[BUG] Regular expressions with hex digits not working as expected

[andygrove]

Describe the bug

I am seeing cuDF fail to match some hex digits correctly. Here is one example. I do not yet know if this is related to #4409 or whether this is a separate issue.

javaPattern=\xA9, cudfPattern=\xA9, input='©', cpu=true, gpu=false

Steps/Code to reproduce bug

  test("compare CPU and GPU: hex") {
    val patterns = Seq(raw"\x61", raw"\xA9", raw"\xa9")
    val inputs = Seq("a", "b", "\u00A9")
    assertCpuGpuMatchesRegexpFind(patterns, inputs)
  }

Expected behavior
Test should pass.

Environment details (please complete the following information)
N/A

Additional context
N/A

[andygrove]

This is perhaps due to cuDF requiring UTF-8 and the fact that the copyright symbol has bytes 0xC2 0xA9 or 11000010 10101001 in UTF-8 encoding, and byte 0xA9 in ANSI encoding.

[andygrove]

Now that #4492 is merged, we fall back to CPU for now for hex digits.

[NVnavkumar]

With cuDF PR #10220 now merged, at least 127 characters should be supported by the cuDF regex hex escape, so this should be re-enabled and tested.