我们可以使用基于lookbehinds的条件吗？

Question

我有一个正则表达式，其任务与此类似。这个人想要查找价格 (demo)，其中美元符号必须在数字之前或之后，但不能两者兼而有之。

  \b       # "word" boundary
  (?<=\$)  # Assert that current position is preceded by $
  \d+\b    # digits, boundary
  (?!\$)   # Assert that current position is not followed by $
|          # alternation
  \b       # boundary
  (?<!\$)  # Assert that current position not preceded by $
  \d+\b    # digits, boundary
  (?=\$)   # Assert that current position is followed by $

PCRE 中有没有一种方法可以使用条件来对条件进行类似操作。 (demo)

(                 # capture 1, because we can't quantify lookarounds
  (?<=\$)         # Asserts that $ precedes current position
)?                # close capture 1, make it "optional" (0 or 1 times)

\b                # boundary
\d+               # digits
\b                # boundary
(?(1)             # conditional that capture #1 was a success
  (?!\$)          # if successful, assert that it is not followed by a #
|
  (?=\$)          # unsuccessful, assert that this position is followed by a $
)

注意：单词边界很重要，因此两者都要捕获整数，否则正则表达式会后退一步，从多位数字中剪掉一个数字。

上面的两个表达式都匹配 $15 和 16$，但不是 $17$，因为它们应该匹配。如果没有边界，它们将匹配 17 美元中的 15 美元、16 美元和 1 美元。

Answer 1

一个可行的解决方案 (demo)。问题的解决方法是 99 步，在这个简单的例子中是不必要的牺牲。在某些情况下，牺牲（如果有的话）是非常谨慎的。

这最终变得非常简单。使捕获组 1拥有。这意味着无论它捕获什么，如果正则表达式试图回溯，它都不会投降。这意味着一旦它决定数字应该以$ 开头，如果数字后面也跟着$，它将不会让正则表达式回溯。它会放弃并继续前进，找到下一个边界，然后是数字并查看其周围环境。

(                 # capture 1, because we can't quantify lookarounds
  (?<=\$)         # Asserts that $ precedes current position
)?+               # close capture 1, make it "optional" (0 or 1 times)
                  # the possessive quantifier is the only change to the regex
                    # + is only 'possessive' when it's the second character of a quantifier

\b                # boundary
\d+               # digits
\b                # boundary
(?(1)             # conditional that capture #1 was a success
  (?!\$)          # if successful, assert that it is not followed by a #
|
  (?=\$)          # unsuccessful, assert that this position is followed by a $
)