pcregrep searches files for character patterns, in the same way as other grep commands do, but it uses the PCRE regular expression library to support patterns that are compatible with the regular expressions of Perl 5. See pcresyntax(3) for a quick-reference summary of pattern syntax, or pcrepattern(3) for a full description of the syntax and semantics of the regular expressions that PCRE supports.

Patterns, whether supplied on the command line or in a separate file, are given without delimiters. For example:

pcregrep Thursday /etc/motd

If you attempt to use delimiters (for example, by surrounding a pattern with slashes, as is common in Perl scripts), they are interpreted as part of the pattern. Quotes can of course be used to delimit patterns on the command line because they are interpreted by the shell, and indeed quotes are required if a pattern contains white space or shell metacharacters.

The first argument that follows any option settings is treated as the single pattern to be matched when neither -e nor -f is present. Conversely, when one or both of these options are used to specify patterns, all arguments are treated as path names. At least one of -e, -f, or an argument pattern must be provided.

If no files are specified, pcregrep reads the standard input. The standard input can also be referenced by a name consisting of a single hyphen. For example:

pcregrep some-pattern /file1 - /file3

By default, each line that matches a pattern is copied to the standard output, and if there is more than one file, the file name is output at the start of each line, followed by a colon. However, there are options that can change how pcregrep behaves. In particular, the -M option makes it possible to search for patterns that span line boundaries. What defines a line boundary is controlled by the -N (--newline) option.

The amount of memory used for buffering files that are being scanned is controlled by a parameter that can be set by the --buffer-size option. The default value for this parameter is specified when pcregrep is built, with the default default being 20K. A block of memory three times this size is used (to allow for buffering "before" and "after" lines). An error occurs if a line overflows the buffer.

Patterns can be no longer than 8K or BUFSIZ bytes, whichever is the greater. BUFSIZ is defined in . When there is more than one pattern (specified by the use of -e and/or -f), each pattern is applied to each line in the order in which they are defined, except that all the -e patterns are tried before the -f patterns.

By default, as soon as one pattern matches a line, no further patterns are considered. However, if --colour (or --color) is used to colour the matching substrings, or if --only-matching, --file-offsets, or --line-offsets is used to output only the part of the line that matched (either shown literally, or as an offset), scanning resumes immediately following the match, so that further matches on the same line can be found. If there are multiple patterns, they are all tried on the remainder of the line, but patterns that follow the one that matched are not tried on the earlier part of the line.

This behaviour means that the order in which multiple patterns are specified can affect the output when one of the above options is used. This is no longer the same behaviour as GNU grep, which now manages to display earlier matches for later patterns (as long as there is no overlap).

Patterns that can match an empty string are accepted, but empty string matches are never recognized. An example is the pattern "(super)?(man)?", in which all components are optional. This pattern finds all occurrences of both "super" and "man"; the output differs from matching with "super|man" when only the matching substrings are being shown.

If the LCALL or LCCTYPE environment variable is set, pcregrep uses the value to set a locale when calling the PCRE library. The --locale option can be used to override this.