Problematic behaviour of #scan_integer(base: 16) when parses `"0x"`

[andrykonchin]

The method #scan_integer called with base: 16 is supposed to parse hexadecimal integer literals. There is an edge case when "0x" is not followed by any hex character (e.g. 0x.!@#). In this case #scan_integer doesn't match an integer literal and returns nil:

StringScanner.new("0x.!@#").scan_integer(base: 16)
# => nil

But there are might be pretty valid cases when 0 is expected to be returned instead:

StringScanner.new("0xyz").scan_integer(base: 16) # `"xyz"` is a plain text and `0` is expected

Wondering if it makes sense to match "0" and return 0 in the following cases:

• "0x<non-hex-char>"
• "+0x<non-hex-char>"
• "-0x<non-hex-char>"
• "0x<end-of-string>"
• "+0x<end-of-string>"
• "-0x<end-of-string>"

[eregon]

cc @byroot
This was found when implementing scan_integer on TruffleRuby, where the natural way to do it would be using scan(/[+-]?(0x)?[0-9a-fA-F]+/).

[byroot]

Do you have a repro?

I'm trying to fix this but if I add a test for it it passes:

diff --git a/test/strscan/test_stringscanner.rb b/test/strscan/test_stringscanner.rb
index eb35dfa..d939a2d 100644
--- a/test/strscan/test_stringscanner.rb
+++ b/test/strscan/test_stringscanner.rb
@@ -1056,6 +1056,11 @@ module StringScannerTests
     assert_equal(0, s.pos)
     refute_predicate(s, :matched?)
 
+    s = create_string_scanner('0xZZZZ')
+    assert_nil(s.scan_integer(base: 16))
+    assert_equal(0, s.pos)
+    refute_predicate(s, :matched?)
+
     s = create_string_scanner('-0x')
     assert_nil(s.scan_integer(base: 16))
     assert_equal(0, s.pos)

[byroot]

Nevermind, I kinda misread.

[byroot]

#141