Detailed Description

Regular expressions, or regexps, are a very powerful tool to handle strings and texts. This is useful in many contexts, e.g.,

This document is by no means a complete reference to pattern matching using regular expressions, and the following parts will require the reader to have some basic knowledge about Perl-like regular expressions and their pattern syntax.

Good references about regular expressions include:

• Mastering Regular Expressions (Third Edition) by Jeffrey E. F. Friedl, ISBN 0-596-52812-4;
• the pcrepattern(3) man page, describing the pattern syntax supported by PCRE (the reference implementation of Perl-compatible regular expressions);
• the Perl's regular expression documentation and the Perl's regular expression tutorial.

Introduction

QRegularExpression implements Perl-compatible regular expressions. It fully supports Unicode. For an overview of the regular expression syntax supported by QRegularExpression, please refer to the aforementioned pcrepattern(3) man page. A regular expression is made up of two things: a pattern string and a set of pattern options that change the meaning of the pattern string.

You can set the pattern string by passing a string to the QRegularExpression constructor:

QRegularExpression re("a pattern");

This sets the pattern string to a pattern. You can also use the setPattern() function to set a pattern on an existing QRegularExpression object:

QRegularExpression re;
re.setPattern("another pattern");

Note that due to C++ literal strings rules, you must escape all backslashes inside the pattern string with another backslash:

// matches two digits followed by a space and a word
QRegularExpression re("\\d\\d \\w+");

// matches a backslash
QRegularExpression re2("\\\\");

The pattern() function returns the pattern that is currently set for a QRegularExpression object:

QRegularExpression re("a third pattern");
QString pattern = re.pattern(); // pattern == "a third pattern"

Pattern Options

The meaning of the pattern string can be modified by setting one or more pattern options. For instance, it is possible to set a pattern to match case insensitively by setting the QRegularExpression::CaseInsensitiveOption.

You can set the options by passing them to the QRegularExpression constructor, as in:

// matches "Qt rocks", but also "QT rocks", "QT ROCKS", "qT rOcKs", etc.
QRegularExpression re("Qt rocks", QRegularExpression::CaseInsensitiveOption);

Alternatively, you can use the setPatternOptions() function on an existing QRegularExpressionObject:

QRegularExpression re("^\\d+$");
re.setPatternOptions(QRegularExpression::MultilineOption);
// re matches any line in the subject string that contains only digits (but at least one)

It is possible to get the pattern options currently set on a QRegularExpression object by using the patternOptions() function:

QRegularExpression re = QRegularExpression("^two.*words$", QRegularExpression::MultilineOption
                                                           | QRegularExpression::DotMatchesEverythingOption);

QRegularExpression::PatternOptions options = re.patternOptions();
// options == QRegularExpression::MultilineOption | QRegularExpression::DotMatchesEverythingOption

Please refer to the QRegularExpression::PatternOption enum documentation for more information about each pattern option.

Match Type and Match Options

The last two arguments of the match() and the globalMatch() functions set the match type and the match options. The match type is a value of the QRegularExpression::MatchType enum; the "traditional" matching algorithm is chosen by using the NormalMatch match type (the default). It is also possible to enable partial matching of the regular expression against a subject string: see the partial matching section for more details.

The match options are a set of one or more QRegularExpression::MatchOption values. They change the way a specific match of a regular expression against a subject string is done. Please refer to the QRegularExpression::MatchOption enum documentation for more details.

Normal Matching

In order to perform a match you can simply invoke the match() function passing a string to match against. We refer to this string as the subject string. The result of the match() function is a QRegularExpressionMatch object that can be used to inspect the results of the match. For instance:

// match two digits followed by a space and a word
QRegularExpression re("\\d\\d \\w+");
QRegularExpressionMatch match = re.match("abc123 def");
bool hasMatch = match.hasMatch(); // true

If a match is successful, the (implicit) capturing group number 0 can be used to retrieve the substring matched by the entire pattern (see also the section about extracting captured substrings):

QRegularExpression re("\\d\\d \\w+");
QRegularExpressionMatch match = re.match("abc123 def");
if (match.hasMatch()) {
    QString matched = match.captured(0); // matched == "23 def"
    // ...
}

It's also possible to start a match at an arbitrary offset inside the subject string by passing the offset as an argument of the match() function. In the following example "12 abc" is not matched because the match is started at offset 1:

QRegularExpression re("\\d\\d \\w+");
QRegularExpressionMatch match = re.match("12 abc 45 def", 1);
if (match.hasMatch()) {
    QString matched = match.captured(0); // matched == "45 def"
    // ...
}

Extracting captured substrings

The QRegularExpressionMatch object contains also information about the substrings captured by the capturing groups in the pattern string. The captured() function will return the string captured by the n-th capturing group:

QRegularExpression re("^(\\d\\d)/(\\d\\d)/(\\d\\d\\d\\d)$");
QRegularExpressionMatch match = re.match("08/12/1985");
if (match.hasMatch()) {
    QString day = match.captured(1); // day == "08"
    QString month = match.captured(2); // month == "12"
    QString year = match.captured(3); // year == "1985"
    // ...
}

Capturing groups in the pattern are numbered starting from 1, and the implicit capturing group 0 is used to capture the substring that matched the entire pattern.

It's also possible to retrieve the starting and the ending offsets (inside the subject string) of each captured substring, by using the capturedStart() and the capturedEnd() functions:

QRegularExpression re("abc(\\d+)def");
QRegularExpressionMatch match = re.match("XYZabc123defXYZ");
if (match.hasMatch()) {
    int startOffset = match.capturedStart(1); // startOffset == 6
    int endOffset = match.capturedEnd(1); // endOffset == 9
    // ...
}

All of these functions have an overload taking a QString as a parameter in order to extract named captured substrings. For instance:

QRegularExpression re("^(?<date>\\d\\d)/(?<month>\\d\\d)/(?<year>\\d\\d\\d\\d)$");
QRegularExpressionMatch match = re.match("08/12/1985");
if (match.hasMatch()) {
    QString date = match.captured("date"); // date == "08"
    QString month = match.captured("month"); // month == "12"
    QString year = match.captured("year"); // year == 1985
}

Global Matching

Global matching is useful to find all the occurrences of a given regular expression inside a subject string. Suppose that we want to extract all the words from a given string, where a word is a substring matching the pattern \w+.

QRegularExpression::globalMatch returns a QRegularExpressionMatchIterator, which is a Java-like forward iterator that can be used to iterate over the results. For instance:

QRegularExpression re("(\\w+)");
QRegularExpressionMatchIterator i = re.globalMatch("the quick fox");

Since it's a Java-like iterator, the QRegularExpressionMatchIterator will point immediately before the first result. Every result is returned as a QRegularExpressionMatch object. The hasNext() function will return true if there's at least one more result, and next() will return the next result and advance the iterator. Continuing from the previous example:

QStringList words;
while (i.hasNext()) {
    QRegularExpressionMatch match = i.next();
    QString word = match.captured(1);
    words << word;
}
// words contains "the", "quick", "fox"

You can also use peekNext() to get the next result without advancing the iterator.

It is possible to pass a starting offset and one or more match options to the globalMatch() function, exactly like normal matching with match().

Partial Matching

A partial match is obtained when the end of the subject string is reached, but more characters are needed to successfully complete the match. Note that a partial match is usually much more inefficient than a normal match because many optimizations of the matching algorithm cannot be employed.

A partial match must be explicitly requested by specifying a match type of PartialPreferCompleteMatch or PartialPreferFirstMatch when calling QRegularExpression::match or QRegularExpression::globalMatch. If a partial match is found, then calling the hasMatch() function on the QRegularExpressionMatch object returned by match() will return false, but hasPartialMatch() will return true.

When a partial match is found, no captured substrings are returned, and the (implicit) capturing group 0 corresponding to the whole match captures the partially matched substring of the subject string.

Note that asking for a partial match can still lead to a complete match, if one is found; in this case, hasMatch() will return true and hasPartialMatch() false. It never happens that a QRegularExpressionMatch reports both a partial and a complete match.

Partial matching is mainly useful in two scenarios: validating user input in real time and incremental/multi-segment matching.

Validating user input

Suppose that we would like the user to input a date in a specific format, for instance "MMM dd, yyyy". We can check the input validity with a pattern like:

^(Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec) \d\d?, \d\d\d\d$

(This pattern doesn't catch invalid days, but let's keep it for the example's purposes).

We would like to validate the input with this regular expression while the user is typing it, so that we can report an error in the input as soon as it is committed (for instance, the user typed the wrong key). In order to do so we must distinguish three cases:

• the input cannot possibly match the regular expression;
• the input does match the regular expression;
• the input does not match the regular expression right now, but it will if more characters will be added to it.

Note that these three cases represent exactly the possible states of a QValidator (see the QValidator::State enum).

In particular, in the last case we want the regular expression engine to report a partial match: we are successfully matching the pattern against the subject string but the matching cannot continue because the end of the subject is encountered. Notice, however, that the matching algorithm should continue and try all possibilities, and in case a complete (non-partial) match is found, then this one should be reported, and the input string accepted as fully valid.

This behaviour is implemented by the PartialPreferCompleteMatch match type. For instance:

QString pattern("^(Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec) \\d\\d?, \\d\\d\\d\\d$");
QRegularExpression re(pattern);

QString input("Jan 21,");
QRegularExpressionMatch match = re.match(input, 0, QRegularExpression::PartialPreferCompleteMatch);
bool hasMatch = match.hasMatch(); // false
bool hasPartialMatch = match.hasPartialMatch(); // true

If matching the same regular expression against the subject string leads to a complete match, it is reported as usual:

QString input("Dec 8, 1985");
QRegularExpressionMatch match = re.match(input, 0, QRegularExpression::PartialPreferCompleteMatch);
bool hasMatch = match.hasMatch(); // true
bool hasPartialMatch = match.hasPartialMatch(); // false

Another example with a different pattern, showing the behaviour of preferring a complete match over a partial one:

QRegularExpression re("abc\\w+X|def");
QRegularExpressionMatch match = re.match("abcdef", 0, QRegularExpression::PartialPreferCompleteMatch);
bool hasMatch = match.hasMatch(); // true
bool hasPartialMatch = match.hasPartialMatch(); // false
QString captured = match.captured(0); // captured == "def"

In this case, the subpattern abc\\w+X partially matches the subject string; however, the subpattern def matches the subject string completely, and therefore a complete match is reported.

If multiple partial matches are found when matching (but no complete match), then the QRegularExpressionMatch object will report the first one that is found. For instance:

QRegularExpression re("abc\\w+X|defY");
QRegularExpressionMatch match = re.match("abcdef", 0, QRegularExpression::PartialPreferCompleteMatch);
bool hasMatch = match.hasMatch(); // false
bool hasPartialMatch = match.hasPartialMatch(); // true
QString captured = match.captured(0); // captured == "abcdef"

Incremental/multi-segment matching

Incremental matching is another use case of partial matching. Suppose that we want to find the occurrences of a regular expression inside a large text (that is, substrings matching the regular expression). In order to do so we would like to "feed" the large text to the regular expression engines in smaller chunks. The obvious problem is what happens if the substring that matches the regular expression spans across two or more chunks.

In this case, the regular expression engine should report a partial match, so that we can match again adding new data and (eventually) get a complete match. This implies that the regular expression engine may assume that there are other characters beyond the end of the subject string. This is not to be taken literally -- the engine will never try to access any character after the last one in the subject.

QRegularExpression implements this behaviour when using the PartialPreferFirstMatch match type. This match type reports a partial match as soon as it is found, and other match alternatives are not tried (even if they could lead to a complete match). For instance:

QRegularExpression re("abc|ab");
QRegularExpressionMatch match = re.match("ab", 0, QRegularExpression::PartialPreferFirstMatch);
bool hasMatch = match.hasMatch(); // false
bool hasPartialMatch = match.hasPartialMatch(); // true

This happens because when matching the first branch of the alternation operator a partial match is found, and therefore matching stops, without trying the second branch. Another example:

QRegularExpression re("abc(def)?");
QRegularExpressionMatch match = re.match("abc", 0, QRegularExpression::PartialPreferFirstMatch);
bool hasMatch = match.hasMatch(); // false
bool hasPartialMatch = match.hasPartialMatch(); // true

This shows what could seem a counterintuitve behaviour of quantifiers: since ? is greedy, then the engine tries first to continue the match after having matched "abc"; but then the matching reaches the end of the subject string, and therefore a partial match is reported. This is even more surprising in the following example:

QRegularExpression re("(abc)*");
QRegularExpressionMatch match = re.match("abc", 0, QRegularExpression::PartialPreferFirstMatch);
bool hasMatch = match.hasMatch(); // false
bool hasPartialMatch = match.hasPartialMatch(); // true

It's easy to understand this behaviour if we remember that the engine expects the subject string to be only a substring of the whole text we're looking for a match into (that is, how we said before, that the engine assumes that there are other characters beyond the end of the subject string).

Since the * quantifier is greedy, then reporting a complete match could be an error, because after the current subject "abc" there may be other occurrences of "abc". For instance, the complete text could have been "abcabcX", and therefore the right match to report (in the complete text) would have been "abcabc"; by matching only against the leading "abc" we instead get a partial match.

Error Handling

It is possible for a QRegularExpression object to be invalid because of syntax errors in the pattern string. The isValid() function will return true if the regular expression is valid, or false otherwise:

QRegularExpression invalidRe("(unmatched|parenthesis");
bool isValid = invalidRe.isValid(); // false

You can get more information about the specific error by calling the errorString() function; moreover, the patternErrorOffset() function will return the offset inside the pattern string

QRegularExpression invalidRe("(unmatched|parenthesis");
if (!invalidRe.isValid()) {
    QString errorString = invalidRe.errorString(); // errorString == "missing )"
    int errorOffset = invalidRe.patternErrorOffset(); // errorOffset == 22
    // ...
}

If a match is attempted with an invalid QRegularExpression, then the returned QRegularExpressionMatch object will be invalid as well (that is, its isValid() function will return false). The same applies for attempting a global match.

Unsupported Perl-compatible Regular Expressions Features

QRegularExpression does not support all the features available in Perl-compatible regular expressions. The most notable one is the fact that duplicated names for capturing groups are not supported, and using them can lead to undefined behaviour.

This may change in a future version of Qt.

Notes for QRegExp Users

The QRegularExpression class introduced in Qt 5 is a big improvement upon QRegExp, in terms of APIs offered, supported pattern syntax and speed of execution. The biggest difference is that QRegularExpression simply holds a regular expression, and it's not modified when a match is requested. Instead, a QRegularExpressionMatch object is returned, in order to check the result of a match and extract the captured substring. The same applies with global matching and QRegularExpressionMatchIterator.

Other differences are outlined below.

Different pattern syntax

Porting a regular expression from QRegExp to QRegularExpression may require changes to the pattern itself.

In certain scenarios, QRegExp was too lenient and accepted patterns that are simply invalid when using QRegularExpression. These are somehow easy to detect, because the QRegularExpression objects built with these patterns are not valid (cf. isValid()).

In other cases, a pattern ported from QRegExp to QRegularExpression may silently change semantics. Therefore, it is necessary to review the patterns used. The most notable cases of silent incompatibility are:

• Curly braces are needed in order to use a hexadecimal escape like \xHHHH with more than 2 digits. A pattern like \x2022 neeeds to be ported to \x{2022}, or it will match a space (0x20) followed by the string "22". In general, it is highly recommended to always use curly braces with the \x escape, no matter the amount of digits specified.
• A 0-to-n quantification like {,n} needs to be ported to {0,n} to preserve semantics. Otherwise, a pattern such as \d{,3} would actually match a digit followed by the exact string "{,3}".
• QRegExp by default does Unicode-aware matching, while QRegularExpression requires a separate option; see below for more details.

Porting from QRegExp::exactMatch()

QRegExp::exactMatch() in Qt 4 served two purposes: it exactly matched a regular expression against a subject string, and it implemented partial matching.

Porting from QRegExp's Exact Matching

Exact matching indicates whether the regular expression matches the entire subject string. For example, the classes yield on the subject string "abc123":

Exact matching is not reflected in QRegularExpression. If you want to be sure that the subject string matches the regular expression exactly, you can wrap the pattern using the anchoredPattern() function:

QString p("a .*|pattern");

// re matches exactly the pattern string p
QRegularExpression re(QRegularExpression::anchoredPattern(p));

Porting from QRegExp's Partial Matching

When using QRegExp::exactMatch(), if an exact match was not found, one could still find out how much of the subject string was matched by the regular expression by calling QRegExp::matchedLength(). If the returned length was equal to the subject string's length, then one could conclude that a partial match was found.

QRegularExpression supports partial matching explicitly by means of the appropriate MatchType.

Global matching

Due to limitations of the QRegExp API it was impossible to implement global matching correctly (that is, like Perl does). In particular, patterns that can match 0 characters (like "a*") are problematic.

QRegularExpression::globalMatch() implements Perl global match correctly, and the returned iterator can be used to examine each result.

Unicode properties support

When using QRegExp, character classes such as \w, \d, etc. match characters with the corresponding Unicode property: for instance, \d matches any character with the Unicode Nd (decimal digit) property.

Those character classes only match ASCII characters by default when using QRegularExpression: for instance, \d matches exactly a character in the 0-9 ASCII range. It is possible to change this behaviour by using the UseUnicodePropertiesOption pattern option.

Wildcard matching

There is no direct way to do wildcard matching in QRegularExpression. However, the wildcardToRegularExpression method is provided to translate glob patterns into a Perl-compatible regular expression that can be used for that purpose.

Other pattern syntaxes

QRegularExpression supports only Perl-compatible regular expressions.

Minimal matching

QRegExp::setMinimal() implemented minimal matching by simply reversing the greediness of the quantifiers (QRegExp did not support lazy quantifiers, like *?, +?, etc.). QRegularExpression instead does support greedy, lazy and possessive quantifiers. The InvertedGreedinessOption pattern option can be useful to emulate the effects of QRegExp::setMinimal(): if enabled, it inverts the greediness of quantifiers (greedy ones become lazy and vice versa).

Caret modes

The AnchoredMatchOption match option can be used to emulate the QRegExp::CaretAtOffset behaviour. There is no equivalent for the other QRegExp::CaretMode modes.

Debugging Code that Uses QRegularExpression

QRegularExpression internally uses a just in time compiler (JIT) to optimize the execution of the matching algorithm. The JIT makes extensive usage of self-modifying code, which can lead debugging tools such as Valgrind to crash. You must enable all checks for self-modifying code if you want to debug programs using QRegularExpression (for instance, Valgrind's --smc-check command line option). The downside of enabling such checks is that your program will run considerably slower.

To avoid that, the JIT is disabled by default if you compile Qt in debug mode. It is possible to override the default and enable or disable the JIT usage (both in debug or release mode) by setting the QT_ENABLE_REGEXP_JIT environment variable to a non-zero or zero value respectively.