Good catches from several of you. The original fix dropped the try-catch entirely which was a regression for overflow and edge cases like a bare "-". Updated the post to keep a try-catch around the final parseInt as a safety net. The pre-validation still avoids the expensive path for the common cases, which is the core point. Appreciate the feedback.

Not to mention it's extra work in the case where the input usually is valid.

And it will fail with "-"