Re: Some comments re: the current Turtle working draft

On Mon, Jul 11, 2011 at 7:39 AM, Alex Hall <alexhall@revelytix.com> wrote:
> On Sat, Jul 9, 2011 at 6:15 AM, Mischa Tuffield <mischa.tuffield@garlik.com>
> wrote:
>>
>> <snip/>
>> On 9 Jul 2011, at 01:02, Alex Hall <alexhall@revelytix.com> wrote:
>>
>> On Fri, Jul 8, 2011 at 12:29 PM, Mischa Tuffield
>> <mischa.tuffield@garlik.com> wrote:
>>>
>>> <snip/>
>>> 5. In Section 4.4 - Grammar: there is a distinct lack of whitespacing
>>> here, I am guessing this is based the current grammar is but a first pass.
>>> There is an email thread I started on this list which includes feedback from
>>> a Stefano D'Angelo (parser implementer), I think we should make sure we
>>> address the issues brought forward there [1].
>>
>> There is a related note from Andy at [1].  Basically, whitespace and
>> comments are included in the PASSED TOKENS rule, which indicates that
>> whitespace and comments are allowed as tokens (a.k.a. terminals) anywhere in
>> the grammar but ignored.  This reflects the fact that many tools (javacc,
>> Antlr, etc) can skip whitespace tokens or emit them on a special hidden
>> channel.
>>
>> Note that section 4.1 does talk some about whitespace.  Manually inserting
>> whitespace tokens everywhere they could possibly appear in the grammar would
>> be too difficult and would obscure the meaningful parts of the grammar.  So
>> we just say that it's allowed everywhere (outside of terminals) and only
>> required to disambiguate two terminals that would otherwise be interpreted
>> as one.
>>
>> Note also that the SPARQL grammar [2] handles whitespace in a similar
>> fashion.
>>
>> I have just gone though SPARQL1.1 grammar and agreed the handling of
>> whitespace is best left out to not obscure the meaningful parts. FWIW
>> Section 4.1 is slightly confusing from my point of view perhaps the
>> following statement in [a] should be expanded upon:
>> "White space is significant in tokens IRI_REF and string."
>
> I think all this is saying is that whitespace appearing within an IRI or
> string literal is not ignored as it is in other parts of the grammar, i.e.
> "foo bar" != "foobar".  This does bring up the question of whether the
> mention of IRI_REF should be dropped here, since whitespace is no longer
> allowed in IRI's.
> Also, the IRI_REF production is not displaying correctly in my browser.  I
> see:
> <IRI_REF>   ::=   "<" (( [^<>\"{}|^`\\] - [#0000- ] ) | UCHAR )* ">"
> Note that the class of excluded characters has a lower bound (#0000) but no
> upper bound.  Comparing to the SPARQL grammar, it looks like that part
> should read [#0000-#0020].

Heh, it's sort of "correct" as in #x0020 is ' '. There are still a
fair number of editorial issues with the table displayed grammar.
There may also be real issues with grammar. The ED draft now also has
a link to http://dvcs.w3.org/hg/rdf/raw-file/Turtle-FPWD/rdf-turtle/turtle.bnf
until the issues with the HTML table are resolved. It would help if
any bugs with the grammar were checked vs. the raw grammar and not
just the HTML display at this point.

Thanks,
Gavin

> -Alex
>
>>
>>
>> -Alex
>>
>> [1] http://lists.w3.org/Archives/Public/public-rdf-wg/2011Mar/0297.html
>> [2] http://www.w3.org/TR/sparql11-query/#whitespace
>>
>>
>> The link [2] above doesn't resolve in my browser, and I can find any
>> section entitled whitespace in the document, nevertheless I do prefer
>> SPARQL's style over the older "ws" heavy turtle submission.
>> Regards,
>> Mischa
>> [a]
>> http://dvcs.w3.org/hg/rdf/raw-file/Turtle-FPWD/rdf-turtle/index.html#sec-grammar-ws
>> - sent from a tablet thing
>

Received on Monday, 11 July 2011 15:38:40 UTC