Clojure from the ground-up: debugging | Hacker News Reader

Clojure from the ground-up: debugging

aphyr.com

70 pointsmomo-reina21 comments

Kyle's suggestions have a ton of value, but do note that debugging Clojure(Script) can also be a pedestrian activity if you so desire. All the usual mainstream bell/whistles IDE-based interactive debugging is available today in CursiveClojure - break on all exceptions, eval in frame, etc.

Also ClojureScript has had accurate source mapping in Safari, Chrome, Firefox, and Node.js for quite a long time now and recently moved to support targets like iOS. It's a huge time saver even for an experienced Clojure programmer :)

And if you favour open source IDEs the Eclipse based Counterclockwise http://doc.ccw-ide.org/ will provide the same debugging experience for Clojure.

I love ccw as a coding tool and the full editor repl is great too but I haven't been able to get breakpoints to work on it ever. Another thing that irks me about it lately is no real clojurescript support! That said it is still my primary clojure tool and things like figwheel help close the gap.

My debugging experience with Counterclockwise wasn't very positive. But admittedly the last time I tried was a very long time ago and things may have changed significantly.

I find debugging with Cursiveclojure to be easier.

I'm trying so hard to love Clojure, but I keep running into examples of code like this in the world: http://imgur.com/79FlpTL. I don't write dense Lisp code like that, but unfortunately a lot of other people do. And I spend far more time reading code than writing it.

Think of this code as taking a built-in data structure and repeatedly refining it by transforming it into a different data structure. With only a little more practice, you'll find reading (and writing!) code like this becomes second nature.

I find it much simpler to deal with this style than with the equivalent in other languages, which would scatter the details across a half-dozen files, a dozen classes, and two dozen methods.

Well that code seems pretty readable to me and appears to preserve some interesting properties of the original paper as mentioned below the source :)

As far as spending more time reading Clojure than writing it I experience this myself - and I'm pretty sure this isn't a bad thing.

I meant that in general, we as programmers will spend more time reading (our, others') code than writing it. Nature of the business. Did not mean to imply this was a bad thing.

When I consider adopting a language, part of that consideration is how well I can read others' idiomatic code. It's so important for understanding libraries. I'm still just having a hard time with Clojure. I know it clicks for a lot of people though.

Perhaps readable if you read a lot of Clojure, but hardly anything that is easy to test piece by piece inspecting intermediate data, using a REPL for example.

These are fairly common functions in clojure. You might not be there yet as it is pretty readable. My personal preference would be to work in some let expressions for extra clarity.

Amen to that. I try to write mostly functional Python, but I use a lot of temporary variables to make such things as f(g(args) more readable where each arg is in itself a non-trivial computation that must be passed along to another function.

The problem with your let strategy is this can also impair readability if you use too many as the code becomes too far indented to the right (lets inside of lets inside of lets).

But I guess in these situations you can just use transients that are first defined to nil. Some people will probably yell at you for this.

http://clojure.org/transients

If you don't mind, what document is that?

It's a page from the O'Reilly Clojure Programming book. I hate to point fingers at the authors, because it's generally a really great book. Just that some of the examples are rather dense, and I get the feeling that this is not at all unusual in the community.

It's from Clojure Programming by Chas Emerick, Brian Carper, Christophe Grand - http://shop.oreilly.com/product/0636920013754.do

Error messages and stack traces look pretty ugly.

In Common Lisp:

    CL-USER 51 > (bake (make-instance 'pie :flavor :blackberry) 375 10.25)

    Error: The value 3843.75 of #:|predicate1555| is not a predicate in CONDP.
      1 (continue) Supply a new value of #:|predicate1555|.
      2 (abort) Return to level 0.
      3 Return to top loop level 0.

A clear error message and a way to repair it.

A stack frame:

    Call to BAKE {offset 394}
      PIE               : #<PIE 402000993B>
      TEMP              : 375
      TIME              : 10.25
      DBG::|predicate|  : 3843.75
      DBG::|expression| : #<Function < 411007995C>
      DBG::G            : 3843.75

Which CL is that?

LispWorks, the one Rich Hickey used for quite some time... ;-)

Yeah, I feared that. That's the paid-for one, isn't it?

SBCL at compile time:

    * (defun bake (pie temp time)
      "Bakes a cake for a certain amount of time, returning a cake with a new
      :tastiness level."
      (setf (tastiness pie)
            (condp (* temp time) #'<
                   (400 :burned)
                   (350 :perfect)
                   (300 :soggy))))
    ; in: DEFUN BAKE
    ;     (CONDP (* TEMP TIME)
    ;         #'<
    ;       (400 :BURNED)
    ;       (350 :PERFECT)
    ;       (300 :SOGGY))
    ; ==>
    ;   (LET ((#:PREDICATE2 (* TEMP TIME)) (#:EXPRESSION3 #'<))
    ;     (DECLARE (TYPE FUNCTION #:PREDICATE2))
    ;     (CHECK-TYPE #:PREDICATE2 FUNCTION "a predicate in CONDP")
    ;     (COND ((FUNCALL #:PREDICATE2 #:EXPRESSION3 400) :BURNED)
    ;           ((FUNCALL #:PREDICATE2 #:EXPRESSION3 350) :PERFECT)
    ;           ((FUNCALL #:PREDICATE2 #:EXPRESSION3 300) :SOGGY)))
    ;
    ; caught WARNING:
    ;   Derived type of (* TEMP TIME) is
    ;     (VALUES NUMBER &OPTIONAL),
    ;   conflicting with its asserted type
    ;     FUNCTION.
    ;   See also:
    ;     The SBCL Manual, Node "Handling of Types"
    ;
    ; compilation unit finished
    ;   caught 1 WARNING condition

Pretty cool...

Does anyone know what color theme those code boxes use?

I apologize for being so shallow :)

Reply on news.ycombinator.com