On Wednesday, October 19, 2016 at 7:23:21 AM UTC-5, Paul Gowder wrote: > > Hi folks, > > I have a kinda basic question (I think?) that I should really know the > answer to after writing Clojure for like 8 months, but don't. It's about > requiring namespaces with protocol definitions in them but not the actual > implementations of those protocols, and how it works. > > tl/dr: if a protocol is defined in namespace a and implemented (extended > to the type at issue) in namespace b, can you call functions from it in > namespace c where namespace c only requires namespace a, not namespace b? >
Yes. You are invoking the protocol function in a and as a caller, that's all you need to know about. HOWEVER, the implementation in b must be loaded before you can do so. Your namespace doesn't necessarily have to do it, but someone does. Each protocol is essentially a little piece of hidden state in the runtime. Each time you extend the protocol to an implementation, that state is modified. If you don't load the code that does the extension, the state will not be updated. All of this also applies to multimethods. > Backstory: I was poking around core.matrix recently in pursuit of > restarting work on a months ago issue I promised to work on, and it looks > like at least one namespace calls a function it shouldn't be able to call. > > So: clojure.core.matrix.impl.pprint requires only one namespace within > core.matrix, clojure.core.matrix.protocols. (I'm going to use ... hereafter > to stand for clojure.core.matrix.) ...protocols defines a protocol > PSliceSeq2, which gives the name, docstring, etc. for a function called > get-slice-seq. PSliceSeq2 is actually implemented for the generic object > case in ...impl.defaults and for ISeq in ...impl.sequence. > > But ...impl.pprint calls ...protocols/get-slice-seq. And it seems to work. > Even though I can't find either ...impl.defaults or ...impl.sequence > anywhere in the dependency tree for ...impl.pprint. > > I haven't looked at the code, but something must be loading the extension before you invoke it. > This is a puzzle to me. I mean, the ...impl.pprint namespace will > typically be required from some other namespace, and presumably that > higher-level namespace will require everything that one needs. But still, I > hadn't thought that one would be able to call get-slice-seq in a function > in ...impl.pprint without requiring a namespace that actually defines the > function, as opposed to just defining the protocol. > > So can someone explain how this magic works? Is the compiler just way > smarter than I thought it was, and capable of figuring this stuff out on > its own? > > Thanks! > > -Paul > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
