Improve performance of tagged streams

[nukemberg]

Create tags-coll set only once to avoid expensive set creation and allocation

[sanel]

After some testing, this small change was faster indeed :) Please notice that my tests will not use the code from PR, but simulate the difference when set is created on-the-fly (when event is received, old behavior) and when set is created before stream function is defined (new behavior from this PR).

(require '[criterium.core :as c])

(def tags ["foo" "bar" "baz"])  ; default behavior
(def tags-set (set tags))       ; generated set object before stream was created

(c/bench (tagged-any? tags {:tags ["foo"]}))
; Evaluation count : 95891520 in 60 samples of 1598192 calls.
;              Execution time mean : 628.626019 ns
;     Execution time std-deviation : 8.187710 ns
;    Execution time lower quantile : 622.051208 ns ( 2.5%)
;    Execution time upper quantile : 648.455723 ns (97.5%)
;                    Overhead used : 2.654335 ns
; 
; Found 3 outliers in 60 samples (5.0000 %)
; 	low-severe	 2 (3.3333 %)
; 	low-mild	 1 (1.6667 %)
;  Variance from outliers : 1.6389 % Variance is slightly inflated by outliers

(c/bench (tagged-any? tags-set {:tags ["foo"]}))
; Evaluation count : 319278240 in 60 samples of 5321304 calls.
;              Execution time mean : 184.413915 ns
;     Execution time std-deviation : 1.959476 ns
;    Execution time lower quantile : 181.763322 ns ( 2.5%)
;    Execution time upper quantile : 187.707063 ns (97.5%)
;                    Overhead used : 2.654335 ns
; nil

(c/bench (tagged-all? tags {:tags ["foo"]}))
; Evaluation count : 102135000 in 60 samples of 1702250 calls.
;              Execution time mean : 585.095348 ns
;     Execution time std-deviation : 1.638328 ns
;    Execution time lower quantile : 584.169099 ns ( 2.5%)
;    Execution time upper quantile : 589.706420 ns (97.5%)
;                    Overhead used : 2.654335 ns
; 
; Found 3 outliers in 60 samples (5.0000 %)
; 	low-severe	 3 (5.0000 %)
;  Variance from outliers : 1.6389 % Variance is slightly inflated by outliers

(c/bench (tagged-all? tags-set {:tags ["foo"]}))
; Evaluation count : 204666780 in 60 samples of 3411113 calls.
;              Execution time mean : 290.374668 ns
;     Execution time std-deviation : 0.656243 ns
;    Execution time lower quantile : 287.957636 ns ( 2.5%)
;    Execution time upper quantile : 291.137373 ns (97.5%)
;                    Overhead used : 2.654335 ns
; 
; Found 5 outliers in 60 samples (8.3333 %)
; 	low-severe	 4 (6.6667 %)
; 	low-mild	 1 (1.6667 %)
;  Variance from outliers : 1.6389 % Variance is slightly inflated by outliers

It is >3x improvement for tagged-any? and >2x improvement for tagged-all? when they receive a set object :)

cc @jamtur01 @pyr

[pyr]

Yes, this one is a no-brainer, thanks @nukemberg!