I was wondering if anyone could examine this regular expression and rewrite it without catastrophic backtracking? My SonarQube seems to be upset. ;)

String NEGATIVE = "(.*?[^0-9] ?)-([0-9.] ?.*?)";

According to the Java code, it is attempting use this expression to help find and replace negative number signs with tilde characters.

// Replace any negative number signs as they will
// be confused with the subtraction operator.
expression = expression.replaceAll(NEGATIVE, "$1~$2").replaceFirst("^-", "~");

Thanks.

CodePudding user response：

In the patten (.*?[^0-9] ?)-([0-9.] ?.*?) the non greedy part .*? at the end can be omitted, as it will not match any characters.

This part [0-9.] ? will only match 1 character as the quantifier is non greedy so the quantifier can be omitted (as the following .*? does not match any character)

You can use lookarounds with just a single character in the assertions:

(?<!\d)-(?=\d)

• (?<!\d) Negative lookbehind, assert not a digit directly to the left of the current position
• - Match a hyphen
• (?=\d) Positive lookahead, assert a digit directly to the right of the current positions

Regex demo

CodePudding user response：

This will match all negative signs before digits in basic expressions. Execute, and then replace the entire match with a ~.

You did not post your input set, so there might be other cases that need to be coded.

(?<=[- *\/0-9 ]|^)-(?=[\d] )

Here are some examples of replacements https://regex101.com/r/3H30tU/1https://regex101.com/r/3H30tU/1