String#force_encoding don't clear coderange if encoding is unchanged

Some code out there blind calls `force_encoding` without checking
what the original encoding was, which clears the coderange uselessly.

If the String is big, it can be a rather costly mistake.

For instance the `rack-utf8_sanitizer` gem does this on request
bodies.

3 files changed, 19 insertions, 2 deletions

diff --git a/string.c b/string.c
index b3004624dd..41641c67e8 100644
--- a/string.c
+++ b/string.c
@@ -10843,7 +10843,23 @@ static VALUE
 rb_str_force_encoding(VALUE str, VALUE enc)
 {
     str_modifiable(str);
-    rb_enc_associate(str, rb_to_encoding(enc));
+
+    rb_encoding *encoding = rb_to_encoding(enc);
+    int idx = rb_enc_to_index(encoding);
+
+    // If the encoding is unchanged, we do nothing.
+    if (ENCODING_GET(str) == idx) {
+        return str;
+    }
+
+    rb_enc_associate_index(str, idx);
+
+    // If the coderange was 7bit and the new encoding is ASCII-compatible
+    // we can keep the coderange.
+    if (ENC_CODERANGE(str) == ENC_CODERANGE_7BIT && encoding && rb_enc_asciicompat(encoding)) {
+        return str;
+    }
+
     ENC_CODERANGE_CLEAR(str);
     return str;
 }
diff --git a/test/objspace/test_objspace.rb b/test/objspace/test_objspace.rb
index bc6799b49f..1f1709fb76 100644
--- a/test/objspace/test_objspace.rb
+++ b/test/objspace/test_objspace.rb
@@ -585,7 +585,7 @@ class TestObjSpace < Test::Unit::TestCase
 
   def test_dump_string_coderange
     assert_includes ObjectSpace.dump("TEST STRING"), '"coderange":"7bit"'
-    unknown = "TEST STRING".dup.force_encoding(Encoding::BINARY)
+    unknown = "TEST STRING".dup.force_encoding(Encoding::UTF_16BE)
     2.times do # ensure that dumping the string doesn't mutate it
       assert_includes ObjectSpace.dump(unknown), '"coderange":"unknown"'
     end
diff --git a/test/ruby/test_process.rb b/test/ruby/test_process.rb
index 47228d35e6..0416b20176 100644
--- a/test/ruby/test_process.rb
+++ b/test/ruby/test_process.rb
@@ -2747,6 +2747,7 @@ EOS
     require 'objspace'
     begin;
       obj = "a" * 12
+      obj.force_encoding(Encoding::UTF_16LE)
       obj.force_encoding(Encoding::BINARY)
       assert_include(ObjectSpace.dump(obj), '"coderange":"unknown"')
       Process.warmup