Fix bug in SubstitutionString replace

NEWS.md

@@ -6,6 +6,7 @@ New language features
 
 * Support for Unicode 12.1.0 ([#32002]).
 * Methods can now be added to an abstract type ([#31916]).
+* Added argument `keep` to `unescape_string` ([#27125]).
 * Support for unicode bold-digits and double-struck digits 0 through 9 as valid identifiers ([#32838]).
 
 Language changes

base/regex.jl

@@ -457,12 +457,15 @@ function _write_capture(io, re::RegexAndMatchData, group)
     io.size = max(io.size, io.ptr - 1)
 end
 
+
+const SUB_CHAR = '\\'
+const GROUP_CHAR = 'g'
+const KEEP_ESC = [SUB_CHAR, GROUP_CHAR, '0':'9'...]
+
 function _replace(io, repl_s::SubstitutionString, str, r, re::RegexAndMatchData)
-    SUB_CHAR = '\\'
-    GROUP_CHAR = 'g'
     LBRACKET = '<'
     RBRACKET = '>'
-    repl = repl_s.string
+    repl = unescape_string(repl_s.string, KEEP_ESC)
     i = firstindex(repl)
     e = lastindex(repl)
     while i <= e

base/strings/io.jl

@@ -373,10 +373,12 @@ end
 # TODO: handle unescaping invalid UTF-8 sequences
 """
     unescape_string(str::AbstractString)::AbstractString
-    unescape_string(io, str::AbstractString)::Nothing
+    unescape_string(io, s::AbstractString, keep = ())::Nothing
 
 General unescaping of traditional C and Unicode escape sequences. The first form returns
 the escaped string, the second prints the result to `io`.
+The argument `keep` specifies a collection of characters which (along with backlashes) are
+to be kept as they are.
 
 The following escape sequences are recognised:
  - Escaped backslash (`\\\\`)
@@ -396,17 +398,22 @@ julia> unescape_string("\\\\u03c0") # unicode
 
 julia> unescape_string("\\\\101") # octal
 "A"
+
+julia> unescape_string("aaa \\\\g \\\\n", ['g']) # using `keep` argument
+"aaa \\\\g \\n"
 ```
 
 ## See also
 [`escape_string`](@ref).
 """
-function unescape_string(io, s::AbstractString)
+function unescape_string(io::IO, s::AbstractString, keep = ())
     a = Iterators.Stateful(s)
     for c in a
         if !isempty(a) && c == '\\'
             c = popfirst!(a)
-            if c == 'x' || c == 'u' || c == 'U'
+            if c in keep
+                print(io, '\\', c)
+            elseif c == 'x' || c == 'u' || c == 'U'
                 n = k = 0
                 m = c == 'x' ? 2 :
                     c == 'u' ? 4 : 8
@@ -456,6 +463,8 @@ function unescape_string(io, s::AbstractString)
         end
     end
 end
+unescape_string(s::AbstractString, keep = ()) =
+    sprint(unescape_string, s, keep; sizehint=lastindex(s))
 unescape_string(s::AbstractString) = sprint(unescape_string, s, sizehint=lastindex(s))
 
 """

test/regex.jl

@@ -46,6 +46,12 @@
     @test_throws ArgumentError match(r"test", GenericString("this is a test"))
     @test_throws ArgumentError findfirst(r"test", GenericString("this is a test"))
 
+    # Issue 27125
+    msg = "#Hello# from Julia"
+    re = r"#(.+)# from (?<name>\w+)"
+    subst = s"FROM: \g<name>\n MESSAGE: \1"
+    @test replace(msg, re => subst) == "FROM: Julia\n MESSAGE: Hello"
+
     # findall
     @test findall(r"\w+", "foo bar") == [1:3, 5:7]
     @test findall(r"\w+", "foo bar", overlap=true) == [1:3, 2:3, 3:3, 5:7, 6:7, 7:7]

test/strings/io.jl

@@ -143,6 +143,10 @@
         @test "\x01" == unescape_string("\\x01")
         @test "\x0f" == unescape_string("\\x0f")
         @test "\x0F" == unescape_string("\\x0F")
+
+        str= "aaa \\g \\n"
+        @test "aaa \\g \n" == unescape_string(str, ['g'])
+        @test "aaa \\g \\n" == unescape_string(str, ['g', 'n'])
     end
 end
 @testset "join()" begin