v/vlib/regex

V RegEx (Regular expression) 0.9h

[TOC]

introduction

Write here the introduction... not today!! -_-

Basic assumption

In this release, during the writing of the code some assumptions are made and are valid for all the features.

1. The matching stops at the end of the string not at the newline chars.
2. The basic elements of this regex engine are the tokens, in a query string a simple char is a token. The token is the atomic unit of this regex engine.

Match positional limiter

The module supports the following features:

• $ ^ delimiter

^ (Caret.) Matches at the start of the string

$ Matches at the end of the string

Tokens

The tokens are the atomic units used by this regex engine and can be ones of the following:

Simple char

this token is a simple single character like a.

Char class (cc)

The cc matches all the chars specified inside, it is delimited by square brackets [ ]

the sequence of chars in the class is evaluated with an OR operation.

For example, the following cc [abc] matches any char that is a or b or c but doesn't match C or z.

Inside a cc is possible to specify a "range" of chars, for example [ad-f] is equivalent to write [adef].

A cc can have different ranges at the same time like [a-zA-z0-9] that matches all the lowercase, uppercase and numeric chars.

It is possible negate the cc using the caret char at the start of the cc like: [^abc] that matches every char that is not a or b or c.

A cc can contain meta-chars like: [a-z\d] that matches all the lowercase latin chars a-z and all the digits \d.

It is possible to mix all the properties of the char class together.

Note: In order to match the - (minus) char, it must be located at the first position in the cc, for example [-_\d\a] will match - minus, _underscore, \d numeric chars, \a lower case chars.

Meta-chars

A meta-char is specified by a backslash before a char like \w in this case the meta-char is w.

A meta-char can match different type of chars.

• \w matches an alphanumeric char [a-zA-Z0-9_]
• \W matches a non alphanumeric char
• \d matches a digit [0-9]
• \D matches a non digit
• \smatches a space char, one of [' ','\t','\n','\r','\v','\f']
• \S matches a non space char
• \a matches only a lowercase char [a-z]
• \A matches only an uppercase char [A-Z]

Quantifier

Each token can have a quantifier that specify how many times the char can or must be matched.

Short quantifier

• ? matches 0 or 1 time, a?b matches both ab or b
• + matches at minimum 1 time, a+ matches both aaa or a
• * matches 0 or more time, a*b matches both aaab or ab or b

Long quantifier

• {x} matches exactly x time, a{2} matches aa but doesn't match aaa or a
• {min,} matches at minimum min time, a{2,} matches aaa or aa but doesn't match a
• {,max} matches at least 0 time and maximum max time, a{,2} matches a and aa but doesn't match aaa
• {min,max} matches from min times to max times, a{2,3} matches aa and aaa but doesn't match a or aaaa

a long quantifier may have a greedy off flag that is the ? char after the brackets, {2,4}? means to match the minimum number possible tokens in this case 2.

dot char

the dot is a particular meta char that matches "any char", is more simple explain it with an example:

suppose to have abccc ddeef as source string to parse with regex, the following table show the query strings and the result of parsing source string.

query string	result
.*c	abc
.*dd	abcc dd
ab.*e	abccc dde
ab.{3} .*e	abccc dde

the dot char matches any char until the next token match is satisfied.

OR token

the token | is a logic OR operation between two consecutive tokens, a|b matches a char that is a or b.

The OR token can work in a "chained way": a|(b)|cd test first a if the char is not a then test the group (b) and if the group doesn't match test the token c.

note: The OR work at token level! It doesn't work at concatenation level!

A query string like abc|bde is not equal to (abc)|(bde)!! The OR work only on c|b not at char concatenation level.

Groups

Groups are a method to create complex patterns with repetition of blocks of tokens.

The groups are delimited by round brackets ( ), groups can be nested and can have a quantifier as all the tokens.

c(pa)+z match cpapaz or cpaz or cpapapaz .

(c(pa)+z ?)+ matches cpaz cpapaz cpapapaz or cpapaz

let analyze this last case, first we have the group #0 that are the most outer round brackets (...)+, this group has a quantifier that say to match its content at least one time +.

After we have a simple char token c and a second group that is the number #1 :(pa)+, this group try to match the sequence pa at least one time as specified by the + quantifier.

After, we have another simple token z and another simple token ? that is the space char (ascii code 32) followed by the ? quantifier that say to capture the space char 0 or 1 time.

This explain because the (c(pa)+z ?)+ query string can match cpaz cpapaz cpapapaz .

In this implementation the groups are "capture groups", it means that the last temporal result for each group can be retrieved from the RE struct.

The "capture groups" are store as couple of index in the field groups that is an []int inside the RE struct.

example:

text := 'cpaz cpapaz cpapapaz'
query := r'(c(pa)+z ?)+'
mut re := regex.regex_opt(query) or { panic(err) }
println(re.get_query())
// #0(c#1(pa)+z ?)+  // #0 and #1 are the ids of the groups, are shown if re.debug is 1 or 2
start, end := re.match_string(text)
// [start=0, end=20]  match => [cpaz cpapaz cpapapaz]
mut gi := 0
for gi < re.groups.len {
	if re.groups[gi] >= 0 {
		println('${gi / 2} :[${text[re.groups[gi]..re.groups[gi + 1]]}]')
	}
	gi += 2
}
// groups captured
// 0 :[cpapapaz]
// 1 :[pa]

note: to show the group id number in the result of the get_query() the flag debug of the RE object must be 1 or 2

In order to simplify the use of the captured groups it possible to use the utility function: get_group_list.

This function return a list of groups using this support struct:

pub struct Re_group {
pub:
	start int = -1
	end   int = -1
}

Here an example of use:

/*
This simple function convert an HTML RGB value with 3 or 6 hex digits to an u32 value,
this function is not optimized and it si only for didatical purpose
example: #A0B0CC #A9F
*/
fn convert_html_rgb(in_col string) u32 {
	mut n_digit := if in_col.len == 4 { 1 } else { 2 }
	mut col_mul := if in_col.len == 4 { 4 } else { 0 }
	// this is the regex query, it use the V string interpolation to customize the regex query
	// NOTE: if you want use escaped code you must use the r"" (raw) strings,
	// *** please remember that the V interpoaltion doesn't work on raw strings. ***
	query := '#([a-fA-F0-9]{$n_digit})([a-fA-F0-9]{$n_digit})([a-fA-F0-9]{$n_digit})'
	mut re := regex.regex_opt(query) or { panic(err) }
	start, end := re.match_string(in_col)
	println('start: $start, end: $end')
	mut res := u32(0)
	if start >= 0 {
		group_list := re.get_group_list() // this is the utility function
		r := ('0x' + in_col[group_list[0].start..group_list[0].end]).int() << col_mul
		g := ('0x' + in_col[group_list[1].start..group_list[1].end]).int() << col_mul
		b := ('0x' + in_col[group_list[2].start..group_list[2].end]).int() << col_mul
		println('r: $r g: $g b: $b')
		res = u32(r) << 16 | u32(g) << 8 | u32(b)
	}
	return res
}

Groups Continuous saving

In particular situations it is useful have a continuous save of the groups, this is possible initializing the saving array field in RE struct: group_csave.

This feature allow to collect data in a continuous way.

In the example we pass a text followed by a integer list that we want collect. To achieve this task we can use the continuous saving of the group that save each captured group in a array that we set with: re.group_csave = [-1].repeat(3*20+1).

The array will be filled with the following logic:

re.group_csave[0] number of total saved records

re.group_csave[1+n*3] id of the saved group re.group_csave[1+n*3] start index in the source string of the saved group re.group_csave[1+n*3] end index in the source string of the saved group

The regex save until finish or found that the array have no space. If the space ends no error is raised, further records will not be saved.

fn example2() {
	test_regex()
	text := 'tst: 01,23,45 ,56, 78'
	query := r'.*:(\s*\d+[\s,]*)+'
	mut re := new() or { panic(err) }
	// re.debug = 2
	re.group_csave = [-1].repeat(3 * 20 + 1) // we expect max 20 records
	re.compile_opt(query) or {
		println(err)
		return
	}
	q_str := re.get_query()
	println('Query: $q_str')
	start, end := re.match_string(text)
	if start < 0 {
		println('ERROR : ${re.get_parse_error_string(start)}, $start')
	} else {
		println('found in [$start, $end] => [${text[start..end]}]')
	}
	// groups capture
	mut gi := 0
	for gi < re.groups.len {
		if re.groups[gi] >= 0 {
			println('${gi / 2} ${re.groups[gi]},${re.groups[gi + 1]} :[${text[re.groups[gi]..re.groups[gi +
				1]]}]')
		}
		gi += 2
	}
	// continuous saving
	gi = 0
	println('num: ${re.group_csave[0]}')
	for gi < re.group_csave[0] {
		id := re.group_csave[1 + gi * 3]
		st := re.group_csave[1 + gi * 3 + 1]
		en := re.group_csave[1 + gi * 3 + 2]
		println('cg id: $id [$st, $en] => [${text[st..en]}]')
		gi++
	}
}

The output will be:

Query: .*:(\s*\d+[\s,]*)+
found in [0, 21] => [tst: 01,23,45 ,56, 78]
0 19,21 :[78]
num: 5
cg id: 0 [4, 8] => [ 01,]
cg id: 0 [8, 11] => [23,]
cg id: 0 [11, 15] => [45 ,]
cg id: 0 [15, 19] => [56, ]
cg id: 0 [19, 21] => [78]

Named capturing groups

This regex module support partially the question mark ? PCRE syntax for groups.

(?:abcd) non capturing group: the content of the group will not be saved

(?P<mygroup>abcdef) named group: the group content is saved and labeled as mygroup

The label of the groups is saved in the group_map of the RE struct, this is a map from string to int where the value is the index in group_csave list of index.

Have a look at the example for the use of them.

example:

import regex

fn main() {
	test_regex()
	text := 'http://www.ciao.mondo/hello/pippo12_/pera.html'
	query := r'(?P<format>https?)|(?:ftps?)://(?P<token>[\w_]+.)+'
	mut re := new()
	re.debug = 2
	// must provide an array of the right size if want the continuos saving of the groups
	re.group_csave = [-1].repeat(3 * 20 + 1)
	re.compile_opt(query) or {
		println(err)
		return
	}
	q_str := re.get_query()
	println('O.Query: $query')
	println('Query  : $q_str')
	re.debug = 0
	start, end := re.match_string(text)
	if start < 0 {
		err_str := re.get_parse_error_string(start)
		println('ERROR : $err_str, $start')
	} else {
		text1 := text[start..end]
		println('found in [$start, $end] => [$text1]')
	}
	// groups
	mut gi := 0
	for gi < re.groups.len {
		if re.groups[gi] >= 0 {
			println('${gi / 2} ${re.groups[gi]},${re.groups[gi + 1]} :[${text[re.groups[gi]..re.groups[gi +
				1]]}]')
		}
		gi += 2
	}
	// continuous saving
	gi = 0
	println('num of group item saved: ${re.group_csave[0]}')
	for gi < re.group_csave[0] {
		id := re.group_csave[1 + gi * 3]
		st := re.group_csave[1 + gi * 3 + 1]
		en := re.group_csave[1 + gi * 3 + 2]
		println('cg id: $id [$st, $en] => [${text[st..en]}]')
		gi++
	}
	println('raw array: ${re.group_csave[0..gi * 3 + 2 - 1]}')
	// named capturing groups
	println('named capturing groups:')
	for g_name in re.group_map.keys() {
		s, e := re.get_group(g_name)
		if s >= 0 && e > s {
			println("'$g_name':[$s, $e] => '${text[s..e]}'")
		} else {
			println("Group [$g_name] doesn't exist.")
		}
	}
}

Output:

O.Query: (?P<format>https?)|(?:ftps?)://(?P<token>[\w_]+.)+
Query  : #0(?P<format>https?)|{8,14}(?:ftps?)://#1(?P<token>[\w_]+.)+
found in [0, 46] => [http://www.ciao.mondo/hello/pippo12_/pera.html]
0 0,4 :[http]
1 42,46 :[html]
num of group item saved: 8
cg id: 0 [0, 4] => [http]
cg id: 1 [7, 11] => [www.]
cg id: 1 [11, 16] => [ciao.]
cg id: 1 [16, 22] => [mondo/]
cg id: 1 [22, 28] => [hello/]
cg id: 1 [28, 37] => [pippo12_/]
cg id: 1 [37, 42] => [pera.]
cg id: 1 [42, 46] => [html]
raw array: [8, 0, 0, 4, 1, 7, 11, 1, 11, 16, 1, 16, 22, 1, 22, 28, 1, 28, 37, 1, 37, 42, 1, 42, 46]
named capturing groups:
'format':[0, 4] => 'http'
'token':[42, 46] => 'html'

In order to simplify the use of the named groups it possible to use names map in the re struct using the function re.get_group.

Here a more complex example of use:

// This function demostrate the use of the named groups
fn convert_html_rgb_n(in_col string) u32 {
	mut n_digit := if in_col.len == 4 { 1 } else { 2 }
	mut col_mul := if in_col.len == 4 { 4 } else { 0 }
	query := '#(?P<red>[a-fA-F0-9]{$n_digit})(?P<green>[a-fA-F0-9]{$n_digit})(?P<blue>[a-fA-F0-9]{$n_digit})'
	mut re := regex.regex_opt(query) or { panic(err) }
	start, end := re.match_string(in_col)
	println('start: $start, end: $end')
	mut res := u32(0)
	if start >= 0 {
		red_s, red_e := re.get_group('red')
		r := ('0x' + in_col[red_s..red_e]).int() << col_mul
		green_s, green_e := re.get_group('green')
		g := ('0x' + in_col[green_s..green_e]).int() << col_mul
		blue_s, blue_e := re.get_group('blue')
		b := ('0x' + in_col[blue_s..blue_e]).int() << col_mul
		println('r: $r g: $g b: $b')
		res = u32(r) << 16 | u32(g) << 8 | u32(b)
	}
	return res
}

Flags

It is possible to set some flags in the regex parser that change the behavior of the parser itself.

// example of flag settings
mut re := regex.new()
re.flag = regex.F_BIN

• F_BIN: parse a string as bytes, utf-8 management disabled.

• F_EFM: exit on the first char matches in the query, used by the find function.

• F_MS: matches only if the index of the start match is 0, same as ^ at the start of the query string.

• F_ME: matches only if the end index of the match is the last char of the input string, same as $ end of query string.

• F_NL: stop the matching if found a new line char \n or \r

Functions

Initializer

These functions are helper that create the RE struct, a RE struct can be created manually if you needed.

Simplified initializer

// regex create a regex object from the query string and compile it
pub fn regex_opt(in_query string) ?RE

Base initializer

// new_regex create a REgex of small size, usually sufficient for ordinary use
pub fn new() RE

// new_regex_by_size create a REgex of large size, mult specify the scale factor of the memory that will be allocated
pub fn new_by_size(mult int) RE

After a base initializer is used, the regex expression must be compiled with:

// compile compiles the REgex returning an error if the compilation fails
pub fn (re mut RE) compile_opt(in_txt string) ?

Operative Functions

These are the operative functions

// match_string try to match the input string, return start and end index if found else start is -1
pub fn (re mut RE) match_string(in_txt string) (int,int)

// find try to find the first match in the input string, return start and end index if found else start is -1
pub fn (re mut RE) find(in_txt string) (int,int)

// find_all find all the "non overlapping" occurrences of the matching pattern, return a list of start end indexes
pub fn (re mut RE) find_all(in_txt string) []int

// replace return a string where the matches are replaced with the replace string, only non overlapped matches are used
pub fn (re mut RE) replace(in_txt string, repl string) string

Debugging

This module has few small utilities to help the writing of regex expressions.

Syntax errors highlight

the following example code show how to visualize the syntax errors in the compilation phase:

query := r'ciao da ab[ab-]'
// there is an error, a range not closed!!
mut re := new()
re.compile_opt(query) or { println(err) }
// output!!
// query: ciao da ab[ab-]
// err  : ----------^
// ERROR: ERR_SYNTAX_ERROR

Compiled code

It is possible to view the compiled code calling the function get_query(). The result will be something like this:

========================================
v RegEx compiler v 0.9c output:
PC:  0 ist: 7fffffff [a]      query_ch {  1,  1}
PC:  1 ist: 7fffffff [b]      query_ch {  1,MAX}
PC:  2 ist: 88000000 PROG_END {  0,  0}
========================================

PC:int is the program counter or step of execution, each single step is a token.

ist:hex is the token instruction id.

[a] is the char used by the token.

query_ch is the type of token.

{m,n} is the quantifier, the greedy off flag ? will be showed if present in the token

Log debug

The log debugger allow to print the status of the regex parser when the parser is running.

It is possible to have two different level of debug: 1 is normal while 2 is verbose.

here an example:

normal

list only the token instruction with their values

// re.flag = 1 // log level normal
flags: 00000000
#   2 s:     ist_load PC:   0=>7fffffff i,ch,len:[  0,'a',1] f.m:[ -1, -1] query_ch: [a]{1,1}:0 (#-1)
#   5 s:     ist_load PC:   1=>7fffffff i,ch,len:[  1,'b',1] f.m:[  0,  0] query_ch: [b]{2,3}:0? (#-1)
#   7 s:     ist_load PC:   1=>7fffffff i,ch,len:[  2,'b',1] f.m:[  0,  1] query_ch: [b]{2,3}:1? (#-1)
#  10 PROG_END

verbose

list all the instructions and states of the parser

flags: 00000000
#   0 s:        start PC: NA
#   1 s:     ist_next PC: NA
#   2 s:     ist_load PC:   0=>7fffffff i,ch,len:[  0,'a',1] f.m:[ -1, -1] query_ch: [a]{1,1}:0 (#-1)
#   3 s:  ist_quant_p PC:   0=>7fffffff i,ch,len:[  1,'b',1] f.m:[  0,  0] query_ch: [a]{1,1}:1 (#-1)
#   4 s:     ist_next PC: NA
#   5 s:     ist_load PC:   1=>7fffffff i,ch,len:[  1,'b',1] f.m:[  0,  0] query_ch: [b]{2,3}:0? (#-1)
#   6 s:  ist_quant_p PC:   1=>7fffffff i,ch,len:[  2,'b',1] f.m:[  0,  1] query_ch: [b]{2,3}:1? (#-1)
#   7 s:     ist_load PC:   1=>7fffffff i,ch,len:[  2,'b',1] f.m:[  0,  1] query_ch: [b]{2,3}:1? (#-1)
#   8 s:  ist_quant_p PC:   1=>7fffffff i,ch,len:[  3,'b',1] f.m:[  0,  2] query_ch: [b]{2,3}:2? (#-1)
#   9 s:     ist_next PC: NA
#  10 PROG_END
#  11 PROG_END

the columns have the following meaning:

# 2 number of actual steps from the start of parsing

s: ist_next state of the present step

PC: 1 program counter of the step

=>7fffffff hex code of the instruction

i,ch,len:[ 0,'a',1] i index in the source string, ch the char parsed, len the length in byte of the char parsed

f.m:[ 0, 1] f index of the first match in the source string, m index that is actual matching

query_ch: [b] token in use and its char

{2,3}:1? quantifier {min,max}, :1 is the actual counter of repetition, ? is the greedy off flag if present.

Custom Logger output

The debug functions output uses the stdout as default, it is possible to provide an alternative output setting a custom output function:

// custom print function, the input will be the regex debug string
fn custom_print(txt string) {
	println('my log: $txt')
}

mut re := new()
re.log_func = custom_print
// every debug output from now will call this function

Example code

Here there is a simple code to perform some basically match of strings

struct TestObj {
	source string // source string to parse
	query  string // regex query string
	s      int // expected match start index
	e      int // expected match end index
}

const (
	tests = [
		TestObj{'this is a good.', r'this (\w+) a', 0, 9},
		TestObj{'this,these,those. over', r'(th[eio]se?[,. ])+', 0, 17},
		TestObj{'test1@post.pip.com, pera', r'[\w]+@([\w]+\.)+\w+', 0, 18},
		TestObj{'cpapaz ole. pippo,', r'.*c.+ole.*pi', 0, 14},
		TestObj{'adce aabe', r'(a(ab)+)|(a(dc)+)e', 0, 4},
	]
)

fn example() {
	for c, tst in tests {
		mut re := regex.new()
		re.compile_opt(tst.query) or {
			println(err)
			continue
		}
		// print the query parsed with the groups ids
		re.debug = 1 // set debug on at minimum level
		println('#${c:2d} query parsed: $re.get_query()')
		re.debug = 0
		// do the match
		start, end := re.match_string(tst.source)
		if start >= 0 && end > start {
			println('#${c:2d} found in: [$start, $end] => [${tst.source[start..end]}]')
		}
		// print the groups
		mut gi := 0
		for gi < re.groups.len {
			if re.groups[gi] >= 0 {
				println('group ${gi / 2:2d} :[${tst.source[re.groups[gi]..re.groups[gi + 1]]}]')
			}
			gi += 2
		}
		println('')
	}
}

fn main() {
	example()
}

more example code is available in the test code for the regex module vlib\regex\regex_test.v.