Re: [Gzz] ``canon3_file_format``: A canonical, N3-based file format

[Benja Fallenstein]

Tuomas Lukka wrote:
> On Wed, Apr 02, 2003 at 04:11:47PM +0200, Benja Fallenstein wrote:
> > If we want to use it with CVS, requiring LF would mean that the files 
> > would have to be added as binary; AFAIK CVS wouldn't do diffing then.
>
> Ahh, right.
>
> Hmm - maybe should have different subformats: strict and non-strict:
> strict which is absolutely specified, and non-strict which is used for CVS?

Ok.

> > > Capitalize "must" ;)
> >
> > Which definition do you want to use?
>
> RFC one?

Ok.

> Anyway, might make sense to unconvolve the sentence:
> First Literals, then URIrefs, and finally anonymous nodes.

Ok.

> > > > - URIrefs are compared character-by-character,
> > > > in the form as defined in [RFC 2396]
> > > > (i.e., *after* Unicode characters outside
> > > > the ASCII range have been escaped).
> > > > Characters are compared by Unicode code point
> > > > value.
> > >
> > > Is this the same as a lexicographic string comparison
> > > of the UTF-8 encoded one?
> >
> > I don't know.
>
> Need to explain how to compare. I couldn't write a program yet.

In Java: string1.compareTo(string2), on the in-memory representation as 
used by Jena.

The full writer algorithm looks something like this:

- Get all Statements from the Model.
- Put them into a SortedSet. Normalization into NFC is done at this 
stage (note to self: find out how this works in Java.) The comparator 
uses the algorithm specified in the PEG, using Java compareTo to compare 
any strings.
- Create an UTF-8 writer and write the header to it.
- Write each statement in order. Escaping of literals is done at this stage.


- Benja