Fix TrieIterator to correctly handle non-ascii chararacters

Squashed version of commits in JuliaCollections#756 with relevant changes only.

Backport to v0.18, meaning we're not including `find_prefixes`.

Co-authored-by: Du Shiqiao <[email protected]>
Co-authored-by: Reuben Gardos Reid <[email protected]>

Author: lucidfrontier45

src/trie.jl

@@ -111,12 +111,12 @@ end
 # since the root of the trie corresponds to a length 0 prefix of str.
 function Base.iterate(it::TrieIterator, (t, i) = (it.t, 0))
     if i == 0
-        return it.t, (it.t, 1)
-    elseif i == length(it.str) + 1 || !(it.str[i] in keys(t.children))
+        return it.t, (it.t, firstindex(it.str))
+    elseif i > lastindex(it.str) || !(it.str[i] in keys(t.children))
         return nothing
     else
         t = t.children[it.str[i]]
-        return (t, (t, i + 1))
+        return (t, (t, nextind(it.str, i)))
     end
 end
 

test/test_trie.jl

@@ -40,4 +40,17 @@
         @test collect(partial_path(t, "ro")) == [t0, t1, t2]
         @test collect(partial_path(t, "roa")) == [t0, t1, t2]
     end
+
+    @testset "partial_path iterator non-ascii" begin
+        t = Trie(["東京都"])
+        t0 = t
+        t1 = t0.children['東']
+        t2 = t1.children['京']
+        t3 = t2.children['都']
+        @test collect(partial_path(t, "西")) == [t0]
+        @test collect(partial_path(t, "東京都")) == [t0, t1, t2, t3]
+        @test collect(partial_path(t, "東京都渋谷区")) == [t0, t1, t2, t3]
+        @test collect(partial_path(t, "東京")) == [t0, t1, t2]
+        @test collect(partial_path(t, "東京スカイツリー")) == [t0, t1, t2]
+    end
 end # @testset Trie