[Raku PDF Project] / PDF::Grammar

PDF-Grammar

Although PDF documents do not lend themselves to an overall BNF style grammar description; there are areas where these can be put to use, including:

• PDF file header and trailer/xref parsing

• Parsing of objects fetched via the xref index. Top level objects commomly include: dictionarys , streams, arrays or numbers.

• The overall file structure for FDF files (which are not indexed), or for full-scan recovery of PDF files (headers, objects, cross-reference tables and footers).

• Parsing the operands that make up content streams. These are used to markup text, forms, images and graphical elements.

PDF::Grammar is a set of Raku grammars for parsing and validation of real-world PDF examples. There are four grammars:

PDF::Grammar::Content - describes the text and graphics operators that are used to produce page layout.

PDF::Grammar::Content::Fast - is an optimized version of PDF::Grammar::Content.

PDF::Grammar::FDF - this describes the file structure of FDF (Form Data) exchange files.

PDF::Grammar::PDF - this describes the file structure of PDF documents, including headers, trailers, top-level objects and the cross-reference table.

PDF::Grammar::Function - a tokeniser for Postscript Calculator (type 4) functions.

PDF-Grammar has so far been tested against a number of sample of PDF documents and may still be subject to change.

I have been working off the PDF 1.7 reference manual (http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/pdf_reference_1-7.pdf). I've relaxed rules, when needed, to handle real-world examples.

Usage Notes

• PDF input files typically contain a mixture of ASCII directives and binary data, plus byte-orientated addressing. For this reason:

  • files should be read as binary (avoid encoding layers)

  • strings should be decoded as latin1

  % rakudo -MPDF::Grammar::PDF -e"say PDF::Grammar::PDF.parse: slurp($f, :bin).decode('latin-1')"

• This module is put to work by the down-stream PDF module. E.g. to uncompress a PDF, using the installed pdf-rewriter script:

  % pdf-rewriter.raku --uncompress flyer.pdf

Examples

• parse some markup content:

  % raku -M PDF::Grammar::Content -e"say PDF::Grammar::Content.parse('(Hello, world\041) Tj')"

• parse a PDF file:

  % rakudo -MPDF::Grammar::PDF -e"say PDF::Grammar::PDF.parsefile( $f )"

• dump the contents of a PDF

  use v6;
  use PDF::Grammar::PDF;
  use PDF::Grammar::PDF::Actions;
  
  sub MAIN(Str $pdf-file) {
      my $actions = PDF::Grammar::PDF::Actions.new;
  
      if PDF::Grammar::PDF.parsefile( $pdf-file, :$actions ) {
          say $/.ast.raku;
      }
      else {
          say "failed to parse PDF: $pdf-file";
      }
  }

AST Reference

The action methods in this module return AST trees. Each node in the tree consists of a key, value pair, where the key is the AST Tag, indicating the type of the AST node.

For example, here's the AST tree for the following parse:

use PDF::Grammar::PDF;
use PDF::Grammar::PDF::Actions;
my $actions = PDF::Grammar::PDF::Actions.new;

PDF::Grammar::PDF.parse( q:to"--END-DOC--", :rule<ind-obj>, :$actions);
3 0 obj <<
   /Type /Pages
   /Count 1
   /Kids [4 0 R]
>>
endobj
--END-DOC--

say '# ' ~ $/.ast.raku;
# :ind-obj($[3, 0, :dict({:Count(:int(1)), :Kids(:array([:ind-ref($[4, 0])])), :Type(:name("Pages"))})])

This is an indirect object (ind-obj), it contains a dictionary object (dict). Entries in the dictionary are:

• Count with integer value (int) of 1.

• Kids, and array (array) containing one indirect reference (ind-ref).

• Type with name (name) 'Pages'.

In most cases, the node type corresponds to the name of the rule or token that was used to construct the node.

This AST representation is used extensively throughout the PDF tool-chain. For example, as an intermediate format by PDF::Writer for reserialization.

For reference, here is a list of all AST node types:

See also

• PDF - Raku module for PDF manipulation, including compression, encryption and reading and writing of PDF data.