Hi Ron,
This all sounds great. I have a question on your choice of using a tree
structure, can you explain that more? Are you thinking of the file being
the root, then having leaves like: indi, fams, famc, etc and then each
of these have their respective data hanging off those objects? Or are
you thinking the tree would represent the family relationships? I don't
see how the later will work.
-Steve
On 11/5/2012 2:04 AM, Ron Savage wrote:
Hi
The new GEDCOM parser
This document is a collection of ideas which have been percolating
in my mind for a long time.
Comments welcome.
Ideas
Module name
Genealogy::Gedcom::Parser.
A place-holder, Genealogy::Gedcom
<http://metacpan.org/release/Genealogy-Gedcom>, is already on CPAN.
Note: This module was written before the new, major tools now
available were released. See Tools below.
ETA
There is no ETA for the parser.
However, certain Perl-based tools are now available which will make
coding a simple task. See Tools below.
See also 'Famous Last Words' :-).
UTF-8
The code will accept input files in utf-8, and generate files
containing utf-8 characters.
Apache and mod_perl
These will not be required. I only mention these because references
to them appear in the Gedcom.pm distro.
Logging
The code will have a built-in logger, so debugging, e.g., can be
turned on with a parameter to new().
This logger will use Log::Handler. See Tools below.
Sub-classing
Sub-classing the main module will be trivial, and samples will be
provided.
Sub-classing will be done with Hash::FieldHash. See Tools and the
FAQ below.
Grammars and grammar generators
Like Gedcom.pm, the code will read a GEDCOM grammar in BNF from a
file.
I'll run this phase before shipping the module, so you don't have to.
See Tools below, specifically Marpa::Rules::Simple.
Bascially, this means the startling complexity of the code in
Gedcom.pm is a thing of the past.
Operating the parser
Using Marpa, callbacks are triggered when input is recognized.
So, when lines like these are encountered:
1 @<XREF:FAM>@ FAM
2 RIN <AUTOMATED_RECORD_ID>
Marpa will automatically call the callback attached to each tag.
Callbacks will probably have names like 'do_fam' and 'do_rin', i.e.
of the format 'do_$tag'.
The parameters passed to the callback include the non-tag text on
the line.
Default callbacks for all tags will be provided, each one doing its
part in parsing the parameters to the tag, and storing the result.
The result will probably be stored in a tree. See Tools below,
specifically Tree::DAG_Node.
Database support
A DBD::SQLite database is possible.
Tools
o Hash::FieldHash
Simplifies class-building.
As for the obvious question, why not use Moose, see the FAQ below.
o Log::Handling
Simplifies logging.
o Marpa::R2
This is the modern way to do parsing.
Home page <http://jeffreykegler.github.com/Marpa-web-site/>.
Jeffrey's blog about Marpa
<http://jeffreykegler.github.com/Ocean-of-Awareness-blog/>.
My recent article about lexing and parsing with Marpa
<http://www.perl.com/pub/2012/10/an-overview-of-lexing-and-parsing.html>.
o MarpaX::Simple::Rules
This module reads a grammar in BNF and generates a Marpa grammar.
Hence it will read a BNF version of the GEDCOM spec and output
the matching Marpa grammar.
o Tree::DAG_Node
The most sophisticated tree-handling code on CPAN. I've recently
become co-maintainer of this module.
FAQ
Why did you choose Hash::FieldHash over Moose?
My policy is to use the light-weight Hash::FieldHash for stand-alone
modules and Moose for applications.
Why did you choose to store the data in a tree?
A GEDCOM file's structure can be viewed as a tree, so my initial
plan is to store the data likewise.