A type assertion is dangerous because if the thing you are asserting to doesn't match the type of the value inside the interface, what type should the result be? With the map example the type of the result is known.
On Thu, 29 Sep 2016, 19:36 Axel Wagner <axel.wagner...@googlemail.com> wrote: > On Thu, Sep 29, 2016 at 9:31 AM, Dave Cheney <d...@cheney.net> wrote: > >> Sorry, I misspoke, this logic does not apply to map lookup > > > But why? It seems to me, everything you said can just as easily be applied > to map lookups. Why is a zero-value on a type-assertion potentially > dangerous, while it's fine on a map-lookup? > I don't see anything in your E-Mail that is specific to type-assertions (I > mean. It's specific to type-assertions as implemented. Just not as > theorized). > > Indeed, you could argue that, as we deal just fine with the behavior for > maps, we would probably deal just fine with the behavior for type > assertions. And that the same conveniences we get from *not* panic'ing on > map-accesses also would apply to type-assertions. > > So… I have to agree, that there is an inconsistency here. I don't think > it's a *bad* thing and I prefer my type-assertions panic'ing and my > map-accesses not-panic'ing, but I would be hard pressed to give a technical > argument why they should be different. > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
