Split mathematical expression on every operator/bracket

Question

I'm trying to split math expressions using regex although no matter what I try, there's always one condition that I'm missing.
I need to split the string on all operators +,-,*,/ and on all types of brackets (,),[,]. If there's an increment or decrement ++,-- I need to keep it as one string. I also have to keep multidigit numbers as one string.

Right now, the closest thing I have is:
(?<![+-[0-9]])(?=[*/+-\[\]\(\)])|(?<=[*/+-\[\]\(\)])(?![+-[0-9]])

For example, [(([++[[(--2)]*33]])/22)+1] should give
[,(,(,[,++,[,[,(,--,2,),],*,33,],],),/,22,),+,1,] but instead gives
[,(,(,[,++,[,[,(,--2,),],*,33,],],),/,22,),+1,]

I wouldn't use a regex for this. Should be much easier (and more performant) using a normal loop. — marstran, Oct 26 '17 at 16:22
@marstran Mmm always thought a loop would be slower. I'd still like to understand how I could make it work using regex though — Girrafish, Oct 26 '17 at 16:28
Refer to this - https://stackoverflow.com/questions/3481828/how-to-split-a-string-in-java — Germa Vinsmoke, Oct 26 '17 at 17:28

score 2 · Answer 1 · answered Oct 26 '17 at 16:29

tldr: Other developers that inherit your code will be able to understand the \d+ regex... But they're very unlikely to be able to look at your monster of a regex — or any other long regex, for that matter — and understand what's going on.

If there's a simpler way to do it without regexes, that's usually the better option for both readability and performance. But if you must use regex...

My recommendation is to use the much simpler \d+ to find all digits, or \D to find all non-digits; then, you should loop through the result and combine any subsets of the array tokens where tokens[i] and tokens[i+1] are both + or -.

As commenters have noted, this is still likely less performant than ignoring RegExes altogether and splitting using a loop.

While regular expressions are interesting and very useful for certain tasks, matching-bracket tasks like this is particularly poorly suited for regex (because the math syntax we commonly use is not a regular language).

Split mathematical expression on every operator/bracket

1 Answers1