Expired events sometimes have the `:time` field set to a value of type `Ratio` about riemann

[sanel]

I don't think this is a bug. Notice that Clojure Ratio is lazy - actual computation will not be done until is explicitly casted or converted. Take a look at these measurements - explicit conversion adds some overhead:

(require 'criterium.core)

;; no explicit conversion
(criterium.core/bench (doall (map #(/ 1234 %) (range 2 25))))

; Evaluation count : 26816220 in 60 samples of 446937 calls.
;              Execution time mean : 2.232489 µs
;     Execution time std-deviation : 29.268225 ns
;    Execution time lower quantile : 2.217370 µs ( 2.5%)
;    Execution time upper quantile : 2.318537 µs (97.5%)
;                    Overhead used : 7.253383 ns
; 
; Found 4 outliers in 60 samples (6.6667 %)
; 	low-severe	 1 (1.6667 %)
; 	low-mild	 3 (5.0000 %)
;  Variance from outliers : 1.6389 % Variance is slightly inflated by outliers

;; explicit conversion to long

(criterium.core/bench (doall (map #(long (/ 1234 %)) (range 2 25))))

; Evaluation count : 18773880 in 60 samples of 312898 calls.
;              Execution time mean : 3.206310 µs
;     Execution time std-deviation : 39.479470 ns
;    Execution time lower quantile : 3.180998 µs ( 2.5%)
;    Execution time upper quantile : 3.310231 µs (97.5%)
;                    Overhead used : 7.253383 ns
; 
; Found 5 outliers in 60 samples (8.3333 %)
; 	low-severe	 2 (3.3333 %)
; 	low-mild	 3 (5.0000 %)
;  Variance from outliers : 1.6389 % Variance is slightly inflated by outliers

In your case, I'd explicitly convert it to desired format before you print it. The same is done for other destinations (e.g. InfluxDB) - numbers are converted before they are saved at desired precision. In between, they are flowing as lazy unevaluted number type.

[v0lat1le]

I was leaning towards updating riemann.time/unix-time-real to (quot (System/currentTimeMillis) 1000)
It is also inconsistent with riemann.common.event which does (long (round (unix-time)))