > >> lists are different lengths; they all either stop with the shorter of > >> the two lists, or fill in the missing values with a nil value. > > > > That may be coincidence, however, since the page lists only a handful of > languages. As a counter example, OCaml, which calls it 'combine', throws. > Standard ML even provides two variants, 'zip' and 'zipEq', the latter > throwing. (And as an additional data point, nowhere in my SML code have I > ever had a need for the non-throwing version.) > > Fair point. Perhaps Rust should also provide both. I like the SML names, > too. > > Lindsey > _______________________________________________ > Rust-dev mailing list > [email protected] > https://mail.mozilla.org/listinfo/rust-dev > In the name of preventing obvious mistakes I would strongly suggest implementing the reverse logic from SML: it's best when the shortest name provides the safe behavior and "unsafe" behaviors have more descriptive names indicating in what they are unsafe. It forces people to consciously choose the unsafe alternatives. I would therefore propose: - zip: only on collections of equal length - zipcut: stop iteration as soon as the shortest collection is exhausted - zipfill: fill the "void" (somehow: default value, Option<T>, ...) This way we have all 3 variants, with descriptive names for those 2 who introduce specific behavior. -- Matthieu
_______________________________________________ Rust-dev mailing list [email protected] https://mail.mozilla.org/listinfo/rust-dev
