Update: you want 0.9.19 not 0.9.18 after package configuration issue found. Links fixed.
This release of Rasqal adds draft syntax
support for the SPARQL 1.1 Update language being developed by the W3C
SPARQL Working Group. The SPARQL 1.1 Update W3C Working
2010-01-26 introduces the first syntax design with some uncertainties
and gray areas still present (no grammar spec section yet). I added what
I thought would work, avoiding the ambiguous
WITH forms where
everything is optional. Since this is draft work, this extra parsing is
only done when the 'laqrs' query language syntax is chosen. LAQRS stands
for LAQRS adds to Querying RDF in SPARQL.
This is just syntax and API support in Rasqal, so it means you can
prepare the upload queries, but there is no code to execute it. The API
allows getting access to the decoded sparql update (
with or without
DATA) and graph operations (
There is still more to do, when the syntax gets changed in later drafts
and there is no API to stream triple insert/deletes during parsing, to
handle uploading and downloading large triple blocks. That would
required a rewrite of the SPARQL parser to use a different technology
than flex+bison (maybe lemon, maybe
Ragel) as well as new APIs.
Rasqal has several things to finish for SPARQL 1.0 support (UNION and nested OPTIONALs don't work) but the recent rewrite of the query engine internals should make other SPARQL 1.1 parts such as aggregate functions and nested queries, a lot easier to do than with the old query engine. I will probably remove the old query engine from the codebase soon.
The second substantial change is a set of APIs moved from private to
rasqal.h to enable the construction of query result sets and
query result set rows (
rasqal_row) via the public API. This allows
query results to be read from a syntax or constructed by API as well as
serialized to result formats, without any query being executed. Rasqal
can be used with this addition to provide the sparql results syntax
support for other applications that may have created query results via a
different method. It can read query results formats from the SPARQL XML
format (the standard format), and write or serialize them to SPARQL XML,
SPARQL JSON, CSV, TSV and an ASCII Table format. This functionality is
all available via Triplr where you can make HTTP
GET URLs for saved queries.
The final change is in the area of resilience. The functions in the
public API have been updated so that when invalid or
NULL pointers are
given, the functions return failure or
NULL / false rather than try to
use the pointer and probably crash. Hopefully I caught all of them. The
release testing (as usual) included valgrind memory leak checking of all
of the 100s of tests and there were no leaks or buffer overruns found.
This is also the first Rasqal release since switching to GIT as the source control for the Redland libraries so the source pointers have moved to git.librdf.org where details of how to check it out can be found.
So in summary, the main changes in this release are:
- 0.9.19: Fix rasqal.pc to Requires raptor again.
- Add initial draft parsing and API (NOT execution) support for SPARQL 1.1 Update W3C Working Draft of 2010-01-26.
- Add public APIs (row, results, result formatter, variables table) so that query results can be built, read and written without a query.
- Add API resilience checks for invalid NULL pointer arguments.
- Many other bug fixes and improvements were made.
- 0000320: Add a void* user_data field to rasqal_variable
- 0000323: Official MIME Type for JSON isn't text/json
- 0000343: Mime type for 'table' results format is text/plan
- 0000345: MIME Type and URI for TSV and CSV
- 0000347: rasqal linking fix
See the Rasqal 0.9.19 Release Notes for the full details of the changes.