正则表达式：对\b边界行为的困惑及十六进制匹配问题

Got it, let's fix this regex issue for your Atom custom syntax highlighting. The core problem here is that \b (word boundary) behaves differently for $ vs 0x—since $ isn't a word character (word chars are a-z, A-Z, 0-9, _), \b doesn't work as expected to block unwanted preceding characters for the $ prefix.

Here's the single regex that will handle both cases correctly, avoiding false matches like t0x0t:

(?<!\w)(?:\$|0x)[A-Fa-f0-9]+

Breakdown of how this works: #

• (?<!\w): A negative lookbehind assertion that ensures there is no word character immediately before our hex prefix. This blocks cases like t0x0t (where t is a word character before 0x) and also prevents unwanted matches like a$123 (where a is a word character before $).
• (?:\$|0x): A non-capturing group that matches either the $ or 0x prefix. Using ?: avoids creating an unnecessary capture group, which is cleaner for syntax highlighting.
• [A-Fa-f0-9]+: Matches one or more hexadecimal digits (upper/lowercase letters + numbers).

Test cases this handles correctly: #

• ✅ Matches $affe (whether at the start of a line or after a non-word character like !$affe)
• ✅ Matches 0xeffa
• ❌ Does NOT match t0x0t (the t before 0x triggers the negative lookbehind, so the regex skips this)
• ❌ Does NOT match foo$123 (the o before $ is a word character, so the lookbehind blocks the match)

Why this is better than using \b:
\b relies on the transition between word and non-word characters. For 0x, \b0x works because 0 is a word character—if there's a word character before it (like t), there's no word boundary between t and 0, so it doesn't match. But for $, \b$ would actually match cases like t$affe (since t is a word character and $ is non-word, creating a boundary), which we don't want. The negative lookbehind fixes this by explicitly checking for the absence of a preceding word character, regardless of the prefix type.

内容的提问来源于stack exchange，提问作者koalag