Bugfix: Allow regex captures of substrings of stringviews

A substring of `SubString{T}` is not `SubString{SubString{T}}`, which the
previous code assumed, but just `SubString{T}`.

Author: jakobnissen

src/regex.jl

@@ -28,8 +28,9 @@ function Base.match(re::Regex, str::T, idx::Integer, add_opts::UInt32=UInt32(0))
     end
     n = div(PCRE.ovec_length(data), 2) - 1
     p = PCRE.ovec_ptr(data)
-    mat = SubString(str, unsafe_load(p, 1)+1, prevind(str, unsafe_load(p, 2)+1))
-    cap = Union{Nothing,SubString{T}}[unsafe_load(p,2i+1) == PCRE.UNSET ? nothing :
+    SS = T <: SubString ? T : SubString{T}
+    mat = SubString(str, unsafe_load(p, 1)+1, prevind(str, unsafe_load(p, 2)+1))::SS
+    cap = Union{Nothing,SS}[unsafe_load(p,2i+1) == PCRE.UNSET ? nothing :
                                            SubString(str, unsafe_load(p,2i+1)+1,
                                            prevind(str, unsafe_load(p,2i+2)+1)) for i=1:n]
     off = Int[ unsafe_load(p,2i+1)+1 for i=1:n ]
@@ -192,4 +193,4 @@ end
 
 Base.iterate(m::SVRegexMatch, args...) = iterate(m.captures, args...)
 Base.length(m::SVRegexMatch) = length(m.captures)
-Base.eltype(m::SVRegexMatch) = eltype(m.captures)
+Base.eltype(m::SVRegexMatch) = eltype(m.captures)

test/runtests.jl

@@ -96,6 +96,10 @@ end
     sv = StringView(codeunits("foo 1234 bar"))
     @test match(r"[0-9]+", sv).match.string === sv
     @test eltype(eachmatch(r"[0-9]+", sv)) == SVRegexMatch{typeof(sv)}
+
+    # Regex match of substring of stringview
+    s = only(match(r"^([a-z]+)$", SubString(StringView((b"abc")))))
+    @test typeof(s) == SubString{StringView{typeof(b"abc")}}
 end
 
 @testset "named subpatterns" begin