How to split a command line like string?

Question

Basically, I need to split the string like

"one quoted argument" those are separate arguments "but not \"this one\""

to get in result the list of arguments

"one quoted argument"
those
are
separate
"but not \"this one\""

This regex "(\"|[^"])*"|[^ ]+ nearly does the job but the issue is that regular expression always (at least in java) tries to match the longest string possible.

In consequence, when I apply the regex to a string that starts and ends with a quoted arguments, it matches the whole string and does not create a group for each argument.

Is there a way to tweak this regex or the matcher or the pattern or whatever to handle that?

Note: don't tell me I could use GetOpt or CommandLine.parse or anything else similar.
My concern is about pure java regex (if possible but I doubt it...).

I don't think it's possible with regular expressions, but I may be wrong. — arshajii, Nov 21 '12 at 14:33
What about using `*?` so that the regular expression is not greedy. `"(\\"|[^"])*?"|[^ ]+` matches what you need. — Alex, Nov 21 '12 at 14:34
How should it handle `"one quoted argument" those are separate arguments "but not \\"this one\\""`? — Hans Then, Nov 21 '12 at 14:37
A. R. S. is correct, you can't use Java (Perl 5 Compatible) regular expressions to parse the full set of possible command lines. The reason is that the escaping you're referring to can nest recursively. You will either need a parser, or a different regular expression engine (see the regex system in newer versions of Perl, which can do this). — Kyle Burton, Nov 21 '12 at 14:41

score 4 · Answer 1 · answered Nov 21 '12 at 14:37

regular expression always (at least in java) tries to match the longest string possible.

Um... no.

That is controlled by if you use greedy or non-greedy expressions. See some examples. Using a non-greedy one (by adding a question mark) should do it. It's called lazy quantification.

The default is greedy, but it certainly doesn't mean it is always that way.

score 4 · Accepted Answer · answered Nov 21 '12 at 14:40

4

You may use the non greedy qualifier *? to make it work:

"(\\"|[^"])*?"|[^ ]+

See this link for an example in action: http://gskinner.com/RegExr/?32srs

answered Nov 21 '12 at 14:40

Alex

25,147
6
59
55

1

Suggestion: replace `[^ ]+` with `[^\s]+`, so it can handle tabs or other whitespace characters – Rui Jarimba Nov 21 '12 at 15:08

score 2 · Answer 3 · answered Aug 29 '13 at 07:23

2

public static String[] parseCommand( String cmd )
{
    if( cmd == null || cmd.length() == 0 )
    {
        return new String[]
        {};
    }

    cmd = cmd.trim();
    String regExp = "\"(\\\"|[^\"])*?\"|[^ ]+";
    Pattern pattern = Pattern.compile( regExp, Pattern.MULTILINE | Pattern.CASE_INSENSITIVE );
    Matcher matcher = pattern.matcher( cmd );
    List< String > matches = new ArrayList< String >();
    while( matcher.find() ) {
        matches.add( matcher.group() );
    }
    String[] parsedCommand = matches.toArray(new String[] {});
    return parsedCommand;
}

answered Aug 29 '13 at 07:23

Igor

21
1

Maybe throw NPE on null ? Also you should cache the pattern as per [Markus answer](http://stackoverflow.com/a/22472588/281545) - and `cmd.length() == 0` will be taken care by the regex – Mr_and_Mrs_D Jul 07 '14 at 13:53
@Igor this is a great start at the problem, but doesn't work for the last case. – vallentin Apr 04 '16 at 12:21

score 2 · Answer 4 · edited Jul 06 '14 at 22:46

I came up with this one (thanks Alex for giving me the good starting point :))

/**
 * Pattern that is capable of dealing with complex command line quoting and
 * escaping. This can recognize correctly:
 * <ul>
 * <li>"double quoted strings"
 * <li>'single quoted strings'
 * <li>"escaped \"quotes within\" quoted string"
 * <li>C:\paths\like\this or "C:\path like\this"
 * <li>--arguments=like_this or "--args=like this" or '--args=like this' or
 * --args="like this" or --args='like this'
 * <li>quoted\ whitespaces\\t (spaces & tabs)
 * <li>and probably more :)
 * </ul>
 */
private static final Pattern cliCracker = Pattern
    .compile(
       "[^\\s]*\"(\\\\+\"|[^\"])*?\"|[^\\s]*'(\\\\+'|[^'])*?'|(\\\\\\s|[^\\s])+",
       Pattern.MULTILINE);

Will be using it and test. FWIW here is the implementation of [translateCommandline](https://commons.apache.org/proper/commons-exec/apidocs/src-html/org/apache/commons/exec/CommandLine.html) - line 337 - see also http://stackoverflow.com/questions/3259143/split-a-string-containing-command-line-parameters-into-a-string-in-java — Mr_and_Mrs_D, Jul 07 '14 at 13:50

How to split a command line like string?

4 Answers4