generating a regular expression from a string

Question

I wish to generate a regular expression from a string containing numbers, and then use this as a Pattern to search for similar strings. Example:

String s = "Page 3 of 23"

If I substitute all digits by \d

    StringBuilder sb = new StringBuilder();
    for (int i = 0; i < s.length(); i++) {
    char c = s.charAt(i);
    if (Character.isDigit(c)) {
        sb.append("\\d"); // backslash d
    } else {
        sb.append(c);
        }
    }

    Pattern numberPattern = Pattern.compile(sb.toString());

//    Pattern numberPattern = Pattern.compile("Page \d of \d\d");

I can use this to match similar strings (e.g. "Page 7 of 47"). My problem is that if I do this naively some of the metacharacters such as (){}-, etc. will not be escaped. Is there a library to do this or an exhaustive set of characters for regular expressions which I must and must not escape? (I can try to extract them from the Javadocs but am worried about missing something).

Alternatively is there a library which already does this (I don't at this stage want to use a full Natural Language Processing solution).

NOTE: @dasblinkenlight's edited answer now works for me!

Here's an answer to the which characters question, I'm not aware of any libraries to generate regexs though: http://stackoverflow.com/questions/399078/what-special-characters-must-be-escaped-in-regular-expressions — Evan Knowles, Apr 16 '13 at 10:16
@Evan thanks. I am only interested in Java so that looks like a useful resource. — peter.murray.rust, Apr 16 '13 at 10:18

Sergey Kalinichenko · Accepted Answer · 2013-04-16T13:05:06.440

10

Java's regexp library provides this functionality:

String s = Pattern.quote(orig);

The "quoted" string will have all its metacharacters escaped. First, escape your string, and then go through it and replace digits by \d to make a regular expression. Since regex library uses \Q and \E for quoting, you need to enclose your portion of regex in inverse quotes of \E and \Q.

One thing I would change in your implementation is the replacement algorithm: rather than replacing character-by-character, I would replace digits in groups. This would let an expression produced from Page 3 of 23 match strings like Page 13 of 23 and Page 6 of 8.

String p = Pattern.quote(orig).replaceAll("\\d+", "\\\\E\\\\d+\\\\Q");

This would produce "\QPage \E\d+\Q of \E\d+\Q\E" no matter what page numbers and counts were there originally. The output needs only one, not two slashes in \d, because the result is fed directly to regex engine, bypassing the Java compiler.

edited Apr 16 '13 at 13:05

answered Apr 16 '13 at 10:18

Sergey Kalinichenko

714,442
84
1,110
1,523

Cool, I didn't know about this method. – toniedzwiedz Apr 16 '13 at 10:21
@dasblinkenlight Great! Agreed I might look for repeated digits but there is also heuristics value for me in having exact digit counts. I may use both approaches. – peter.murray.rust Apr 16 '13 at 10:27
@peter.murray.rust See the last edit: the number of slashes required to make two slashes in the output is really ridiculous - times two for the compiler and times two for the regex library, for the total of eight slashes. – Sergey Kalinichenko Apr 16 '13 at 10:29
1

@dasblinkenlight. Which is why I am grateful to you for creating it! Takes me a long time to get the number right! (I tend to use constants to help break it down) – peter.murray.rust Apr 16 '13 at 10:32
I think this answer has everything I need and can be extended so I have accepted it – peter.murray.rust Apr 16 '13 at 10:33
NOTE: I was surprised by the output. `Pattern.quote("3 - 45")` gives `\Q3 - 45\E` . This is quoting the whole string. I am not sure whether this would not then destroy the replacements – peter.murray.rust Apr 16 '13 at 11:17
@peter.murray.rust That's why I wrote that you need to first escape the string, and only then go through with the replacements. – Sergey Kalinichenko Apr 16 '13 at 11:20
@dasblinkenlight So \Q ... \E doesn't turn off matching against \\d+ for example? – peter.murray.rust Apr 16 '13 at 11:30
@peter.murray.rust You're right, it will... Let me look into this to see if there's something better that may work. – Sergey Kalinichenko Apr 16 '13 at 11:32
@dasblinkenlight I like the current solution. I will test it now – peter.murray.rust Apr 16 '13 at 12:40
@dasblinkenlight I find I need `replaceAll("\\d+", "\\\\E\\\\d+\\\\Q")`; – peter.murray.rust Apr 16 '13 at 12:46

generating a regular expression from a string

1 Answers1

Linked