Java Regex . Get substring between certain KEYWORDS

Question

I need help creating the best possible regular expression for this problem.

I have combinations / sets of Starting and End Delimeters and I need to get ALL the substring / any words between the starting delimeter upto the end delimeter.

Assume this table of Delimeters:

START | END

CAT | DOG

APPLE | ORANGE

LION | ZEBRA

PANDA | CAT

sample Input:

substring1 CAT substring2 substring3 DOG substring4 substring5 CAT substring6
APPLE substring7 substring 8 ORANGE ORANGE substring9 DOG substring10 PANDA
substring11 CAT substring12 DOG substring13 LION substring10 substring11 ZEBRA substring12
CAT substring13 substring14 APPLE substring15 substring 16 ORANGE

The output must be:

1. CAT substring2 substring3 DOG
2. APPLE substrin7 substring8 ORANGE
3. PANDA substring 11 CAT
4. LION substring10 substring 11 ZEBRA
5. APPLE substring15 substring16 ORANGE

My regular expression:

 CAT (.)*? DOG | APPLE (.)*? ORANGE | LION (.)*? ZEBRE |  PANDA (.)*? CAT 

I have problem dealing with string that has multiple occurence of other starting delimeter.

take for example:

CAT word1 word2 word3 word4 APPLE word5 word6 word7 DOG 

I know that it will match with this CAT (.)*? DOG but this is wrong since the substring contains one of the starting delimeters.

I just need a regex that that will get all the words between a starting delimeter upto its matching end delimeter if ever the substring does not contain any occurence of other starting delimeters.

any suggestion? Thanks

Answer 1

The technique that helps us here is called "lookaround".

I Updated my answer after clarification of nfinium and feedback from jsobo

CAT ((?!(APPLE|LION|PANA)).)*? DOG|APPLE ((?!(CAT|LION|PANDA)).)*? ORANGE|LION ((?!(CAT|APPLE|PANDA)).)*? ZEBRA|PANDA ((?!(APPLE|LION)).)*? CAT

Given the input:

substring1 CAT substring2 substring3 DOG substring4 substring5 CAT substring6 APPLE substring7 substring 8 ORANGE ORANGE substring9 DOG substring10 PANDA substring11 CAT substring12 DOG substring13 LION substring10 substring11 ZEBRA substring12 CAT substring13 substring14 APPLE substring15 substring 16 ORANGE  string CAT dkdkdkdkdk CAT dkdkdk dkdkdk ORANGE dkdkdkdk DOG etc. CAT word1 word2 word3 word4 APPLE word5 word6 word7 DOG wordx

It matches

CAT substring2 substring3 DOG
APPLE substring7 substring 8 ORANGE
PANDA substring11 CAT
LION substring10 substring11 ZEBRA
APPLE substring15 substring 16 ORANGE
CAT dkdkdkdkdk CAT dkdkdk dkdkdk ORANGE dkdkdkdk DOG

Specificaly, it will not match the following as indicated by nfinium

CAT word1 word2 word3 word4 APPLE word5 word6 word7 DOG 

And also matches as you pointed out

CAT dkdkdkdkdk CAT dkdkdk dkdkdk ORANGE dkdkdkdk DOG 

You say that it should match the following

CAT substring12 DOG

but I dont think it should not since the CAT from above is the end delimiter of

PANDA substring11 CAT

This regex produces the expected result of nfinium

Note that as per the requirments of nfinium CAT can be a starting and an ending delimiter

CAT | DOG
PANDA | CAT

Answer 2

I think the key to this is the 2nd and output of:

 "APPLE substrin7 substring8 ORANGE" 

which is contained in:

 "CAT substring6 APPLE substring7 substring 8 ORANGE ORANGE substring9 DOG"

so bascially you have to catch Cat not followed by APPLE | ORANGE | LION | ZEBRA | PANDA | CAT as those would start another group. This is potentially possible but writing a regex to do this is akin to trying to parse HTML with a regex.

See: RegEx match open tags except XHTML self-contained tags

It could be done but the regex is going to be very complicated this problem is best handled in a code...

Here is an example of what I think you want that handles the first two start end combos.

(CAT(?!.+(?:APPLE|ORANGE|LION|ZEBRA|PANDA|CAT).+DOG).*?DOG)|(APPLE(?!.+(?:APPLE|LION|ZEBRA|PANDA|CAT|DOG).+ORANGE).*?ORANGE)

Just the first group is...

(CAT(?!.+(?:APPLE|ORANGE|LION|ZEBRA|PANDA|CAT).+DOG).*?DOG)

so you can see if this had more combinations it just gets very verbose.