Greg London > Parse-Gnaw > Parse::Gnaw



Annotate this POD

View/Report Bugs
Module Version: 0.601   Source  


Parse::Gnaw - Write extensible, recursive, grammars in pure perl code (grammar rules are perl arrays) and apply them to whatever parsee you want.


Write extensible, recursive, grammars using pure perl code. Grammar rules are perl arrays. Apply them to whatever parsee you want. Normal parsees would be strings. Interesting parsees might be a three-dimensional array of characters.

        no strict 'vars';
        use Parse::Gnaw;
        use Parse::Gnaw::String;

        rule('SayHello', 'Hello', 'World');
        my $string=Parse::Gnaw::String->New('So Hello World of mine');

This is the second generation of Parse::Gnaw starting from revision 0.600. Gen1 stored rules as code references and that prevented recursive calls within a rule as calling the code ref for the rule would go into an infinite loop. Gen2 uses array references to store rule, with the name of the array reference variable matching the name of the rule.

        our $rulename = [ .... rule content .... ];

It should allow recursive rules, although it will probably get hung in an infinite loop trying to match a left recursive rule.

Define a Grammar

Before you can parse anything, you have to create a grammar. Grammars are created with the "rule" subroutine, which is imported when you use Parse::Gnaw.

        # see t/doc_ex_rule_hi.t
        use Parse::Gnaw;
        rule('SayHello', 'H', 'I');

This will create a package scalar in your current package. The name of the scalar will be the name of the rule. The scalar will be a reference to an array that contains the rule. You can treat it like any other perl variable.

        print Dumper $SayHello;

This will print out something like:

$VAR1 = [ [ 'rule', 'rule1', { 'methodname' => 'rule', 'filename' => 't/doc_ex_rule_hi.t', 'linenum' => 18, 'payload' => 'rule1', 'quantifier' => '', 'package' => 'main' } ], [ 'lit', 'H', { 'methodname' => 'lit', 'filename' => 't/doc_ex_rule_hi.t', 'linenum' => 18, 'payload' => 'H', 'package' => 'main' } ], [ 'lit', 'I', { 'methodname' => 'lit', 'filename' => 't/doc_ex_rule_hi.t', 'linenum' => 18, 'payload' => 'I', 'package' => 'main' } ] ];

The array shows three elements. The first is a "rule" which defines the name of the rule and also holds extra information about the rule. The next two elements are literals looking for 'H' and then 'I'.

Create Something to be Parsed

A grammar is half of the puzzle. You also need to create the thing you want to parse. A simple example is a string:

        # see t/doc_ex_string_dog.t
        use Parse::Gnaw::LinkedListDimensions1;
        my $ab_string=Parse::Gnaw::LinkedListDimensions1->new("dog");

What this does is take the string 'dog' and turn it into a linked list that can be parsed.

Because Data::Dumper() does not handle linked lists well (they do not display in an easy-to-read format), the display() method was created. It will output a Parse::Gnaw string-ish object of some kind in a more readable format

        Dumping LinkedList object
        LETPKG => Parse::Gnaw::Blocks::Letter # package name of letter objects
        CONNMIN1 => 0 # max number of connections, minus 1
        HEADING_PREVNEXT_INDEX  => 0    
        FIRSTSTART => 

                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa08c820)
                payload: 'FIRSTSTART'
                from: unknown
                         [ ........... , ........... ]
        LASTSTART => 
                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa18d70c)
                payload: 'LASTSTART'
                from: unknown
                         [ ........... , ........... ]
        CURRPTR => 
                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa08c820)
                payload: 'FIRSTSTART'
                from: unknown
                         [ ........... , ........... ]
        letters, by order of next_start_position()
        letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa252d2c)
                payload: 'd'
                from: file t/doc_ex_string_dog.t, line 22, column 0     
                         [ ........... , (0xa252de0) ]

                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa252de0)
                payload: 'o'
                from: file t/doc_ex_string_dog.t, line 22, column 1
                         [ (0xa252d2c) , (0xa252ef8) ]

                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa252ef8)
                payload: 'g'
                from: file t/doc_ex_string_dog.t, line 22, column 2
                         [ (0xa252de0) , ........... ]

                letterobject: Parse::Gnaw::Blocks::Letter=ARRAY(0xa18d70c)
                payload: 'LASTSTART'
                from: unknown
                         [ ........... , ........... ]

Apply the Grammar to the Grammee.

Now that you have a Grammar and a Grammee, you can parse.

The parse() method is something that Parse::Gnaw::LinkedList type objects have available. It takes in one argument, a string containing the name of the top level rule or grammar that you want to apply to the string. If the rule matches the string, parse() will return true 1. If the rule does NOT match the string, parse() will return false ''.


The parse() method is used for parsing an an entire string from the beginning. It is similar to putting ^ or \A at the front of a regular expression:

        m/^(rule)/ or m/\A(rule)/ 

Here's a full example of parsing a string:

        # see t/doc_ex_rule_and_string.t
        use Parse::Gnaw;
        use Parse::Gnaw::LinkedListDimensions1;

        # A Simple Rule Example
        rule( 'rule1', 'H', 'I' );

        # A simple string example
        my $histring=Parse::Gnaw::LinkedListDimensions1->new("HI THERE");

        ok($histring->parse('rule1'), "This is like regex   'HI THERE' =~ m/HI/ ");


The rule function is used to create rules. Rules are created as package scalar in caller's namespace. The name of the scalar is the name of the rule.

        package main;
        rule( 'rule1', 'H', 'I' );

The above example will create a rule called "main::rule1". You can call Data::Dumper on $rule1 and see that it is an array reference.

Rules by themselves don't match anything in a string or block of text. Rules are just a way to handle a grammar in managable chunks. They could be thought of as similar to a perl subroutine, a container for the code that does something.

The first parameter is a string with the name of the rule.

Everything after that defines what the rule does. These can be string literals or character classes or alternations or quantifiers, and so on. Another thing you can do inside a rule is call another rule.

Rules and Subrules

This rule


turns into a $rule1 scalar holding a reference to an array.

        $VAR1 = [
              'methodname' => 'rule',
              'filename' => 't/doc_ex_rule_hi.t',
              'linenum' => 18,
              'payload' => 'rule1',
              'quantifier' => '',
              'package' => 'main'
              'methodname' => 'lit',
              'filename' => 't/doc_ex_rule_hi.t',
              'linenum' => 18,
              'payload' => 'H',
              'package' => 'main'
              'methodname' => 'lit',
              'filename' => 't/doc_ex_rule_hi.t',
              'linenum' => 18,
              'payload' => 'I',
              'package' => 'main'

You may have noticed that the rule array is just an array of smaller arrays. These smaller arrays are created by the subrules passed into rule(). For example, a lit() subrule:


might create a subarray that looks like this:

              'methodname' => 'lit',
              'filename' => 't/doc_ex_rule_hi.t',
              'linenum' => 18,
              'payload' => 'H',
              'package' => 'main'

All subrule functions return a subarray of this format so that the rule() function can easily handle them.

The first element in the subarray is the method name associated with the subrule. In the above example, we used the lit() function to create the literal subarray. When the rule1 gets parsed, it will see this ['lit', 'H', {}] array and call the 'lit' *method*. Because the first element will be used as a methodname, the first element is always a string.

The second element in the subarray is the payload. The payload is whatever is the important bit of information for the method. For a lit(), the important information is the actual literal you're looking for, such as 'H' above. i.e. a capital letter H. Some payloads are more complicated. For a character class, the payload is a hash reference, and the keys are the different letters in the character class.

The third element is a hash reference that contains all the information for the subrule, including stuff that is only used for error reporting. For example, if a subrule throws a die"" while parsing a string, it would be helpful to know where the original subrule was defined and put that in the error message. Therefore, a number of entries in the hashref is location information as to where the subrule was originally defined in the code. In the above example, if we went to file t/doc_ex_rule_hi.t line number 18, we should see something that looks like:


or possibly just


If while parsing a string, an error occurs while looking for this lit('H'), the hashref contains the information needed to point back to where the original subrule was declared.

Subrule Function versus Subrule Method

The lit() function in Parse::Gnaw returns an array where the first element is 'lit'. When parsing a rule, the parser will take 'lit' and call that method on the text object being parsed.

This does get a little confusing from time to time.

        The lit() function is contained in the Parse::Gnaw package.

        The 'lit' method is contained in Parse::Gnaw::Blocks::ParsingMethods package.

And the Parse::Gnaw::Blocks::ParsingMethods package is a child package of Parse::Gnaw::LinkedList. The Parse::Gnaw::LinkedList package is the base package for defining any string/text object that you want to parse. As a string is parsed, the subarrays in the rule array is iterated through and whatever method is contained in the subrule is called.

There is a subrule function defined in Parse::Gnaw for defining rules.

There is a subrule method defined in Parse::Gnaw::Blocks::ParsingMethods for parsing the string.

One reason for this split is because the way the subrule is defined in the rule is usually the same regardless of what kind of string we're parsing. But depending on what kind of string we're parsing, we might have to handle the subrule method differently.

If you're just starting to use Parse::Gnaw, this split between rule/function and string/method won't stop you from using the module. But if you want to do more advanced things with the package, like create your own subrule, then you'll need to know that you need to create a subrule function and a string method.


The next set of documentation covers the subrule functions that can be called to define a rule. The subrule returns an array reference which is then passed into the rule function.

        rule('rulename', subrule(..blah..)  );

The subrule functions are defined in Parse::Gnaw. For every subrule, there is some corresponding method defined in Parse::Gnaw::Blocks::ParsingMethods.


Use the 'call;' subroutine to have one rule call another rule.

        rule( 'rule1', 'a', 'b');
        rule( 'rule2', 'c', call('rule1') );

Note: if you call a rule that doesn't exist, script will throw a warning. You can pre-declare a rule with the predeclare() function:

        rule( 'rule2', 'c', call('rule1') );
        rule( 'rule1', 'a', 'b');

Recursive calls currently work as long as some text in the string is consumed before making the recursive call. This will work fine:

        rule( 'myrule', 'a', call('myrule') );

The above example will work fine because it has to match something (the literal 'a') before it recursively calls itself ('myrule').

However, this example below will compile, but will get stuck in an infinite loop if you try to parse with the rule:

        rule( 'myrule', call('myrule'), 'a');

The last example above won't work because the first thing 'myrule' does is call 'myrule' again, which then wants to call 'myrule', which then wants to call 'myrule'. The parser currently doesn't detect this is happening, and so your code will get infinite recursion until the stack crashes.


When declaring rule1 that calls rule2, and you haven't yet declared rule2, you will get a warning message about the rule not existing. You can ignore this warning as long as you declare rule2 before you start parsing.

But if you want to supress the warning, use predeclare() and pass in the name of the rule you want to predeclare.

        rule( 'rule2', 'c', call('rule1') );

        ... later ...

        rule( 'rule1', 'a', 'b');


Pass the lit() function a string containing the literal value you want to match.

        rule( 'greeting', lit('hello') );

As a shorthand, any string passed into rule() will be assumed to be a lit().

        rule( 'greeting', 'hello' );

Note that 'greeting' is the name of the rule looking for a literal 'hello'.


Call this and pass in a string defining a character class.


This is like [aeiou] in perl regular expressions.


Call this and pass in a string defining an inverted character class.


This is like [^aeiou] in perl regular expressions.


The alt() function is for defining grammars that contain alternations or alternatives.

The rule 'fruit' might be a choice between 'banana', 'apple', and 'orange'. The three possible choices are an alternation.

        rule('fruit', alt([apple'], ['banana'], ['orange']));

In a perl regular expression, this might look like:


The problem is that we can't use pipe '|' as a separator. So, instead, we have to use array references to bundle the different alternatives. It's a bit more typing, but we need some way to associate different pieces of alternatives, because most alternatives won't be just alternatives of just one word.

        rule('greetings', alt(['howdy','partner'], ['hello', 'friend'], ['hey', 'sport']));

In the 'greetings' example, the only way to know which literals are bundled together is to put them in array references. With the array references acting to bundle the alternatives, the rule is functionally equivalent to the following regexp:

        m/(howdy partner)|(hello friend)|(hey sport)/

without the array references, we might assume each individual word is an alternative, leading to a regexp that might look like this:


The alt() function will create rules for each alternative which will follow the pattern "alternate_" followed by an integer.


Quantifier. Pass in a series of subrules to thrifty and it will attempt to match that series as defined by the last entry in the elements passed into the function call.

A perl regular expression /(abc)+/ becomes thrifty('a', 'b', 'c', '+');

All arguments but the last one are essentially put in parenthesis and associated with the quantity specifier i.e. /(abc)+/ becomes thrifty('a','b','c','+')

Note the only quantifier mode supported is thrifty. Parse::Gnaw does not support greedy quantifiers.

Here is a list of ways you can define the last element passed into thrifty:

        thrifty( ... , [3,9]    );      3 to 9
        thrifty( ... , [3,]     );      3 or more
        thrifty( ... , [,9]     );      0 to 9
        thrifty( ... , '3,9'    );      3 to 9
        thrifty( ... , '3,'     );      3 or more
        thrifty( ... , ',9'     );      0 to 9
        thrifty( ... , '3'      );      3, no more, no less
        thrifty( ... , '+'      );      1 or more
        thrifty( ... , '*'      );      0 or more
        thrifty( ... , '?'      );      0 or 1

Note that there is more than one way to express the min/max pair.

'3', could also be specified as '3,3' as well as [3,3].

The thrifty function depends greatly on the internal 'fragment_a_rule' function.


Internal subroutine.

This processes the various ways to call the various Parse::Gnaw functions and fills in the pieces the caller doesn't pass in. Should always return a hash reference will all info filled in.

        rule('rulename', ... );
        rule('rulename', {ruleinfo}, ... );

        thrifty({quantifierinfo}, ...);

The purpose of the function is to support all the above forms of calling Parse::Gnaw functions, extract the information regardless of the format, and return a generic hashref of information that can be used by any function.


Internal subroutine. Used to break up a rule into pieces so that a quantifier can operate correctly.

the rest of the code in this subroutine is to "reorder" the grammar. for example, this grammar:

        rule1 : 'a' rule2 'b'
        rule2 : 'c' thrifty('d') 'e'

needs to rearrange the thrifty so that it can try to match a number of 'd' then it has to match 'e', then it has to match 'b' from the previous rule. if the thrifty quantifier fails, it has to try to match another 'd', then match 'e', then match 'b' from the previous rule.

This can't be done treating each rule as a subroutine/function as they appear because a quantifier can't return after it's matched 'd'. it has to match 'd', then match anything in the grammar anywhere in the grammar that occurs after it, and THEN it can return.

The way we're going to do this is by fragmenting/chopping up the rules any time we have a CALL or QUANTIFIER (quantifiers are actually calls) we are going to take everything AFTER THE CALL, and put it in its own rule fragment.

the original call gets modified with a thencall=>rulefragment added to it.

for example:

        rule1 : 'a' call('rule2') 'c' qty(thrifty1) 'e'
        rule2 : 'b'
        thrifty1 : 'd'

we need to fragment rule1

        rule1 : 'a' call('rule2') 'c' qty(thrifty1) 'e'

It can be viewed as getting fragmented as follows: rule1 : 'a' call('rule2') [ 'c' qty(thrifty1) ['e']] ^frag1 ^frag2

therfore it becomes rule1 : 'a' call('rule2',thencall=>rule1frag1) rule1frag1 : 'c' call('thrifty1', thencall=>rule1frag2) rule1frag2 : 'e'

this will allow all calls and quantifiers to treat the rest of the grammar after the call/quantifier as if it were part of a nested function call. The "thrifty" call doesn't return until it matches all the way to the end of the grammar, therefore, everything after the thrifty call needs to be treated as part of the thrifty function call.

in the above example, when we call rule 'thrifty1', we also pass in the fact that the rule after that is 'rule1frag2' this means 'thrifty1' can match 1 'd', and then call rule1frag2 to see if the rest of the grammar matches. if it fails, we can trap the failure in the 'thrifty1' call, and then we can try to match another 'd', and then try calling rule1frag2 again to see if the rest of the rule matches THAT.

Explaining it from another angle, fragment_a_rule() breaks up the rules to rearrange the *associativity* while maintaining the same functionality.

The original rule:

        rule1 : 'a' rule2 'b'
        rule2 : 'c' thrifty('d') 'e'

Could be flattened to look like this:

        rule1: 'a' ('c' thrifty('d') 'e') 'b'

However the quantifier will not operate with the above associativity. What fragment_a_rule() does is break up the rule into fragments and rearrange the associativity so that the rule can be reformed as function calls.

The original rules look like this:

        rule1 : 'a' rule2 'b'
        rule2 : 'c' thrifty('d') 'e'

And the fragmentation turns it into this:

        rule1 : 'a' call('rule2',thencall=>'b')
        rule2 : 'c' call(('d', {thrifty}), thencall=> 'e')

The 'b' and 'e' fragments in the "thencall" sections get turned into their own rule fragments, with their own rulenames.

And the "thrifty" because a call() that passes in {thrifty} specific information in a hashref.

The calls then can be handled like subroutines, with extra information being passed in, such as the {thrifty} information and the "thencall" rule name.

For a third example, imagine this rule:

        rule( 'myrule', thrifty('a','+'), thrifty('b','+'), 'c');

And then imagine trying to apply the above rule to the following string:


The rule would be fragmented to look more like this:

        myrule : call('a',{thrifty,'+'}, thencall=>('b',{thrifty,'+'}, thencall=>'c'));

The first thrifty 'a' in the rule would match the first 'a' in teh string:


THe rule would thencall the second fragment, which would start with the thrifty 'b'. This would match the first 'b' in the string.


The rule would thencall the 'c' rule, which would fail. THe thrifty 'b' would then expand to include the second 'b' in the string.


The thrifty 'b' would thencall the 'c' fragment again, which would fail. The thrifty 'b' would then expand to include the third 'b' in the string.


The thrifty 'b' would thencall=> the 'c' fragment, which this time would match.


The 'c' thencall rule would return successfully


The thrifty 'b' would return successfully


And then the thrifty 'a' would return successfully


At which point, we've returned all the way to the very first rule, therefore the parse matched and succeeded.

So, while the original rule might look at the matches like this:


The above associativity doesn't allow the rules to be handled like subroutines. If we fragment the rules and change the associativity to this:


Then the rules match the flow of a subroutine, and we can parse a rule simply by treating each rule call() as a subroutine call.


Internal subroutine.

given a hash created by process_first_arguments_and_return_hash_ref(), extract all location information and copy it to a newly created hash.


Internal subroutine.

Pass in a string. Will call eval("") on it. If you want the eval to return a value, assign it to a special variable $eval_return. The value of $eval_return will be returned by eval_string() function.


Internal subroutine. returns a string that will be used to fragment any rules.

When a rule is fragmented, the fragments are named


Call this subroutine to return the string value for fragment_suffix().


Internal subroutine. Used to get a reference to the rulebook in the caller's package.

All rules for a package are placed into a package variable called (packagename)::rulebook.

This variable is a hash reference where the keys are the names of the rules and the data is an array reference for each rule.


Internal subroutine. Used to get a reference to a specific rulename in the caller's package.

Each rule generated for a package is placed into the package as a scalar containing an array reference.

The array reference contains the rule information needed to parse a string.


Internal subroutine. Formats the package name into consistent string.


Internal subroutine. Formats the filename into a consistent string.


Internal subroutine. Formats the line number into a consistent string.


Please report any bugs or feature requests to bug-parse-Gnaw at, or through the web interface at I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.


You can find documentation for this module with the perldoc command.

    perldoc Parse::Gnaw

You can also look for information at:



Copyright 2013 Greg London.

This program is free software; you can redistribute it and/or modify it under the terms of the the Artistic License (2.0). You may obtain a copy of the full license at:

Any use, modification, and distribution of the Standard or Modified Versions is governed by this Artistic License. By using, modifying or distributing the Package, you accept this license. Do not use, modify, or distribute the Package, if you do not accept this license.

If your Modified Version has been derived from a Modified Version made by someone other than you, you are nevertheless required to ensure that your Modified Version complies with the requirements of this license.

This license does not grant you the right to use any trademark, service mark, tradename, or logo of the Copyright Holder.

This license includes the non-exclusive, worldwide, free-of-charge patent license to make, have made, use, offer to sell, sell, import and otherwise transfer the Package with respect to any patent claims licensable by the Copyright Holder that are necessarily infringed by the Package. If you institute patent litigation (including a cross-claim or counterclaim) against any party alleging that the Package constitutes direct or contributory patent infringement, then this Artistic License to you shall terminate on the date that such litigation is filed.


syntax highlighting: