Re: official orgmode parser

[Ken Mankoff <mankoff@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1889 bytes --]


On 2020-10-26 at 07:21 -07, Nicolas Goaziou <mail@nicolasgoaziou.fr> wrote...
> Ken Mankoff <mankoff@gmail.com> writes:
>
>> I question if this is possible because EBNF is for context-free
>> grammars, but I *think* Org syntax is context-free.
>
> It's not as explained in a footnote in the Org syntax document.

Yes, I meant to write that I think Org syntax is maybe *not* context-free, and therefore EBNF can't capture all of it. But it could still be very helpful and capture most of it.

But the more I think about it, the more I think Org may be context-free.

For the footnotes, I'm not sure that "(1) In particular, the parser requires stars at column 0 to be quoted by a comma when they do not define a headline" violates context. An "*" in the first column defines a header. It can be escaped by anything else too (" *" works too). If ",*" has a special meaning, that can be captured elsewhere in the syntax.

I'm also not sure (2) violates context-freeness, at least in the EBNF sense where a context can include a newline. See for example:

section ::= "*"+ string (tag+) newline (planning newline)? (property_drawer newline)?

planning ::= ("SCHEDULED:" "<" date_or_time ">")? ("DEADLINE:" "<" date_or_time ">")?

property_drawer ::= ":PROPERTIES:" newline drawer_contents newline ":END:"

drawer_contents ::= ":" property ":" whitespace string

Where the first line, "section" is represented graphically as the attached image.

I guess I'm not 100% clear what "context-free" means. EBNF can represent a language where a for loop has an opening and closing brace. The closing brace is context-dependent, just as the planning or property drawers are.

I recently used EBNF to represent a CSV file with header, and I was unable to capture the requirement that the header column must have the same number of fields or commas as the data section. I think that is context-free. 



[-- Attachment #2: tmp_20201026_090940.png --]
[-- Type: image/png, Size: 8197 bytes --]

[-- Attachment #3: Type: text/plain, Size: 7 bytes --]


  -k.