Support IRIs but what about double quotes?

[idimopoulos]

I am opening an issue because because I would like to first have a small feedback if I am thinking this correctly to not lose too much time here. This is kind of a follow up from #29 where we improved the support for literals and we cleared out a lot about what characters are supported and what are not.

Recently, I had the same issue but for the subject. From what I can read online, the RDF data, support IRIs (Internaltional RIs). That means that IRIs like http://example.com/person#name="Ηλίας" is perfectly valid as a resource identifier.

2 issues here:

1. We did not update in Successor of #25: Boost Ntriples performance by @idimopoulos #29 the subjects, we only handled objects. That means that all characters are currently encoded, and the URL above is converted completely using the old system (characters are replaced by their \0x#### version or their \u#### version.
2. The second problem, is the double quotes. While it should be perfectly fine, there is a problem when running queries and how they are transported if there is a way. For example, in Virtuoso (community edition 7.2) the queries

WITH <http://my-example-graph>
INSERT { <http://example.com/person#"Ηλίας"> <http://example.com/predicates#hasName> "Ilias" } 

WITH <http://my-example-graph>
INSERT { <http://example.com/person#\"Ηλίας\"> <http://example.com/predicates#hasName> "Ilias" } 

WITH <http://my-example-graph>
INSERT { <http://example.com/person#\u0022Ηλίας\u0022> <http://example.com/predicates#hasName> "Ilias" } 

WITH <http://my-example-graph>
INSERT { <http://example.com/person#%22Ηλίας%22> <http://example.com/predicates#hasName> "Ilias" } 

either throw an error, or the last one works, but the %22 is not decoded, is passed as is. However, if I am to do this:

WITH <http://my-example-graph>
INSERT { ?identifier <http://example.com/predicates#hasName> "Ilias" }
WHERE { SELECT ?identifier WHERE {
  BIND ( IRI("http://example.com/person#\"Ηλίας\"") as ?identifier )
}}

I am able to insert the quotes in the database as a normal IRI. Do you have @k00ni any experience with this? Does @zozlak has any experience on how this should be handled?

[zozlak]

Just as a note - I'm a little busy this week and will read this issue carefully only on Friday.

[idimopoulos]

Sure, take your time, I will deal with the current status in our project first.
Mainly, I was questioning if we should proceed with the following:

diff --git a/lib/Serialiser/Ntriples.php b/lib/Serialiser/Ntriples.php
index 4e7a8ab..4db84d8 100644
--- a/lib/Serialiser/Ntriples.php
+++ b/lib/Serialiser/Ntriples.php
@@ -209,7 +209,7 @@ class Ntriples extends Serialiser
      */
     protected function serialiseResource($res)
     {
-        $escaped = $this->escapeString($res);
+        $escaped = NtriplesUtil::escapeIri($res);
         if ('_:' == substr($res, 0, 2)) {
             return $escaped;
         } else {

[zozlak]

I was questioning if we should proceed with the following:

Makes a lot of sense to me. The more I look at how it's done now in the easyrdf, the more I think, it's broken :-)

When it comes to your Virtuso issue, I'm afraid it's complex. On the normative side the SPARQL specification says:

RDF URI references containing "<", ">", '"' (double quote), space, "{", "}", "|", "", "^", and "`" are not IRIs. The behavior of a SPARQL query against RDF statements composed of such RDF URI references is not defined.

so basically it can be vendor-specific and you just need to do it in a way your vendor (here Virtuoso) implemented it.
Which brings us to an important remark - the serialization of IRIs and literals in SPARQL has its own rules which are not exactly the same as for n-triples nor turtle (hurray! we all love the RDF ecosystem, don't we? reading RDF specifications is admittedly depressing).

[k00ni]

Any news on your site?