The only hard parts about property based testing are the properties and the generators. ;)
On Wednesday, April 30, 2014 6:38:19 AM UTC-5, henry w wrote: > > Hi, I wanted to get started with clojure.test.check (formerly > simple-check) and I am new to property based testing. > > I plucked clojure.core/group-by for no particular reason as a function to > test. > > I started by stating some properties i think should hold: > > ;; 1. applying the grouping key function to each member in a grouping > should result in the grouping key > ;; 2. flattening the vals of the group-by result should give back the > contents of the original collection. > ;; 3. no element appears in more than one grouping. > those seem good > > so far so good I think. there may be others but this seems ok for now. > > now, how to generate some data. > > for group-by we need two params: > 1) a grouping function > 2) a collection of items to be grouped > > If I start by naively generating collections of maps (containing keyword > keys and int vals, for example), the data is of the right shape to use in > group by, but there is no guarantee that: > 1) any of the maps share a key that I could use for grouping > 2) the values under a common key are shared > > This is really the crux of my problem.... ideally I would have the > generator *mostly* produce data which is actually doing to result in the > sort of collection i might want to call group-by on in real life (ie not > have everything grouped under nil on each generation). So should i create a > generator that creates keywords (which i will want to use as grouping > function) then have another generator that produces what are going to be > the values under this grouping key, then a generator that uses both of > these to create collections of maps from these. then i would have to find > out what the grouping keyword was that was generated.... this could all > work, I have read enough about generators to have a stab at this... but is > it the right approach? > You don't seem to be leveraging the possibilities of the grouping function. If you create a grouping function that maps many random values into a small number of groups, then everything may get easier. Some candidate functions: first character of the keyword, length of the keyword, etc. Or working backwards is often useful with a generator - generate the grouping values first, then use the inverse of the grouping function to generate data that maps to that group and populate the input with that. > > as far as implementing tests for the properties so far, I have done > property 2 above, using a basic generator and yanking out an arbitrary key > from it.... clearly a flawed approach as not much 'realistic' grouping is > going to happen here. > > (def vector-of-maps (gen/such-that not-empty (gen/vector (gen/such-that > not-empty (gen/map gen/keyword gen/int))))) > > (def all-elements-are-grouped > (prop/for-all [group-by-input vector-of-maps] > (let [a-map-key (-> group-by-input first keys first)] ;; > hmm, seems far from ideal > (= (set group-by-input) (-> (group-by a-map-key > group-by-input) vals flatten set))))) > > help appreciated... perhaps I need to learn more about the paradigm first, > but resources linked from the readme are all a bit more basic than this. so > if you know of some more advanced tutorials please let me know. > > Thanks > -- 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.
