mKR serializer for Raptor

[rhmccullough]

New files for mKR writer, serializer, tests. Minimal changes to other programs to add a new language. I have some questions re: Turtle serializer handling of @base and rdf:nil.

[dajobe]

the README.md should not be at the top level. Maybe in the tests area. If it's just about this pull request, delete the file and put in the PR description

[dajobe]

How different is the mkr work compared to turtle. Is a whole new serializer / writer needed or could the turtle one do it with some flags?

[dajobe]

Is mkr really generating 30,000 line results? e.g. test-14.mkr that seems huge

[rhmccullough]

Dave,

I'll try to cover all 4 emails in this reply.

1) It is possible the turtle serializer could do mKR's work with flags. The basic rules are: "." => ";" "( x y z )" => "[ x, y, z ]" and "( )" => "[ ]" "[ ... ]" => "{ ... }" subject => . is subject ; predicate object1, object2, ... ; => . has predicate = [ object1, object2, ... ] ; except rdf:type object1, object2, ... ; => . isu object1, object2, ... ; rdfs:subClassOf object1, object2, ... ; => . iss object1, object2, ... ; rdfs:subPropertyOf object1, object2, ... ; => . iss object1, object2, ... ;

The basic philosophy is: "." stands for current concept, and is short enough to make mKR as readable as Turtle. Everything is a complete sentence, and all the more advanced features can be expressed in more complicated sentences. FYI: mKR uses "isa" to mean either "isu" or "iss".

2) re Makefile.am etc. I'm a real novice with the automake tools. I'll do it anyway you want.

3) re 30,000 line mKR file. That's for real. It's a translation of your 10,000 line Turtle file. I suggest deleting test-16.ttl and test-16.mkr

4) re test-08.ttl Turtle serializer changes "( )" to "rdf:nil". That seems contrary to the spirit of Turtle. I put a 3 line patch in raptor_mkr_emit_resource() that could put out "( )" instead.

5) listing4-3.ttl from "Linked Data" book has @base declaration. I do not know what is supposed to be done with it. Turtle serializer outputs no @base declaration. My hack of Turtle serializer outputs @base with local file name replacing original URI.

6) re building with github files I have not been able to do a successful build -- missing dependency & .am files. For now, I can only build by copying github files into the files I downloaded from librdf.org.

I'll start work on the changes you suggested. How do I keep my changes from affecting you?

[dajobe]

1. If the delta is small then it seems a good idea to me to reuse the turtle writer/serializer and add flags for the if(mkr) parts
2. & 6. The INSTALL.html file in the distribution describes how to build from GIT, what automake tools you need etc. If you look at the Raptor CI (continuous integration) runs at http://ci.aelius.com/job/raptor/641/console for example, you can see how a build from GIT is automated.
3. The test-08.ttl serializer thing sounds like a separate issue, you could make a small PR (pull request) for that alone.
4. The turtle serializer writing @base is done in raptor_turtle_writer_base() and only if the write_base_uri argument to raptor_new_turtle_writer() is true.

Your changes should be applied and the existing tests must all pass. Also I'd like a configure option to enable/disable mKR; defaulting to disable, and the new code made conditional with something like #ifdef RAPTOR_MKR. You can use an automake conditional for that; see configure.ac

[rhmccullough]

Re-implemented mKR serializer using raptor_serialize_turtle.c & raptor_turtle_writer.c. I reverted all the changes in rhmccullough/raptor master. All the latest changes are in the mkr-serializer branch of rhmccullough/raptor. I use context->emit_mkr = 1 for mKR and = 0 for Turtle. Since writer does not use context, I added emit_mkr argument to _prefix and _base procedures. I have been using tests/mkr/Makefile.mkr for testing, since your latest build aborts in scripts (can't find raptor2.h).

[dajobe]

Since you've reverted everything, I'll close this PR and you can make a new one from the mkr-serializer branch.