hrm

So what exactly are "collections". I often see fn docs mentions, take a collection, but I'm not sure which type it refers too. Does it mean clojure.lang.IEditableCollection ?

or is it IPersistentCollection?

OR even broader, the java.lang.Collection?

I would usually interpret that to mean IPersistentCollection

Cool, thanks

(noobie so confirm anything I write): I think alexmiller is correct -- a collection is anything that implements IPsersistentCollection. Another way would be to say it (coll? coll) returns true. This includes (at least) vector, list, map, set, and records (last surprised me.)

Deeper shower thoughts: What if - instead of a simple eval bot - an nrepl bot could be created that permits communal visibility and access to a running repl on a given Slack channel?

Hey

What's the difference between a dynamic binding and just using def with let to create scopes?

http://stackoverflow.com/questions/1523240/let-vs-binding-in-clojure

@herbm right - the tricky one is whether seqs are included

So I'm rendering a maze of interconnected rooms in SVG. The basic rectangles are trivial; you just write out some Hiccup like [:rect {:x 0 :y 0 :width 10 :height 10}]. And figuring out anchor points is trivial as well; it's not hard to calculate that the northeast corner of that rectangle is [10 0]. I've captured a bunch of calculations about the multipliers involved by calling it a "render environment", a map with keys like :margin and :cell-h-spacing and :cell-width. With those values, I can calculate rects for all the cells:

(into [:svg {...base attrs...}] (map (partial render-cell render-environment) all-cells))

But the problem comes when I start drawing the lines between them. Now I need to capture context while rendering. If I have already drawn a line between the cell at [0 0] and its eastern neighbor, I don't need to draw a reciprocal line when I get to that cell.

I was thinking of using reduce to keep track of a set of all drawn lines. Thanks to how Clojure handles values, two [:line {:x1 0, :y1 0, :x2 10, :y2 10}] vectors are equal. But reduce gets cumbersome if I want to build up the result collection and accumulate some other value.

I'd be doing something like (reduce (partial render-cell render-environment) {:cells [], :lines #{}} all-cells), where render-cell now has responsibility for accepting and returning the {cells, lines} map.

So I’m trying to compile clojure master branch and I get this weird compile error java.lang.NoClassDefFoundError: clojure/spec$map_spec_impl can anyone help me out? 🙂

@deiga, if nobody answers here, you might also ask in the #clojure-spec room, where @alexmiller is holding court on spec issues.

@deiga how are you compiling?

mvn clean compile should usually be sufficient

Hmm, now it works. Guess I had some broken artifact somewher

thanks 🙂

if you didn’t clean, sometimes that can cause problems

@amacdougall on using reduce… I’m a fan, but I agree that there is a tipping point when you’re trying to accumulate more than one kind of data. Eventually, I find that my code gets clearer when I rewrite as a loop/`recur`. But I don’t have a clear rule about when to make that switch

That's a good point. I think I can get away without it, though. I've defined render-cell so that it simply returns a Hiccup SVG like:

[:g
 [:rect ...]
 [:line ...]
 [:line ...]]

If this is too stupid to answer, feel free to ignore. But if i had a datastructure I wanted to save to disk for later. I could use transit right?

yes

I don't know if they've lifted that "don't persist transit until we nail down the spec" warning yet

good point. Thanks for the answer. I can always convert formats later. Very small proj for now. JSON would work too, just thought more native would be nice

edn is probably a bit easier

we haven’t (re transit) but there was some chatter recently about whether that should be done

Yeah, I tried EDN a while back and it didnt' work out, but was probably my ignorance. Will consider that again. thanks all for giving me a direction.

I can't imagine it's more difficult than transit

Is there an elegant way to print threading macro intermediate results?

(doto prn)

i.e., (-> a f1 f2 (doto prn) f3 f4)

How does Transit differ itself form EDN? Is a binary format, compressed, more expressive?

If you wanted to print every intermediate, you could create a macro that interleaves the (doto prn) expression between the list of expressions... Gonna try that as an exercise

@didibus Transit has both JSON (text) and MsgPack (bin) serializations

First macro 😄

if you are going to the browser, the JSON parser is highly optimized. because it employs optimizations like key caching, it can surprisingly be faster than conveying the same information over just JSON.

(defmacro prn->
 [x & forms]
 `(-> ~x
   ~@(interleave forms (repeat '(doto prn)))))

there are some libs out there that contain printing variants of the thread macros btw if you’re interested in using or comparing

Would be good to see how more well-considered versions pan out, aye. 🙂

sorry I can’t remember what those libs are atm :)

I'll dig around

So Transit is a competitor to something like Google's protocol-buffer

or Amazon ION??

I guess I'm finding my answers from the comments on the cognitech Transit blog

Though it does feel like Amazon ION fills an exact use case

I'm not sure I see where Transit really fits in with clojure. Is it planned to have it be a first class citizen? It really feels to me more like a library side project.

Transit differs from protobuff in that you don’t need a predefined schema - it is self-describing as it comes over the wire

I haven’t looked at Ion

Ion is self-describing. But its fair enough, Transit was announced before the reveal of Ion

yeah, looks similar

I find it funny that Ion does extensibility via s-exprs

I think Transit’s extensibility story is better based on a quick skim :)

I'll be keeping my eyes open on both 😛

Transit lets you build new tagged types based on existing types. readers that don’t have a knowledge of the tagged type can still read it and understand it’s parts (and send it along) even if they don’t have a known reader for the type.

man, adhoc hierarchies in conjunction with isa?-based multimethod dispatch is so cool