Help-Site Computer Manuals
  Algorithms & Data Structures   Programming Languages   Revision Control
  Cameras   Computers   Displays   Keyboards & Mice   Motherboards   Networking   Printers & Scanners   Storage
  Windows   Linux & Unix   Mac

Framework for customising JE's parser

JE::Parser - Framework for customising JE's parser


JE::Parser - Framework for customising JE's parser


  use JE;

  use JE::Parser;

  $je = new JE;

  $p = new JE::Parser $je; # or: $p = $je->new_parser

  $p->delete_statement('for', 'while', 'do'); # disable loops

  $p->add_statement(try => \&parser); # replace existing 'try' statement


This allows one to change the list of statement types that the parser looks for. For instance, one could disable loops for a mini-JavaScript, or add extensions to the language, such as the 'catch-if' clause of a try statement.

As yet, delete_statement works, but I've not finished designing the API for add_statement. Currently, JavaScript's eval function always uses the default parser, which will be fixed.

I might provide an API for extending expressions, if I can resolve the complications caused by the 'new' operator. If anyone else wants to have a go at it, be my guest. :-)


$p = new JE::Parser
Creates a new parser object.

$p->add_statement($name, \&parser);
This adds a new statement (source element, to be precise) type to the list of statements types the parser supports. If a statement type called $name already exists, it will be replaced. Otherwise, the new statement type will be added to the top of the list.

($name ought to be optional; it should only be necessary if one wants to delete it afterwards or rearrange the list.)

If the name of a statement type begins with a hyphen, it is only allowed at the 'program' level, not within compound statements. Function declarations use this. Maybe this convention is too unintuitive.... (Does anyone think I should change it? What should I change it too?)

&parser will need to parse code contained in $_ starting at pos(), then either return an object, list or coderef (see below) and set pos() to the position of the next token[1], or, if it could not parse anything, return undef and reset pos() to its initial value if it changed.

[1] I.e., it is expected to move pos past any trailing whitespace.

The return value of &parser can be one of the following:

  1. )
  2. An object with an eval method, that will execute the statement, and/or an init method, which will be called before the code runs.

  3. )
  4. (Not yet supported!) A coderef, which will be called when the code is executed.

  5. )
  6. (Not yet supported.) A hash-style list, the two keys being eval and init (corresponding to the methods under item 1) and the values being coderefs; i.e.:
      ( init => \&init_sub, eval => \&eval_sub )

Maybe we need support for a JavaScript function to be called to handnle the statement.

Deletes the given statement types and returns $p.

(Not yet implemented.)

Returns an array ref of the names of the various statement types. You can rearrange this list, but it is up to you to make sure you do not add to it any statement types that have not been added via add_statement (or were not there by default). The statement types in the list will be tried in order, except that items beginning with a hyphen always come before other items.

The default list is qw/-function block empty if while with for switch try labelled var do continue break return throw expr/

Parses the $code and returns a parse tree (JE::Code object).

Shorthand for $p->parse($code)->execute;


None by default. You may choose to export the following:

Exported Variables

... blah blah blah ...

Exported Functions

These all have () for their prototype, except for expected which has ($).

... blah blah blah ...


(To be written)

  expected 'aaaa'; # will be changed to 'Expected aaaa but found....'

  die \"You can't put a doodad after a frombiggle!"; # complete message

  die 'aoenstuhoeanthu'; # big no-no (the error is propagated)


Mini JavaScript

This is an example of a mini JavaScript that does not allow loops or the creation of functions.

  use JE;

  $j = new JE;

  $p = $j->new_parser;


Since function expressions could still create functions, we need to remove the Function prototype object. Someone might then try to put it back with Function = parseInt.constructor, so we'll overwrite Function with an undeletable read-only undefined property.

  $j->prop({ name     => 'Function',

             value    => undef,

             readonly => 1,

             dontdel  => 1 });

Then, after this, we call $p->eval('...') to run JS code.

Perl-style for(LIST) loop

Well, after writing this example, it seems to me this API is not sufficient....

This example doesn't actually work yet.

  use JE;

  use JE::Parser qw'$s ident expr statement expected';


  $j = new JE;

  $p = $j->new_parser;


      sub {

          /\Gfor$s/cog or return;

          my $loopvar = ident or return;

          /\G$s\($s/cog or return;

          my @expressions;

          do {

              # This line doesn't actually work properly because

              # 'expr' will gobble up all the commas

              @expressions == push @expressions, expr

                  and return; # If nothing gets pushed on  to  the

                              # list,  we need to give the default

                              # 'for' handler a chance, instead of

                              # throwing an error.

          } while /\G$s,$s/cog;

          my $statement = statement or expected 'statement';

          return bless {

              var => $loopvar,

              expressions => \@expressions,

              statement => $statement

          }, 'Local::JEx::ForList';




  package Local::JEx::ForList;


  sub eval {

      my $self = shift;

      local $JE::Code::scope =

          bless [@$JE::Code::scope], 'JE::Scope';

          # I've got to come up with a better interface than this.

      my $obj = $JE::Code::global->eval('new Object');

      push @$JE::Code::scope, $obj;

      for (@{$self->{expressions}}) {

          $obj->{ $self->{loopvar} } = $_->eval;





JE and the JE::Code manpage.