Re: Formal description of Org files

[MidLifeXis at PerlMonks <midlifexis@wightmanfam.org>]

[-- Attachment #1: Type: text/plain, Size: 3157 bytes --]

This is something I would like to see adopted by the org community as well.  One 
of my main reasons is to allow other languages to be able to at least, given the 
proper configuration information (for TODO keywords, etc), to be able to deal 
with org files.  I am not quite sure where the lines should be drawn, but I am 
envisioning something like:

* The ORG format definition and actions

The format of the .org file (ORG) itself and a set of actions that are well 
behaved (add node, move node up, node id, property value, etc).  A core set of 
validation tests could be created to exercise any compliant org library.

* The configuration for a set of ORG files

The configuration for a given org (note the small "org") application, defining 
particular tweaks that an application can use to make sense of a properly 
formatted ORG file, such as valid TODO keywords, etc.  This area is muddy in my 
mind.

* The org application itself - how the org files are used in a particular 
application, or the ideas and data that set of ORG files represents, or how an 
exporter handles the data thrown at it, or....

But that is just /my/ dream.






________________________________
From: Rustom Mody <rustompmody@gmail.com>
To: emacs-orgmode <emacs-orgmode@gnu.org>
Sent: Fri, April 15, 2011 12:13:45 PM
Subject: Re: [O] Formal description of Org files


Hi,
>
>I have a question.
>
>At FOSDEM, someone asked me if there was a formal description of the structure 
>of Org files, in some language
>that would be the input for a parser (or parser generator?) so that Org file 
>could be easily parsed.
>
>Unfortunately I did not catch the name of the format description language that 
>could be
>used for something, not did I catch the name of the person who talked to me.
>
>Can anyone help out here?  Let me know what language to use, and maybe help work 
>on such a formal description?  I
>think it would be useful to have....
>
Something like yacc (bison, antlr etc) are all 'executable BNF' languages.  When 
they work they can make the code an order of magnitude smaller and 
development/programming correspondingly easier.

That said I see a couple of hitches.

1. Grammatical handling of languages is based on the assumption of a clear and 
well defined set of tokens/lexemes.  I expect this would be harder in org than 
the typical programming languages for which yacc etc are used.  For example in 
most 'normal' languages there are comments and strings.  These involve some non 
trivial handling which is entirely hidden from the grammar by being pushed into 
the lexer.

2. Parsing a program is done for the full program as a rule (IDEs are the 
exception to the rule). Sensible parsing of program fragments, where the 
fragmenting could be quite arbitrary, is a bit of a research problem

3. As I see it, the main declarative tool (somewhat akin to grammars) that org 
uses is regular expressions.  IOW org is written with re-s strung together with 
programming logic ie vanilla elisp.  An alternative that stays within the 
regular framework (not using the heavy guns of context free parsing) may be 
ragel: http://www.complang.org/ragel/

Rusi

[-- Attachment #2: Type: text/html, Size: 4206 bytes --]