[dannyw]

Language detection works on language trigrams, and are trained on human language. I bet punctuation is stripped during preprocessing. So it's not surprising that code will have false positives.

[vidarh]

Not surprising, but still funny when applied to obfuscated code.