bench

benchmark/benchmark.jl

@@ -10,20 +10,24 @@ function f(out, t, x, y)
     end
 end
 
-# I get 0.12 vs 0.10 in stringdist
-# http://www.markvanderloo.eu/yaRb/2013/09/07/a-bit-of-benchmarking-with-string-distances/
-
+# similar
 @time f(Int, Levenshtein(), x, y)
+@time f(Float64, JaroWinkler(0.1, 0.7, 5), x, y)
+
+# all 5-10x slower
 @time f(Float64, Jaccard(2), x, y)
 @time f(Float64, Cosine(2), x, y)
 @time f(Float64, Cosine(2), x, y)
+@time f(Int QGram(2), x, y)
+
+
 #= Rcode
 library(stringdist)
 x <- sapply(sample(5:25,1e5,replace=TRUE), function(n) paste(sample(letters,n,replace=TRUE),collapse="")) 
 y <- sapply(sample(5:25,1e5,replace=TRUE), function(n) paste(sample(letters,n,replace=TRUE),collapse=""))
-stringdist(x,y,method='lv')
-stringdist(x,y,method='jaccard')
-stringdist(x,y,method='jaccard')
-stringdist(x,y,method='cosine')
+system.time(stringdist(x,y,method='lv'))
+system.time(stringdist(x,y,method='jaccard'))
+system.time(stringdist(x,y,method='cosine'))
+system.time(stringdist(x,y,method='qgram'))
 
 =#

src/qgrams_distances.jl

@@ -115,15 +115,15 @@ function evaluate{T}(dist::Jaccard, s1::T, s2::T)
     length(s1) > length(s2) && return evaluate(dist, s2, s1)
     length(s2) == 0 && return 0.0
 
-    n2 = length(s2) - dist.q + 1
-    n1 = length(s1) - dist.q + 1
    
     set2 = Set{T}()
+    n2 = length(s2) - dist.q + 1
     @inbounds for i2 in 1:n2
         push!(set2, s2[i2:(i2 + dist.q - 1)])
     end
 
     set1 = Set{T}()
+    n1 = length(s1) - dist.q + 1
     @inbounds for i1 in 1:n1
         push!(set1, s1[i1:(i1 + dist.q - 1)])
     end