escaping - regex to match all unescaped '$' in a regex string - Stack Overflow

admin管理员组

文章数量:1345016

I want to build a regex that will match all unescaped $ in strings that represents a regex.

In this case, a character is unescaped if it contains an equal number of backslashes behind it (each pair of backslashes means the backslash char itself, and therefore the next character isn't escaped).

I came up with this pattern: (?<!\\)(\\{2})*\$

Explanation: although this will also match the backslashes preceding the $, this is the closest I came to a solution. This assures an equal number of backslashes before a $ which is not preceded by another backslash, making an odd number of backslashes.

My issues is that it seems that I need 2 consecutive non consuming groups to make the total number of backslashes even, but this is not possible. Is there another way to make that?

I want to build a regex that will match all unescaped $ in strings that represents a regex.

In this case, a character is unescaped if it contains an equal number of backslashes behind it (each pair of backslashes means the backslash char itself, and therefore the next character isn't escaped).

I came up with this pattern: (?<!\\)(\\{2})*\$

Explanation: although this will also match the backslashes preceding the $, this is the closest I came to a solution. This assures an equal number of backslashes before a $ which is not preceded by another backslash, making an odd number of backslashes.

My issues is that it seems that I need 2 consecutive non consuming groups to make the total number of backslashes even, but this is not possible. Is there another way to make that?

• regex
• escaping
• regex-lookarounds
• regex-group
• pcre

Share Improve this question Follow edited 12 hours ago Benny Brudner asked 12 hours ago Benny BrudnerBenny Brudner 3944 bronze badges 5

• "but this is not possible": I don't quite understand what the problem is. – trincot Commented 11 hours ago
• 1 If the issue is that you don't want to get the backslashes in the match, then use \K just before \$. Not sure what you are asking. – trincot Commented 11 hours ago
• \K should work. notionally, you could nest the lookbehinds, but variable-length is probably not supported (?<=(?<!\\)(\\\\)*)\$ – jhnc Commented 11 hours ago
• @CasimiretHippolyte: It deserves an answer – anubhava Commented 11 hours ago
• Nothing to do with Java that allows variable length lookbehinds. It's a PCRE question. – Casimir et Hippolyte Commented 8 hours ago

Add a comment  | 

3 Answers 3

Sorted by: Reset to default 3

One way to do it with PCRE is to avoid all bytes preceded by a backslash using the (*SKIP)(*FAIL) verbs sequence and to catch the DOLLAR SIGN in an other branch:

~ \\ . (*SKIP)(*F) | \$ ~xs

demo

Don't fet that to figure correctly a literal backslash in a quoted PHP string for a regex pattern, this one has to be escaped two times (one time for the regex, since it's a special character to form an escape sequence like \w or \$, and one time for the quoted string since this same character is used to form an escape sequence too like \'), so 4 backslashes to figure a single literal backslash:

$pattern = '~ \\\\ . (*SKIP)(*F) | \$ ~xs';

demo

A nowdoc string avoids the quoted string backslashes:

$pattern = <<<'REGEX'
~ \\ . (*SKIP)(*F) | \$ ~xs
REGEX;

Make the repeating group non-capturing, and add the meta escape \K.

/(?<!\\)(?:\\{2})*\K\$/g

Here it is on Regex101.

If you could not use the \K and (*SKIP*)(*FAIL*) syntax, like in Python (and many other regex flavors), you could.

Logic: Group 1 ($1) will consume and capture the preceding characters required for a match. Group 2 ($2) will consume and capture the unescaped $, \$, preceded by 0, or another even number of, literal backslashes, \.

REGEX PATTERN:

([^\\](?:\\\\)*)(\$)

REPLACEMENT STRING:

$1

Regex Demo: https://regex101/r/qq0Bug/2

TEST STRING:

aaa$aaa\$aaa\\$aaa\\\$
\$ \\$ \\\$ \\\\$ \\\\\$

RESULT:

aaaaaa\$aaa\\aaa\\\$
\$ \\ \\\$ \\\\ \\\\\$

MATCHES AND GROUPS:

MATCH 1:  2-4   a$
GROUP 1:  2-3   a
GROUP 2:  3-4   $

MATCH 2:  11-15 a\\$
GROUP 1:  11-14 a\\
GROUP 2:  14-15 $

MATCH 3:  25-29  \\$
GROUP 1:  25-28  \\
GROUP 2:  28-29 $

MATCH 4:  34-40  \\\\$
GROUP 1:  34-39  \\\\
GROUP 2:  39-40 $

REGEX NOTES:

• ( Begin capture group 1, referred to by $1 in the replacement string.
  • [^\\] Negated character class [^...]. Match any character that is not a literal backslash \.
  • (?:\\\\)* Non-capture group (?:...). Match two consecutive literal backslash characters \\ 0 or more (*) times.
• ) End group 1
• (\$) Group 2 ($2) match a literal $.

本文标签： escapingregex to match all unescaped 3939 in a regex stringStack Overflow