- From: Seaborne, Andy <andy.seaborne@hp.com>
- Date: Wed, 16 Mar 2005 16:04:35 +0000
- To: kendall@monkeyfist.com
- Cc: Eric Prud'hommeaux <eric@w3.org>, public-rdf-dawg@w3.org
Kendall Clark wrote: > On Wed, Mar 16, 2005 at 03:37:09AM -0500, Eric Prud'hommeaux wrote: > >>Bijan was saying that he wants an XML serialization for SPARQL. >> >>While on IRC with Hugo, I doodled: > > > I've rewritten the first 20 queries in rq23 as XML, and I've done > this now 2 times -- trying to get a feel for what this XML > serialization should look-n-feel like. The present state of it is > pretty much useful as an AST for a SPARQL parser, since Bryan Thompson > and I pretty much just synched up on that effort. Better would be to capture the abstraction of the query and not specific to a parser. For example, I do not retain any of the syntax sugaring, expanding it all at parse time once and for all. SPARQL/QL is defined abstractly (basic pattern, group, union, optional, graph, filter) and if the structure of the XML reflected that then tools wishing to analysis/manipulate queries would have an eaiser to use form. If queries are constructed in an XML-based tool, then the abstract form woudl be the output, not the strangeness of the text syntax. Pretty printers are fun to write - it is quite easy to write one if it assumes the triple layout of the parser (i.e. parsrs and pretty printers come in pairs - rather likely in teh same toolkit) yet does not require recording syntactic forms in the parse tree. > > Once I've finished rewriting the remaining dozen or so queries from > rq23, and worked on a schema for all of this, I'll share it with the > group. I've started work on a specification for the serialization, but > it'll be a little while longer yet. > > (One thing that would have made this easier is if rq23 were > well-formed, so I could have extracted all the <pre class="query"> > bits by machine. I used a "tag soup" parser and got most of them, > though.) I'd also like that - to extract all relevant examples as test cases. Fixing and sending the diffs would be great - at least a version ref would help. It passes through "tidy -xml" but that does get run so things can creep in. > > Kendall > > PS--Yeah, I know, this isn't protocol work per se, but there are > connections and dependencies, I think, between our > on-the-wire/exchange serialization and the protocol, so I don't feel > like it's wasted or irrelevant effort. > Could you say what connections and dependencies you are discovering? Just asking but is it proposed that a SPARQL/XML syntax be part of our Last Call bundle of docs? Or is it a WG note? Something else? Should there be an RDF encoding of the query structure? If we want multiple syntaxes, a shared abstract model of a query encoded into various forms would be cool. Andy
Received on Wednesday, 16 March 2005 16:04:50 UTC