SYNOPSIS

    use Text::RewriteRules;

    RULES email
    \.==> DOT
    @==> AT
    ENDRULES

    print email("[email protected]") # prints ambs AT cpan DOT org

    RULES/m inc
    (\d+)=e=> $1+1
    ENDRULES

    print inc("I saw 11 cats and 23 dogs") # prints I saw 12 cats and 24 dogs

ABSTRACT

This module uses a simplified syntax for regexp-based rules for rewriting text. You define a set of rules, and the system applies them until no more rule can be applied.

Two variants are provided:

1.

traditional rewrite (\s-1RULES\s0 function): while it is possible do substitute | apply first substitution rule

2.

cursor based rewrite (RULES/m function): add a cursor to the beginning of the string while not reach end of string | apply substitute just after cursor and advance cursor | or advance cursor if no rule can be applied

DESCRIPTION

A lot of computer science problems can be solved using rewriting rules.

Rewriting rules consist of mainly two parts: a regexp (\s-1LHS:\s0 Left Hand Side) that is matched with the text, and the string to use to substitute the content matched with the regexp (\s-1RHS:\s0 Right Hand Side).

Now, why don't use a simple substitute? Because we want to define a set of rules and match them again and again, until no more regexp of the \s-1LHS\s0 matches.

A point of discussion is the syntax to define this system. A brief discussion shown that some users would prefer a function to receive an hash with the rules, some other, prefer some syntax sugar.

The approach used is the last: we use \*(C`Filter::Simple\*(C' such that we can add a specific non-perl syntax inside the Perl script. This improves legibility of big rewriting rules systems.

This documentation is divided in two parts: first we will see the reference of the module. Kind of, what it does, with a brief explanation. Follows a tutorial which will be growing through time and releases.

SYNTAX REFERENCE

Note: most of the examples are very stupid, but that is the easiest way to explain the basic syntax.

The basic syntax for the rewrite rules is a block, started by the keyword \*(C`RULES\*(C' and ended by the \*(C`ENDRULES\*(C'. Everything between them is handled by the module and interpreted as rules or comments.

The \*(C`RULES\*(C' keyword can handle a set of flags (we will see that later), and requires a name for the rule-set. This name will be used to define a function for that rewriting system.

RULES functioname ... ENDRULES

The function is defined in the main namespace where the \*(C`RULES\*(C' block appears.

In this block, each line can be a comment (Perl style), an empty line or a rule.

Basic Rule

A basic rule is a simple substitution:

RULES foobar foo==>bar ENDRULES

The arrow \*(C`==>\*(C' is used as delimiter. At its left is the regexp to match, at the right side, the substitution. So, the previous block defines a \*(C`foobar\*(C' function that substitutes all \*(C`foo\*(C' by \*(C`bar\*(C'.

Although this can seems similar to a global substitution, it is not. With a global substitution you can't do an endless loop. With this module it is very simple. I know you will get the idea.

You can use the syntax of Perl both on the left and right hand side of the rule, including \*(C`$1...\*(C'.

Execution Rule

If the Perl substitution supports execution, why not to support it, also? So, you got the idea. Here is an example:

RULES foo (\d+)b=e=>'b' x $1 (\d+)a=eval=>'a' x ($1*2) ENDRULES

So, for any number followed by a \*(C`b\*(C', we replace by that number of \*(C`b's\*(C'. For each number followed by an \*(C`a\*(C', we replace them by twice that number of \*(C`a's\*(C'.

Also, you mean evaluation using an \*(C`e\*(C' or \*(C`eval\*(C' inside the arrow. I should remind you can mix all these rules together in the same rewriting system.

Conditional Rule

On some cases we want to perform a substitution if the pattern matches and a set of conditions about that pattern (or not) are true.

For that, we use a three part rule. We have the common rule plus the condition part, separated from the rule by \*(C`!!\*(C'. These conditional rules can be applied both for basic and execution rules.

RULES translate ([[:alpha:]]+)=e=>$dic{$1}!! exists($dic{$1}) ENDRULES

The previous example would translate all words that exist on the dictionary.

Begin Rule

Sometimes it is useful to change something on the string before starting to apply the rules. For that, there is a special rule named \*(C`begin\*(C' (or \*(C`b\*(C' for abbreviate) just with a \s-1RHS\s0. This \s-1RHS\s0 is Perl code. Any Perl code. If you want to modify the string, use $_.

RULES foo =b=> $_.=" END" ENDRULES

Last Rule

As you use \*(C`last\*(C' on Perl to skip the remaining code on a loop, you can also call a \*(C`last\*(C' (or \*(C`l\*(C') rule when a specific pattern matches.

Like the \*(C`begin\*(C' rule with only a \s-1RHS\s0, the \*(C`last\*(C' rule has only a \s-1LHS:\s0

RULES foo foobar=l=> ENDRULES

This way, the rules iterate until the string matches with \*(C`foobar\*(C'.

You can also supply a condition in a last rule:

RULES bar f(o+)b(a+)r=l=> !! length($1) == 2 * length($2);

Rules with /x mode

It is possible to use the regular expressions /x mode in the rewrite rules. In this case:

1.

there must be an empty line between rules

2.

you can insert space and line breaks into the regular expression: RULES/x f1 (\d+) (\d{3}) (000) ==>$1 milhao e $2 mil!! $1 == 1

ENDRULES

POWER EXPRESSIONS

To facilitate matching complex languages Text::RewriteRules defines a set of regular expressions that you can use (without defining them).

Parenthesis

There are three kind of usual parenthesis: the standard parenthesis, brackets or curly braces. You can match a balanced string of parenthesis using the power expressions \*(C`[[:PB:]]\*(C', \*(C`[[:BB:]]\*(C' and \*(C`[[:CBB:]]\*(C' for these three kind of parenthesis.

For instance, if you apply this rule:

[[:BB:]]==>foo

to this string

something [ a [ b] c [d ]] and something more

then, you will get

something foo and something more

Note that if you apply it to

something [[ not ] balanced [ here

then you will get

something [foo balanced [ here

\s-1XML\s0 tags

The power expression \*(C`[[:XML:]]\*(C' match a \s-1XML\s0 tag (with or without children \s-1XML\s0 tags. Note that this expression matches only well formed \s-1XML\s0 tags.

As an example, the rule

[[:XML:]]=>tag

applied to the string

<a><b></a></b> and <more><img src="foo"/></more>

will result in

<a><b></a></b> and tag

TUTORIAL

At the moment, just a set of commented examples.

Example1 \*(-- from number to portuguese words (using traditional rewriting)

Example2 \*(-- Naif translator (using cursor-based rewriting)

Conversion between numbers and words

Yes, you can use Lingua::PT::Nums2Words and similar (for other languages). Meanwhile, before it existed we needed to write such a conversion tool.

Here I present a subset of the rules (for numbers bellow 1000). The generated text is Portuguese but I think you can get the idea. I'll try to create a version for English very soon.

You can check the full code on the samples directory (file \*(C`num2words\*(C').

use Text::RewriteRules;

RULES num2words 100==>cem 1(\d\d)==>cento e $1 0(\d\d)==>$1 200==>duzentos 300==>trezentos 400==>quatrocentos 500==>quinhentos 600==>seiscentos 700==>setecentos 800==>oitocentos 900==>novecentos (\d)(\d\d)==>${1}00 e $2

10==>dez 11==>onze 12==>doze 13==>treze 14==>catorze 15==>quinze 16==>dezasseis 17==>dezassete 18==>dezoito 19==>dezanove 20==>vinte 30==>trinta 40==>quarenta 50==>cinquenta 60==>sessenta 70==>setenta 80==>oitenta 90==>noventa 0(\d)==>$1 (\d)(\d)==>${1}0 e $2

1==>um 2==>dois 3==>tre\*^s 4==>quatro 5==>cinco 6==>seis 7==>sete 8==>oito 9==>nove 0$==>zero 0==> ==> ,==>, ENDRULES

num2words(123); # returns "cento e vinte e tre\*^s"

Naif translator (using cursor-based rewriting)

use Text::RewriteRules; %dict=(driver=>"motorista", the=>"o", of=>"de", car=>"carro");

$word='\b\w+\b';

if( b(a("I see the Driver of the car")) eq "(I) (see) o Motorista do carro" ) {print "ok\n"} else {print "ko\n"}

RULES/m a ($word)==>$dict{$1}!! defined($dict{$1}) ($word)=e=> ucfirst($dict{lc($1)}) !! defined($dict{lc($1)}) ($word)==>($1) ENDRULES

RULES/m b \bde o\b==>do ENDRULES

AUTHOR

Alberto Simo\*~es, \*(C`<[email protected]>\*(C'

Jose\*' Joa\*~o Almeida, \*(C`<[email protected]>\*(C'

BUGS

We know documentation is missing and you all want to use this module. In fact we are using it a lot, what explains why we don't have the time to write documentation.

Please report any bugs or feature requests to \*(C`[email protected]\*(C', or through the web interface at <http://rt.cpan.org>. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.

ACKNOWLEDGEMENTS

Damian Conway for Filter::Simple

COPYRIGHT & LICENSE

Copyright 2004-2012 Alberto Simo\*~es and Jose\*' Joa\*~o Almeida, All Rights Reserved.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.