Find RegEx for Pattern

Question

I am used to write regular expression and do understand it however I tried hard to generate regex for below pattern, but finally need help from you guys.

CREATE FUNCTION ...
...
...
BEGIN

  DECLARE ...
  DECLARE ...
...
END

What I am looking for is to extract 3 groups using Java RegEx:

1st section is from CREATE FUNCITON to BEGIN,

2nd is statements starting with DECLARE and

3rd one is rest everything after DECLARE statements to END.

(See below)

---------------------
CREATE FUNCTION ...
...
...
BEGIN
---------------------
  DECLARE ...
  DECLARE ...
---------------------
...
END
---------------------

Unless you control the source you are trying to parse, regexps are not the best tool. — Thorbjørn Ravn Andersen, Jun 13 '19 at 06:49
Possible duplicate of [Learning Regular Expressions](https://stackoverflow.com/questions/4736/learning-regular-expressions) — Biffen, Jun 13 '19 at 06:52
@Biffen, see 1st comment. parsing source code is a task beyond the capabilities of regex — Sharon Ben Asher, Jun 13 '19 at 06:53
@ThorbjørnRavnAndersen Why can't I use Group of RegEx to extract this ? Can you please elaborate your comment a bit ? — Pratik, Jun 13 '19 at 06:54
@Pratik. think of cases where a function is called "FUNCTION" or cases where a variable is initialized with "BEGIN" string, etc — Sharon Ben Asher, Jun 13 '19 at 07:00
@SharonBenAsher I can assure that we can avoid such corner cases, there won't be any such case. — Pratik, Jun 13 '19 at 07:02
@Pratik, you should take a look into language parsers. the most popular one is [ANTLR](https://www.antlr.org/), but there are [lots of others](https://tomassetti.me/parsing-in-java/). — Sharon Ben Asher, Jun 13 '19 at 07:02

score 1 · Accepted Answer · answered Jun 13 '19 at 07:02

Unless the text always comes in a very regular fashion, regex is not a suitable for parsing source code like this. You should write/use a parser instead.

Assuming:

There will be no DECLARE statements between CREATE FUNCTION and BEGIN
There will not be any other statements in between the DECLARE statements
There will always be at least one DECLARE statement

You can use a regex like this:

([\s\S]+?)((?:\s+DECLARE.+)+)([\s\S]+)

Note that this can cause catastrophic backtracking if there are no DECLARE statements. If the input could have no DECLARE statements, you could just match the (?:\s+DECLARE.+)+ part, and get the other two groups with substring:

String input = "CREATE FUNCTION ...\n" +
        "...\n" +
        "...\n" +
        "BEGIN\n" +
        "\n" +
        "  DECLARE ...\n" +
        "  DECLARE ...\n" +
        "...\n" +
        "END";
Matcher m = Pattern.compile("(?:\\s+DECLARE.+)+").matcher(input);
if (m.find()) {
    String firstPart = input.substring(0, m.start());
    String secondPart = m.group();
    String thirdPart = input.substring(m.end());
    System.out.println(thirdPart);
}

@Sweepwe, Now a days its a trend to downvote questions on stackoverflow very easily, even if you try to explain it very decently. I did spend so many ours behind this but may be I was not very good with backreferencing and lookahead operators. But I really appreciate your answer and the efforts you put in for me. Thank you very much again :) — Pratik, Jun 13 '19 at 07:10

Find RegEx for Pattern

1 Answers1