olivercalder/perg

Grep Regular Expressions: Perg

Perg

Grep Regular Expressions: Perg

About Perg

Perg is a regular expression matching tool built from scratch in C, leveraging nondeterministic finite automata and multithreading.

Expression Syntax

The syntax recognized by perg is inspired by that which is common in computability theory (not the POSIX standard). In particular, the following are metacharacters:

Metachar	Description
( )	Denotes a contained subexpression to which other metacharacters may be applied.
|	The union operator matches either the previous or next expression: i.e. `abc
.	Matches any single symbol.
*	Matches zero or more occurrences of the previous symbol or subexpression: i.e. ab*c matches both ac and abbbbc.
!	Matches any symbol except the next symbol or subexpression: i.e. a!bc matches atc but not abc.
?	Matches zero or one occurrence of the previous symbol or subexpression: i.e. ab?c matches both abc and ac.

Special characters which are not metacharacters, including \t (tab) and \\ (backslash), standard C char symbols are used. To include any metacharacter as its literal symbol, precede it with a backslash, as in 'hello \(greeting\)'.

In the future, additional metacharacters may be supported, including:

Metachar	Description
+	Matches the previous symbol or subexpression one or more times ($s+$ is equivalent to $ss*$).
**n	Matches exactly $n$ occurrences of the previous symbol or subexpression, $n \in \mathbb{Z}^+$.

Additionally, applying ! to subexpressions may be supported once such matching behavior has been defined.

Usage

perg [OPTION...] EXPRESSION [FILE...]

If no file is specified and -r flag is not present, matches expression against stdin.

Options are heavily (and selectively) inspired by those of grep for the benefit of muscle memory. Most are not yet implemented, but they will include:

Option	Description
-i	Ignore case distinctions.
-v	Invert matches to select non-matching lines.
-w	Match whole words, where matching strings must be surrounded by whitespace.
-x	Match whole lines, where the matching string must be the complete line.
-o	Only print the matching parts of a matching line.
-H	Print the filename before each match -- this is the default when multiple files or -r is given.
-h	Suppress printing the filename before each match -- this is the default when one or zero files are given, and -r is not present.
-n	Prefix each matching line with the line number within the input file, after the filename if applicable.
-a	Treat binary files as if they were text files.
-r	Recursively read all files in each given directory and subdirectories -- if no files given, searches the working directory.

Other lower-priority options to be (possibly) implemented later: -c, -L, -l, -q, -A, -B, -C