Redland RDF API Library - Release Notes
The main changes in this release are:
Added a Virtuoso storage by Patrick van Kleef
Support Rasqal 0.9.17 API
Fix compiling against sqlite 3.6.17 or newer
Fix modular storage dynamic loading linking problems.
Building requires installed Raptor (1.4.17 or newer) and Rasqal
(0.9.16 or newer) libraries, preferably with pkg-config
to provide the compile/link details.
Issues Fixed:
- Issue #0000302: Modular storages fail to load from any other library or language binding (drobilla)
- Issue #0000304: librdf_storage_trees_remove_statement returns wrong value (I think)
- Issue #0000311: Incorrect literal comparison in tree structure
- Issue #0000315: cannot compile library on linux 2.6.31
- Issue #0000317: rdf_init_test fails if threads are enables
Configuration and build changes
Default to environment LIBS, CPPFLAGS, LDFLAGS in
configure to address Gentoo issue
267405
Make redland-src-config work for external raptor and rasqal again.
This is for the maintainers only.
Put modular build libltdl includes and libs into INTERNAL macros
Parser Class Changes
librdf_parser_check_name() added to check if a parser name is
valid:
int librdf_parser_check_name(librdf_world* world, const char *name);
Query Class Changes
Support Rasqal 0.9.17 API changes including versioned triple
factory and triple sources, RASQAL_LITERAL_UDT
user-defined datatype literals. Use
rasqal_literal_get_rdf_term_type when available or
fall back to internal version.
Serializer Class Changes
librdf_serializer_check_name() added to check if a
serializer name is valid:
int librdf_serializer_check_name(librdf_world* world, const char *name);
Storage Class Changes
Update storage module link dependencies so that they depend on
librdf library to prevent unresolved symbol errors.
More validation of storage module names.
Storages 'file' and 'uri'. These get new storage option
format with a name value to set the on disk / remote
URI) syntax format. For the both storages it has to be a legal
parser name and for the file storage it has to be a legal serializer
name too. This pretty much limits the values to ntriples, rdfxml and
turtle for the file storage (depending on how raptor is built).
Storage 'postgresql'. Replace deprecated PQescapeString with
PQescapeStringConn that takes a PGconn connection handle and when
encoding takes into consideration any connection properties such as
character encodings in force.
Storage 'sqlite'. Change define SQLITE_API to REDLAND_SQLITE_API
to allow redland to compile with sqlite 3.6.17 or newer where the
SQLITE_API macro is defined. Thanks to Samuli Suominen for the
report and fix.
Storage 'virtuoso'. Added a Virtuoso storage patch from
Issue#0000284
by Patrick van Kleef with some parts omitted for passing through
SPARQL queries that used rasqal internals.
Other changes
Export the list helper class to the public API, primarily for use
by external storage modules.
rdfproc utility: for the 'parse' command, get the
warning count from the warning_count_node, not the (just freed)
error_count_node.
Fix totally bogus memset in librdf_cache_set_common()
that zeroed junk on the stack rather than the memory expected.
The main changes in this release are:
Made storages modular by default - separate dynamically loaded
modules discovered from presence in the
PREFIX/lib/redland directory.
or from the directory defined in the
REDLAND_MODULE_PATH environment variable directory.
Building requires installed Raptor (1.4.18 or newer) and Rasqal libraries.
Issues Fixed:
General Changes
Developer builds from Subversion now require libtool 2.2.0 or
newer to libtoolize from configure.ac.
configure now discovers Berkeley DB 4.7
configure options --with-raptor=internal
and --with-rasqal=internal were removed.
Removed openssl digest support - not used much and an extra large
dependency. MD5 and SHA1 remain builtin.
Statement Class Changes
Fix to librdf_statement_encode_parts() for buffer
overrun error that could occur with certain size buffers. Patch from
Lou Sakey - thanks.
Storage Class Changes
Made the redland storage module API
librdf_storage_factory public. What was called the
storage context is now less confusingly called the storage
instance.
Moved iterator and stream get_method flags to public
API as new enums since they are used by the storage modules API.
Moved librdf_log() and librdf_log_simple()
into public API to allow external storage modules to log messages.
More checks for failure during storage instance allocation.
postgresql storage: Fixes for "Numerous leaks, potential NULL
pointer use, and dropped postgresql errors and buffer overruns in the
postgresql storage module" based on patches in Issue #0000286.
sqlite storage: Fixes to prevent deadlocks with BEGIN
IMMEDIATE instead of the default BEGIN DEFERRED.
No longer writes to the sqlite database when just querying. Several
other fixes.
trees storage: Several fixes.
Other Changes
Multiple Win32 build fixes based on patches from Lou Sakey - thanks.
Added librdf_parser_guess_name2() to SWIG API.
Internal Changes
Added LRU cache for internal use.
Several autogen.sh updates for libltdl.
Use libtool's libltdl to dynamially load storage
modules. Will link in a copy from sources if not present in the
system from libtool.
Internal raptor and rasqal sources removed - now use
pkg-config(1) to find external libraries to compile and
link to.
Fixes for AVL Tree: update parent pointers when deleting - patch
from Aymeric Barthe - thanks.
Added internal support for upcoming Raptor V2 API. Not yet enabled.
Added librdf_world_get_raptor() and
librdf_world_set_raptor() methods.
Added librdf_parser_guess_name2() with librdf_world parameter
and deprecated librdf_parser_guess_name() without it.
Wrapped deprecated functions with REDLAND_DISABLE_DEPRECATED
blocks.
The main changes in this release are:
Updated to use
Rasqal 0.9.16
(Rasqal 0.9.16 release notes)
from 0.9.15. NOTE: Rasqal's ABI and API changed so Redland 1.0.8
requires 0.9.16 or newer.
Updated to use
Raptor 1.4.18
(Raptor 1.4.18 release notes)
from 1.4.16.
Added a new in-memory indexed storage 'trees' using AVL Trees
Issues Fixed:
- 0000256: mysql reconnect not working (MySQL v5 only?)
NOTE: The next release of redland will NOT
include raptor and rasqal in the tarball, they will be separate
download dependencies.
NOTE: In the next release of redland, the
redland-config utility that provides the compile and
link args will be deprecated, generate a warning when run and will
turn into a wrapper for pkg-config redland ...
It will be removed in a future release.
General Changes
configure was updated to ensure that redland 1.0.8+
will work with rasqal 0.9.16 to 0.9.99 only.
Parser Class Changes
librdf_parser_set_feature() now handles setting
feature values as strings or integers, depending on the required
type (although Raptor will handle this too).
Added librdf_parser_parse_file_handle_as_stream() and
librdf_parser_parse_file_handle_into_model() helper
methods.
Serializer Class Changes
librdf_serializer_set_feature() now handles setting
feature values as strings or integers, depending on the required
type (although Raptor will handle this too).
Storage Class Changes
Added a new in-memory storage named "trees" by Dave Robillard that
uses AVLTrees internally for faster indexed triple store/querying.
This is an alpha quality storage since it is brand new and it may be
changed in future.
Other Changes
LIBRDF_URI_RDF_MS, LIBRDF_URI_RDF_SCHEMA
public API macros changed definition to pass in a world parameter.
Added librdf_get_concept_ms_namespace() and
librdf_get_concept_schema_namespace() to make the
concept macros above work.
Added an rss2atom example program.
Update configure to use autoconf 2.5.x AC_CHECK_TYPES
for testing byte, u32 and u64.
Fixes for resilence under low memory and allocation failures.
Internal Changes
Applied some updates from running the Linux 'sparse' utility.
Allow get_context method calls in stream/iterator map functions.
Originally
reported on redland-dev 6 May 2008
by John Fieber.
The storage method
librdf_storage_find_statements_in_context() now uses the
storage factory method find_statements_in_context if it exists,
instead of always using the find_statements factory method and
filtering.
The main changes in this release are:
Updated to use
Rasqal 0.9.15
(Rasqal 0.9.15 release notes)
from 0.9.14. NOTE: Rasqal's API will change in 0.9.16 so Redland 1.0.7
requires 0.9.15 exactly - no newer, no older.
Updated to use
Raptor 1.4.16
(Raptor 1.4.16 release notes)
from 1.4.15.
Added modular storage backends. When
--enable-modular is given to configure, all the storage
major backends are compiled as dynamically loadable modules. The
memory, file and uri storages are always available. This is not
enabled by default as it is a new feature.
Issues Fixed:
- 0000187: "rdf_query_results.c", line 630: void function cannot return value
- 0000189: librdf_uri_is_file_uri: wrong return value
- 0000203: API declarations for Symbian portability
- 0000205: Memory leak in rdf_query_rasqal.c
- 0000209: Null dereference + assertion failures in minimal librdf setup-cleanup
- 0000213: Multiple instances of world object make trouble
- 0000215: librdf_new_node_from_typed_counted_literal() does not respect parameter lengths
- 0000218: librdf_query_execute does not assert model parameter
- 0000229: Segmentation fault when invoking librdf_new_model_from_model()
- 0000230: on Gentoo/FreeBSD redland fails compilation due to a missing #define pthread.h
- 0000234: autogen.sh $dir quoting
- 0000235: $(EXEEXT) in tests
- 0000236: EOL issues when building svn version on cygwin
General Changes
configure was updated to ensure that redland 1.0.7
will work with rasqal 0.9.15 only due to the future API change coming
in the next rasqal version.
Make lots of writable static data truly constant especially from
writable static data from rdf_concepts. This causes a
SOURCE COMPATIBILITY BREAK: the
rdf_concepts.h public API node and uri
macros now require a librdf_world* parameter.
Parser Class Changes
Added librdf_parser_set_uri_filter() and
librdf_parser_get_uri_filter() to set and get parser
URI filters, passed on into raptor.
Storage Class Changes
Several SQLite storage improvements and fixes were made:
- Added transaction support
- Handle SQL errors more gracefully
- the contains_statement method now works with a blank subject node.
- Resource allocation and low mwmory fixes
URI Class Changes
Added librdf_uri_compare() to compare uri objects
rather than rely on stringifying them and comparing the strings.
Internal and Other Changes
Many low memory and error path, resource leak fixes for parsers,
streams, module factory construction and module initialising.
librdf_serializer_set_namespace() now allows NULL uri
and prefix to be given.
librdf_storage_mysql_transaction_commit() now
terminates if there is no work to commit.
autogen.sh version comparsion was fixed so that 1.10 is seen as
newer than 1.9.
The main changes in this release are:
Updated to use
Rasqal 0.9.14
(Rasqal 0.9.14 release notes)
from 0.9.13.
Updated to use
Raptor 1.4.15
(Raptor 1.4.15 release notes)
from 1.4.13.
Plus a new transactions API for changes to the graph, implemented
and tested for MySQL.
Added a new query results formatter class to turn a query results
object into a syntax.
Model Class Changes
Added a model transaction API:
librdf_model_transaction_start() to begin the transaction
librdf_model_transaction_start_with_handle() to begin it
using an existing connection pointer (such as a MySQL connection)
librdf_model_transaction_commit() to commit the changes
librdf_model_transaction_rollback() to abort them (automatic
if not commited before a close)
librdf_model_transaction_get_handle() to
get a handle associated with a connection.
Fixes Issue #0000038
Implemented the Transaction API for MySQL (tested) and PostgreSQL
(not tested).
Invert test in librdf_model_add_submodel().
Fixes Issue#0000175
Parser Class Changes
Added librdf_parser_guess_name() to guess a parser
name from content, similar to how a parser is constructed from
content.
Added functions to return the namespaces declared during a parse:
librdf_parser_get_namespaces_seen_count() to return the
number seen, librdf_parser_get_namespaces_seen_prefix() to
get the prefix string and
librdf_parser_get_namespaces_seen_uri() to get the
librdf_uri*.
Query and Query Results Class Changes
Added a new Query Results Formatter class building
librdf_query_results_formatter* objects
to turn a query results into a syntax such as SPARQL query results
formats, JSON. New functions:
- Constructors:
librdf_new_query_results_formatter() and
librdf_new_query_results_formatter_by_mime_type().
- Destructor:
librdf_free_query_results_formatter()
- Method:
librdf_query_results_formatter_write()
to write the query results to an iostream.
- Helper functions:
librdf_query_results_formats_check()
and librdf_query_results_formats_enumerate()
Added librdf_query_results_is_syntax() to
check if a query result is a syntax, and not any of the other
possibilities - bindings, boolean or graph.
Serializer Class Changes
Default to serializer named "rdfxml" if no name is given to ensure
another rdfxml serializer such as XMP is not returned.
Storage Class Changes
The MySQL storage implements the transaction API. It batches up
all the pending insertions and aggregates them so that any set of
triple writes during the transaction sequence are all done at one
point, with 4 INSERTs only.
Added a supporting
functions librdf_hash_interpret_template() to
interpolate variables into strings using a librdf_hash of
key/values.
The MySQL storage now takes an optional storage option 'layout'
that can choose between the v1 and v2 SQL schemas (only the table
type has changed). The SQL fragments are described in Turtle files
that are installed and read on startup.
Configuration and Build Changes
Packagers note: two new files are installed into
PREFIX/lib/redland for the MySQL schemas. Without them installed,
the MySQL storage will not work.
Removed the configure script option
--enable-parsers since Raptor is always required and
always used inside Redland whether or not any parsers were configured.
Use pkg-config Libs.private in
redland.pc for internal dynamically linked libraries.
redland-config gains an --private-libs option to hold
private libraries moved from --libs which now only contains -L and a
-lrdf.
Added a building utility touch-mtime.pl
to handle touch -r which does not work on SunOS.
Fixes
Issue #0000159
Other Changes
Added librdf_world_open() for every public API
constructor or function that mentions librdf_world* and could have
failed to run librdf_world_open() after
librdf_new_world(). This makes
librdf_world_open() entirely optional.
Fixes
Issue #0000173
rdfproc gains an --transactions / -T option for
enabling transactions on the command line and an --results / -r
option to set the query results formatter name.
Do not throw away the rasqal node type information now, it is needed
later on in FILTER comparisons.
Fixes
Issue #0000153
Fix reference count problem in storages that caused connections
to fail to free in MySQL.
Fixes
#0000150
Fix reference count / memory leak in file storage.
The main changes in this release are:
Updated to use
Rasqal 0.9.13
(Rasqal 0.9.13 release notes)
from 0.9.12.
Updated to use
Raptor 1.4.13
(Raptor 1.4.13 release notes)
from 1.4.9.
Plus several API additions and many bug fixes.
Configuration and Build Changes
configure now removes un-necessary tests for C++ or
F77++ compilers that libtool stupidly insists on. Add lots of
gcc -W flags when they are supported and in maintainer
mode.
Made configure --with-threads work.
Fixes
Issue #0000125
Model Class Changes
Methods
librdf_model_add,
librdf_model_add_statement,
librdf_model_add_typed_literal_statement,
librdf_model_add_string_literal_statement and
librdf_model_contains_statement
now enforce only allowing legal RDF triples.
Added librdf_model_enumerate to list all the
model implementations - currently just one, a model over
a storage.
Node Class Changes
The node class can now encode and decode long literals with length
>65535 using new encoding type 'N'.
Added new constructor
librdf_new_node_from_typed_counted_literal
to build a typed node with a counted literal string.
Parser Class Changes
Changes to allow NULL base uris in parsing methods, but to fail if
it is omitted but required. Applies to
librdf_parser_parse_as_stream,
librdf_parser_parse_into_model,
librdf_parser_parse_string_as_stream,
librdf_parser_parse_string_into_model,
librdf_parser_parse_counted_string_as_stream
and
librdf_parser_parse_counted_string_into_model
Added librdf_parser_get_accept_header to get
the Accept header that would be sent with a parser
HTTP fetch of content.
Query Class Changes
Graph names in SPARQL now connect up to redland context nodes
to allow SPARQL GRAPH to work.
Serializer Class Changes
Added new methods to serialize librdf_streams of triples rather
than just librdf_model:
librdf_serializer_serialize_stream_to_file_handle,
librdf_serializer_serialize_stream_to_file,
librdf_serializer_serialize_stream_to_string,
librdf_serializer_serialize_stream_to_counted_string and
librdf_serializer_serialize_stream_to_iostream.
Fixes
Issue #0000092
Storage Class Changes
Methods librdf_storage_add_statement,
librdf_storage_add_statements and
librdf_storage_contains_statement
now enforce only allowing legal RDF triples.
The Memory storage now supports deleting while iterating or
streaming the storage.
The PostgreSQL storage now accepts option 'database' like
MySQL and other Redland relational databases in addition to the
existing option it accepted, 'dbname'.
SQLite storage fixes:
Fix crash in librdf_model_context_remove_statements
Fixes
Issue #0000103
Wrap all sqlite_FREE calls on error messages so that
they are only called with SQLITE API v2.
Fixes
Issue #0000105
Check for no language or no datatype when they are NULL
in librdf_storage_sqlite_literal_helper.
Fixes
Issue #0000136
Store pending queries while the database is locked doing
a SELECT and do them once the select is done. Such as
deleting triples during a find or serialise.
Fixes
Issue #0000139
SQLite storage now has a 'synchronous' storage option which takes
values 'off', 'normal' or 'full' allowing access to the sqlite pragma
to control when commits are synchronised to file.
Other changes
rdfproc now allows the base URI for parsing to be set
to null with - and adjust messages to handle this.
Added some more const fixes to heuristics.
Fixes
Issue #0000107
Autodocs for librdf_new_node_from_blank_identifier.
Fixes
Issue #0000114
If librdf_storage_sqlite_statement_operator_helper fails,
free the stringbuffer.
Fixes
Issue #0000116
Do not read from stream when parsing returns a NULL pointer.
Fixes
Issue #0000130
Fix debug call in librdf_query_register_factory,
librdf_storage_hashes_add_remove_statement and
librdf_storage_hashes_contains_statement
Fixes
Issue #0000132
Added rss2ical.c example to turn RSS/Atom 1.0 feeds into
iCalendar using raptor, redland and SPARQL.
Many minor fixes and changes due to adding lots of -W flags
and using gcc4.
SWIG Bindings Interface
Fixed typo to declare librdf_query_results_get_binding_value
to return a new object.
Export new function librdf_parser_get_accept_header
Export version strings and integer constants for raptor:
(raptor_version_string,
raptor_version_major,
raptor_version_minor,
raptor_version_release,
raptor_version_decimal)
and rasqal
(rasqal_version_string,
rasqal_version_major,
rasqal_version_minor,
rasqal_version_release,
rasqal_version_decimal)
The main changes in this release are:
Updated to use
Rasqal 0.9.12
(Rasqal 0.9.12 release notes)
to update to SPARQL 2006-04-06 for JSON results
Updated to use
Raptor 1.4.9
(Raptor 1.4.9 release notes)
Plus fixing some crash bugs in parsing and serializing.
Configuration and Build Changes
Now using Subversion
for version control and the Redland installation instructions explain how to get Redland from Subversion.
Win32 fixes and VC build files updates from John Barstow.
Let configure find BDB 4.4
Node Class Changes
librdf_new_node_from_literal
and librdf_new_node_from_typed_literal now accept empty
strings as an alternative to a NULL language pointer.
Parser Class Changes
Add one more place where it needed handling old and new raptor
statement predicate values
RAPTOR_IDENTIFIER_TYPE_RESOURCE and
RAPTOR_IDENTIFIER_TYPE_PREDICATE.
Fixed a bad uses of flose() causing a crash in parsing a file or
from a FILE* file handle when run under
newer glibcs.
Serializer Class Changes
Fixed a bad uses of flose() causing a crash in serializing to a
file when run under newer glibcs.
Storage Class Changes
Introduced a UINT64_T_FMT to handle formatting a 64-bit
unsigned int in storages
as %llu is not portable.
The main changes in this release are:
Updated to use
Rasqal 0.9.11
(Rasqal 0.9.11 release notes)
to update to SPARQL 2005-11-23
Updated to use
Raptor 1.4.8
(Raptor 1.4.8 release notes)
A large source re-arrangement was performed for raptor and rasqal
inside redland. All C sources and headers that build the libraries
were moved to the src dir, general documentation in the doc dir and
utilities in the utils dir for each sub-library.
Version Control change: Redland will be switching to use
Subversion
for version control after the 1.0.3 release. Please check the
Redland Subversion site
for the latest status or the online
Redland installation notes
for the redland specific subversion installation information.
A new PostgreSQL storage backend was added
contributed by Shi Wenzhong based on the MySQL storage backend.
Fixes issue #0000046
Configuration Changes
The autogen.sh script for building from CVS was revamped to be
more modular.
configure now takes an --enable-gtk-doc
option to enable building of the documentation using the
gtk-doc utility. It is by default enabled only if the
utility is available.
Documentation Changes
The GNOME gtk-doc program is now used to
automatically extract documentation from source comments into
reference documentation. This is then merged with templates and
additional documentation to provide a reference manual for redland
as XML document which is turned into HTML along with GNOME devhelp
support.
Portability Changes
Some win32 build fixes were made.
World Class Changes
librdf_world_get_genid() was changed to include the
process ID in the generated blank ID.
Fixes Issue #0000037
(patch from Marc Powell).
Node Class changes
librdf_node_encode(): Give an error when a node cannot
be encoded due to string data being too long.
Serializer Class changes
Added librdf_serializer_serialize_model_to_iostream(), not
exposed in the SWIG bindings.
Storage Class Changes
Added a PostgreSQL storage backend contributed by Shi Wenzhong
based on the MySQL storage backend.
Fixes Issue #0000046
Other Changes
Deprecated librdf_files_temporary_file_name().
Deleted some function prototypes defined in headers but never had
code:
librdf_node_init(),
librdf_iterator_finished(),
librdf_storage_get() and
librdf_world_set_uris_hash()
Added examples/example7.c to read from an RDF/XML file, add a
triple and write it back.
The changes in this release are:
- Fix a crash when querying or searching with the model
find_statements method and getting no results
- Update to use
Rasqal 0.9.10
(Rasqal 0.9.10 release notes)
to provide query results sorting such as SPARQL
ORDER BY
and some SPARQL syntax updates.
- Update to use
Raptor 1.4.7
(Raptor 1.4.7 release notes)
to fix some crashes in the RSS tag soup parser and serializer.
The main changes in this release are to update to use
Rasqal 0.9.9
(Rasqal 0.9.9 release notes)
to provide improved SPARQL query support for the revised syntax
Raptor 1.4.6
(Raptor 1.4.6 release notes)
to
provide a new GRDDL parser as well as updated RSS enclosures support
with the RSS tag soup parser and RSS 1.0 serializer.
The digest and hash classes now expose public API constructors,
destructors and methods via their header files.
Configuration changes
Added --with-memory-signing to check when memory is
allocated / freed across Redland libraries.
Portability Fixes
Added Win32 portability fixes and build configuration (John Barstow)
Binding API changes
The following changes were made to the bindings API as
exported by the SWIG interface file Redland.i:
The digest class was added:
constructor librdf_new_digest,
destructor librdf_free_digest
and methods librdf_digest_init,
librdf_digest_update,
librdf_digest_update_string,
librdf_digest_final and
librdf_digest_to_string,.
The hash class was added as selected functions:
constructors librdf_new_hash_from_string and
librdf_new_hash_from_array_of_strings and destructor
librdf_free_hash.
The query class methods
librdf_query_get_limit,
librdf_query_set_limit,
librdf_query_get_offset and
librdf_query_set_offset were added.
Functions were added for accessing parts of structured log messages
as librdf_log_message objects:
librdf_log_message_code,
librdf_log_message_level,
librdf_log_message_facility,
librdf_log_message_message,
librdf_log_message_locator
and Raptor raptor_locator
objects:
raptor_locator_line,
raptor_locator_column,
raptor_locator_byte,
raptor_locator_file and
raptor_locator_uri.
Digest Class changes
This class was added to the public API in this release adding
constructor librdf_new_digest,
destructor librdf_free_digest
and methods
librdf_digest_init,
librdf_digest_update,
librdf_digest_update_string,
librdf_digest_final,
librdf_digest_get_digest,
librdf_digest_get_digest_length,
librdf_digest_to_string and
librdf_digest_print
Hash Class changes
This class was added to the public API in this release adding
constructors
librdf_new_hash_from_string,
librdf_new_hash_from_array_of_strings and
librdf_new_hash_from_hash;
destructor
librdf_free_hash;
and methods
librdf_hash_get,
librdf_hash_get_as_boolean,
librdf_hash_get_as_long,
librdf_hash_get_del,
librdf_hash_put_strings,
librdf_hash_print,
librdf_hash_print_keys and
librdf_hash_print_values.
Iterator Class changes
Added a librdf_new_empty_iterator helper to make an
always-empty iterator.
Log Class changes
Added LIBRDF_FROM_MEMORY facility.
Model Class changes
All methods returning an librdf_iterator or
librdf_stream now return an empty iterator/stream rather
than NULL when returning an empty result sequence. Errors are still
returned as NULL.
Node Class changes
Removed unused LIBRDF_NODE_TYPE_RESERVED1 (#3) from
librdf_node_type enum.
Query Class changes
Update to use Rasqal 0.9.9 API. Redland 1.0.1 will not build with
older Rasqal releases.
The query API gains methods librdf_query_get_limit,
librdf_query_set_limit,
librdf_query_get_offset and
librdf_query_set_offset for setting and getting query
result limit (max results) and offsets (result to start from).
Fixed some across-library memory allocation/frees between redland
and rasqal that caused problems on win32. Requires rasqal 0.9.9 or
newer.
Stream Class changes
Added a librdf_new_empty_strea helper to make an
always-empty strea.
Storage Class changes
The MySQL store now uses a portable method across endiannesses to
create a 64bit key. This fixes
Issue #0000023
(Morten Frederiksen)
NOTE: This matches the algorithm used for
little-endian systems but big-endian stores will need to serialize
the store before upgrading and reload afterwards.
The MySQL store now uses connection pooling to prevent making a new
connection to the database for each request.
Fixes Issue #0000027
(Morten Frederiksen)
The SQLite store now does not skip items with the factory methods
for serializing, finding statements, serializing contexts and getting
context statements.
Fixes Issue #0000006
All store methods that return an librdf_iterator or
librdf_stream now return an empty iterator/stream rather
than NULL when returning an empty result sequence. Errors are still
returned as NULL.
Other changes
Added REDLAND_INLINE macro, which can be overridden if
inline is not wanted.
This is Redland 1.0.0.
The main changes in this release are to update to use
Rasqal 0.9.6
to provide improved SPARQL query support for optionals, CONSTRUCT, ASK, builtins,
Raptor 1.4.4 to
provide better serializing to RDF/XML and RSS 1.0 and better
RSS tag soup parsing and
a new experimental SQLite
storage supporting SQLite V2 and V3 libraries.
Configuration changes
The storages compiled into redland can be individually selected
with the --enable-storages=LIST configure
option.
The default for --with-dmalloc is now no. Reason:
valgrind.
Portability Fixes
Removed configure altering the search PATH - which was a bad idea
as it makes the configure environment different from the user's.
Added /opt/local to the BerkeleyDB / SleepycatDB
search path to help darwinports.
To help Redland built as an OSX Framework with ObjectiveC, Redland
public headers can be put in a different directory structure if
LIBRDF_OBJC_FRAMEWORK is defined in the build
environment. In this case the headers are found in a subdirectory
Redland. This only applies to the public Redland header
files that #include other public header files -
redland.h including librdf.h,
librdf.h including several rdf_*.h files
and Rasqal's rasqal.h including
raptor.h. (Patch from René Puls)
rdfproc RDF processor utility program changes
Added a size command calling
librdf_model_size.
Model Class changes
Added librdf_model_contains_context to find a context
node in a model.
The librdf_model_sync now returns a success or failure
int.
Parser Class Changes
Updated to Raptor 1.4.4
providing an updated RSS tag soup parser with fixes for RSS 0.9,
RSS.1.1 and Atom.
The constructor now finds the correct parser when given NULL
arguments for mime type or type URI.
Query Class changes
Update to Rasqal 0.9.5
providing improved SPARQL support: new optionals handling,
CONSTRUCT to make RDF graphs, formatting of query
results in the SPARQL XML format along with many other improvements
and fixes.
Added base_uri argument to librdf_new_query
Added
librdf_query_results_to_counted_string,
librdf_query_results_to_string,
librdf_query_results_to_file_handle and
librdf_query_results_to_file
to get the results of a query written to a syntax using
Rasqal's rasqal_query_results_write and Raptor's XML writer API.
Added boolean returning methods:
librdf_query_results_is_bindings,
librdf_query_results_is_boolean,
librdf_query_results_is_graph for testing result formats
and librdf_query_results_get_boolean to get a boolean
result back.
Serializer Class changes
Updated to Raptor 1.4.4
providing a new XML writer API, an improved RDF/XML serializer
allowing user namespace declarations and relative URIs, a new RSS 1.0
serializer.
The serializers now respect any user-declared namespaces done with
librdf_serializer_set_namespace by passing it on to the
Raptor serializer with raptor_serialize_set_namespace.
At present only the Raptor RDF/XML serializer uses these
user-declared namespace hints.
The constructor now finds the correct serializer when given NULL
arguments for mime type or type URI.
Storage Class changes
The storages compiled into redland can be individually selected
with the --enable-storages=LIST configure option.
SQLite Storage. A new experimental storage supporting
SQLite V2 and V3 libraries.
MySQL storage. Return failure early if the connection failed
rather than try to continue opening the storage.
MySQL storage. Only use mysql_real_escape_string if
a connection was made. Do not try to free resources such as the
mysql DB connection owned by iterators after an error, let
librdf_free_iterator tidy up. (Patch from Morten
Frederiksen)
URI Class changes
Handle resolving against a base URI with an omitted path.
Other changes
Removed librdf_concept_labels - never used or
exported.
This is Redland 1.0 Release Candidate 2.
Redland's License was changed from LGPL 2.1/MPL 1.1 to
LGPL 2.1/Apache 2
The main changes in this release are to update to use
Rasqal 0.9.5
to provide initial SPARQL query support and to use
Raptor 1.4.2 to
provide serializing to RDF/XML and N-Triples.
rdfproc RDF processor utility program changes
Updated to allow an optional base URI for the
serialize command and to handle the three forms
of query results - bindings, graph or boolean.
Model Class changes
Added librdf_model_to_string and
librdf_model_to_counted_string to serialize a model to a
string using the Serializer class functionality. Exported
librdf_model_to_string to the language bindings.
Query Class changes
Now requires
Rasqal 0.9.5
which provides initial support for the
SPARQL Query Language for RDF,
W3C Working Draft, 12 October 2004
from the W3C RDF Data Access Working Group (DAWG).
Serializer Class changes
Pass on errors and warnings via the existing Redland logging
mechanism. Previously they all went to stderr, the default.
Switched to use Raptor 1.4
series serializing classes so now provides RDF/XML and N-Triples
serializing and will gain any fugure serializing formats as Raptor
adds them.
Using the Raptor support, the serializers can now serialize to
strings in addition to the existing serializing to FILE* and files.
This is performed by the new methods:
librdf_serializer_serialize_model_to_file_handle
(deprecating librdf_serializer_serialize_model),
librdf_serializer_serialize_model_to_string
and librdf_serializer_serialize_model_to_counted_string.
Exported librdf_serializer_serialize_model_to_string
to the language bindings.
Storage Class changes
librdf_storage_context_add_statement and
librdf_storage_context_add_statements -
With a NULL context, call the non-context method as documented.
Fixed bogus '+' in get_contexts method in the mysql
storage.
World Class changes
librdf_world_get_feature now returns a
librdf_node*, librdf_world_set_feature
takes a librdf_node* value. Export these to
the language bindings.
Revert the statement class usage counting - it broke things.
This is Redland 1.0 Release Candidate 1.
The main changes in this release are the addition of the
Query and
Query Results classes, and
the removal of the language bindings into the new
Redland Bindings
package.
The interface between the Redland library and Redland Bindings
when done via swig is by the
swig interface file Redland.i which is now installed in
/usr/share/redland/Redland.i.
Portability changes
Several patches for building Redland natively on Win32
were provided by Jose Kahan. These included
adding an win32_rdf_config.h hard-coded config
for Win32, using that in the all the C source files and adding an
REDLAND_STATIC define to help statically linking Win32
builds of Redland.
Configuration changes
Redland 0.9.17 requires Raptor 1.3.2 and Rasqal 0.9.1. Installed
versions are used if available by default, otherwise the copies in
the source tree will be used. The
--with-raptor= and
--with-rasqal= configure options with
values internal or system can be
used to select when both are available.
Removed all the language binding-specific configuration code
including options such as --with-perl that are now used
by Redland Bindings.
Packaging changes
Updated the RPM spec file to remove building of the redland-perl
and redland-python RPMs which are now made in the redland-bindings
package.
Added -lrdf to the pkgconfig redland.pc file (Curtis Hovey)
Redland requires automake 1.7 and autoconf 2.54 to build from CVS.
Query and Query Result Classes
The Query
(librdf_query) and
Query Results
(librdf_query_results) classes are new in Redland 0.9.17.
They provide RDF query support with
RDQL
using Rasqal
0.9.1, which is now included in the Redland distribution.
The Query class has the following functions and methods:
librdf_new_query (constructor)
librdf_new_query_from_query (constructor)
librdf_new_query_from_factory (factory constructor)
librdf_free_query (destructor)
librdf_query_execute
The Query Results class has the following methods:
librdf_query_results_as_stream
librdf_query_results_get_count
librdf_query_results_next
librdf_query_results_finished
librdf_query_results_get_bindings
librdf_query_results_get_binding_value
librdf_query_results_get_binding_name
librdf_query_results_get_binding_value_by_name
librdf_query_results_get_bindings_count
librdf_free_query_results (destructor)
A new example example5.c was created to demonstrate using the
query classes.
New Logging API
An enhanced logging API was added to provide enhanced information
from log messages, error and warnings such as from parsing. The new
world method librdf_world_set_logger allows setting of
the log handler callback which gives a
librdf_log_message. Parts of this the structure can be
read by using the following new accessors methods:
librdf_log_message_code
librdf_log_message_level
librdf_log_message_facility
librdf_log_message_message
librdf_log_message_locator
The locator returns an raptor_locator that gives the
line, column, byte offset, URI and/or filename of the message.
Raptor 1.3.2+ provides accessor methods for that structure
that are described in
libraptor.3.
Model Class changes
Check that duplicate statements are not added to models with the
librdf_model_add_statement and
librdf_model_add_statements methods. The methods for
adding statements with a context do not check.
Added librdf_model_query_execute method to execute a
librdf_query against a model, returning an
librdf_query_results
Added librdf_model_load method that uses Raptor
parser guessing to load content from a URI into a model.
The contexts methods now give a run time warning if called when
the model does not support contexts or the are not enabled. The
methods affected are:
librdf_model_context_add_statement
librdf_model_context_add_statements
librdf_model_context_as_stream
librdf_model_context_remove_statement
librdf_model_context_remove_statements
librdf_model_context_serialize
librdf_model_find_statements_in_context
librdf_model_get_contexts
Fixed a memory leak
inside librdf_model_find_statements_in_context.
Node Class changes
Fix assert bug in librdf_node_get_blank_identifier
that caused it to always fail.
A bug was fixed in librdf_node_get_li_ordinal
to not always fail, returning -1
Parser Class changes
Deprecated methods librdf_parser_set_error and
librdf_parser_set_warning which never worked. The
new logging API works correctly with this code.
Storage Class changes
Where possible, check that duplicate statements are not added to
stores with the librdf_storage_add_statement and
librdf_storage_add_statements methods. The methods for
adding statements with a context do not check.
The "mysql" storage was updated by
Morten Frederiksen
to revise the schema and to add two new features,
controlled by new boolean storage options:
bulk: whether model/storage method add_statements
should be optimized, until a model/storage sync operation
merge: whether to maintain a table with merged models
Context methods will now give a run time warning if called and
contexts are not supported or enabled. The methods affected are:
librdf_storage_context_add_statement
librdf_storage_context_remove_statement
librdf_storage_context_serialise
librdf_storage_get_contexts
Fixed a memory leak
inside librdf_storage_find_statements_in_context.
A bug was fixed in method
librdf_storage_node_stream_to_node_create to allow the
second node parameter to be NULL, so that methods
librdf_storage_get_arcs_out and
librdf_storage_get_arcs_in that can use it to work.
A small but important fix to make the installed C headers work
correctly again. The rdf_uri.h file had #include
rdf_digest.h which is an internal header and not installed.
This required removing two functions from the public API that
returned librdf_digest*, but neither of them could be
used since rdf_digest.h was not shipped.
Patched librdf.h and raptor.h to #include stdio.h so
that the empty program #include <redland.h>
main() {} now works (ditto for raptor).
This version of Redland updates to use
Raptor
1.2.0 (raptor news)
along with the new syntax
Turtle Terse RDF Triple Language.
Three new storage modules were added:
- MySQL store written by Morten Frederiksen supporting the entire storage API include full contexts support.
- AKT Triplestore using MySQL inside 3store
- File and URI-backed in-memory store - read/write file, read only URI.
The details of these modules are explained in much greater detail
in the storage modules overview
including the features that they support.
The runtime is now compiled with assertions to check for illegal
use of NULL pointers in arguments, these will give warning messages
(which can be disabled with --disable-assert-messages).
The storage and model classes now track when open iterator and
streams are operating on them and will not release resources until
all the open iterators and streams are finished.
URI and Literal string type changes
There are many changes are to properly use and return UTF-8
strings as unsigned char* for URIs (looking forward to
Internationalized Resource Identifiers, IRIs emerging) and for RDF
literal strings, plus other fixes to work better with C++ correcting
the use of const. Redland and Raptor now both compile
cleanly with g++ 3.2, 3.3 and all warnings enabled.
rdfproc RDF processor utility program changes
rdfproc was extensively updated
and it can now more easily work with any type of Redland store and
has better support for working with contexts. The detailed changes
are as follows:
- Added an -s TYPE option to specify a storage type
- Added an -t option for giving storage options
- Added an -p option to read the value of
storage option
password from standard input, to prevent
exposing it as a parameter.
- The default of creating a new store was
replaced; -n enables truncation or overwriting (the
write=yes storage option).
- Now recognises
_:abc as blank node arguments.
- Parsing now return counts of errors and warnings if possible.
- The parse-stream command can take an optional context node
- Added a contexts command, to list the contexts in a model.
- Added support for environment variables RDFPROC_STORAGE_OPTIONS and
RDFPROC_STORAGE_TYPE to provide default storage options (-t) and type (-s)
- Help message updated and correctly lists supported parsers
Configuration changes
Redland 0.9.15 requires Raptor 1.2.0 either installed otherwise
the copy in the source tree will be installed.
--with-mysql=CONFIG added to point to the
mysql_config program for the MySQL backing store
--with-threestore added to enable the
AKT Triplestore
--disable-assert added to disable
compiling run-time assertions.
--disable-assert-messages added to disable
compiling run-time assertion messages.
Packaging changes
A redland-python RPM package is now provided and the default RPM
packaging now provides the MySQL store, in addition to the in-memory
and BDB previously packaged. The new file and uri stores are also
available. The full list of store modules is described further in the
storage modules overview.
Model Class changes
Methods
librdf_model_add_string_literal_statement and
librdf_model_add_typed_literal_statement
now take a const unsigned char* string.
Method
librdf_model_query_string now takes
a const unsigned char* string query.
New methods were added
librdf_model_find_statements_in_context
to search for a statement matching only in a particular context,
librdf_model_find_statements_with_options to
allow options with searching (none used at present),
and librdf_model_get_contexts to
list the context nodes in a graph.
Methods librdf_model_get_feature,
librdf_model_set_feature were modified to take and return
librdf_node* values.
Node Class changes
Constructors
librdf_new_node_from_uri_string,
librdf_new_node_from_uri_local_name and
librdf_new_node_from_normalised_uri_string
now take a const unsigned char*
string URI string.
Constructors
librdf_new_node_from_literal and
librdf_new_node_from_typed_literal
now take a const unsigned char* string RDF literal.
Constructor librdf_new_node_from_blank_identifier
now takes a const unsigned char* string blank node identifier.
Methods
librdf_node_get_literal_value,
librdf_node_get_literal_value_as_counted_string and
librdf_node_get_blank_identifier
now return a const unsigned char* string RDF literal,
blank node identifier
(Debugging methods librdf_node_to_string and
librdf_node_to_counted_stringwhere changed to return a
const unsigned char* string)
Parser Class changes
Added methods
librdf_parser_parse_counted_string_as_stream and
librdf_parser_parse_counted_string_into_model
for parsing a string with a given length rather than
a C NUL-terminated string.
Methods
librdf_parser_parse_string_as_stream and
librdf_parser_parse_string_into_model
now take a const unsigned char* string syntax buffer.
Methods
librdf_parser_get_feature
and librdf_parser_set_feature
were modified to take and return librdf_node* values.
Query Class changes
Constructor librdf_new_query now takes a const
unsigned char* string query.
Serializer Class changes
Added methods librdf_serializer_get_feature and
librdf_serializer_set_feature taking and returning
librdf_node* values like the methods for the model and
parser classes.
Statement Class changes
The statement objects are now usage counted like most other
objects so they are cheap to copy. No user-API level changes
are needed.
(Debugging method librdf_statement_to_string now
returns a const unsigned char* string)
URI Class changes
Constructors
librdf_new_uri,
librdf_new_uri_from_uri_local_name,
librdf_new_uri_normalised_to_base and
librdf_new_uri_relative_to_base.
now take a const unsigned char* string URI string.
Methods
librdf_uri_as_string,
librdf_uri_as_counted_string,
librdf_uri_to_string and
librdf_uri_to_counted_string
now return a const unsigned char* string URI string.
Java API Changes
Updated to match API changes above. Fixes for OSX 10.3
Perl API changes
Updated to match API changes above. Fixes for perl 5.8.1 and
later which changed how MakeMaker installed. Let perl
install the files rather than try to force it based on the prefix
argument to configure. Make Parser method parse_string_into_model
call the correct Redland function. Added OSX Java directories
to JDK search path.
PHP API Changes
Updated to match API changes above. Redland is now started up and
shutdown once in the PHP module code and does not need to called
directly.
Python API changes
Updated to match API changes above. Fix Node literal_values
to work with datatype URIs. Export the NS class. Added
a FileStorage class for the storage type file.
Removed use of the python Distutils which was often
getting the linking wrong.
Ruby API changes
Updated to match API changes above.
Tcl API changes
Updated to match API changes above.
Other changes
Redland now ships with both portable MD5 as well as (new in
0.9.15) SHA1 digest implementations used if neither is available via
an optimised library.
This version of Redland is an update mostly to synchronise
with the
Raptor version 1.0.0 (RDF/XML and N-Triples parser) - see
Raptor NEWS
for the changes since 0.9.12.
New in this version is the rdfproc utility
which allows command line creation of redland stores, parsing
of syntaxes and manipulation of the model.
Configuration changes
Added --with-raptor to either choose the system
or internal raptor library. If omittted, redland will guess
and choose either the system one, if new enough or the internal
one (always present).
Hash Class changes
Added a librdf_hash_put_strings helper method to put a pair of
string values into a hash. Useful for making model or storage
options.
Node Class changes
Resource (URI) / literal / blank nodes are now interned on
construction so there is only one object for each distinct
librdf_node object. This reduces memory and makes
copying cheaper at the cost of an increase in construction cost.
Copying nodes is very common, since not only is it done for nodes but
any time a statement is used or returned. This interning was already
done for the URI class but the change required
librdf_node to be made immutable, which was done in
0.9.13.
Storage Class changes
Added librdf_storage_context_as_stream deprecating
librdf_storage_context_serialise to be consistent with the
*as_stream method name changes to model.
Added librdf_new_storage_with_options taking an
librdf_hash*
of options rather than the string of librdf_new_storage.
Added new optional storage factory methods context_add_statements
and context_remove_statements. If omitted, the storage class
implements them with context_add_statement and context_remove_statement.
Stream and Iterator Classes changes
Optimised the stream and iterator internals to do less factory
method calls when looking for new items or the end of
stream/iterator. This should make them slightly faster.
Miscellaneous changes
Renamed Redland statics to be librdf_* rather than
redland_*. Added a one-line
librdf_short_copyright_string.
Perl API Changes
Fixed the broken use of UNIVERSAL::isa
This version of Redland is a major update to 0.9.12 changing the
node class. librdf_node objects are now immutable once created,
higher-level language APIs no longer have deal with sharing or
copying internals, some additional API support for parsing content
from strings.
This includes the stable and complete version 0.9.12 of the
Raptor parser
which supports all of the revised RDF/XML syntax.
Initial work was added on an ECMA CLI interface using
the C# ("C sharp") language.
Blank nodes generated by parsing the syntaxes are now
generated more unique (not guaranteed) based on the Redland
startup time and a serial number.
The Perl, Python and Java APIs now always return new nodes
and statements removing the need to consider sharing at this
level. This means that the following operation is legal in
Perl, Python and Java:
# perl
$n=RDF::Redland::Node->from_uri("http://example.org/foo");
$s=new RDF::Redland::Statement($n, $n, $n);
$model->add($s);
# python
n=RDF.Uri("http://example.org/foo")
s=RDF.Statement(n, n, n) # n is used as a resource RDF.Node here
model.add_statement(s)
Configuration
Several improvements were made to the building and configuring
including a better attempt to find matching Sleepycat/Berkeley DB
header and library files (and support BDB up to 4.1), searching for
installed Java JDKs and Tcl headers. Each of the language interfaces
can be individually enabled to be build, tested and installed with
the main C library.
Node Class changes
The librdf_node class was modifed to make objects
immutable once constructed, they cannot be modified. The following
set methods were consequently removed.
librdf_node_set_blank_identifier
librdf_node_set_li_ordinal
librdf_node_set_literal_value
librdf_node_set_type
librdf_node_set_typed_literal_value
librdf_node_set_uri
The following utility methods were added (covering
all the librdf_node_get_type values):
librdf_node_is_blank - return non-0 if the node is a blank node
librdf_node_is_literal - return non-0 if the node is a literal
librdf_node_is_resource - return non-0 if the node is a URI
(These are reflected into the same methods on
the Perl, Python and Java node classes)
LIBRDF_NODE_TYPE_LI as returned by librdf_node_get_type
was deleted, it was never used.
To support datatyped literals, the following method was added:
- librdf_node_get_literal_value_datatype_uri - get the typed literal datatype URI of the literal node
Model Class changes
Added ibrdf_model_sync to sync the model to the backing store.
Added librdf_model_as_stream and
librdf_model_context_as_stream to replace
librdf_model_context_serialise and
librdf_model_context_serialize respectively. This was
done both to reduce the confusion with the
librdf_serializer class and the mixture of US/UK
spelling of that word. The old functions will remain for now.
Parser Class changes
Added librdf_parser_parse_string_as_stream and
librdf_parser_parse_string_into_model to allow parsing
content from strings.
Statement Class changes
Added librdf_statement_is_complete to test if a
statement has all the subject, predicate and object fields assigned.
Partial statements cannot be added to a model but can be used in
matching statement queries with
librdf_model_find_statements.
Storage Class changes
Added librdf_storage_sync to sync the backing store.
(This is slightly internal to Redland, you should not be calling this
directly)
Perl Interface Changes
Updated to make the object constructors less verbose and to allow
Redland URI objects and native perl URIs to be used in place of
Redland Nodes where convienent (such as in making Statements). All
existing code should work but can be shortened.
A new --with-perl configure argument to enable the perl
interface.
Python API changes
A major update to use Python 2.2+ idioms and features by Edd
Dumbill with help from Matt Biddulph for unit tests. A new
--with-python configure argument to enable the python
interface. Edd wrote the following change log.
General changes
- Removed string exceptions and replaced with RedlandError ones
- Where underlying Redland C objects were unexpectedly null, raise
exceptions rather than returning None, "" or [] as appropriate.
- Raise exceptions if Redland C constructors fail in Python
constructors.
- Did quite a lot of documentation updating, where possible
demonstrating preferred Pythonic usages.
- Added more unit tests to test changed functionality.
RDF.Node class
- Constructor now accepts Uri to make a resource/property node, or
a string to make a string literal
- Removed usage of node_type() in favour of is_resource(), is_blank()
etc methods.
- DEPRECATED get_blank_identifier() etc. methods in favour of
node.blank_identifier etc properties. Raise exception if properties
inconsistent with the node type are requested.
RDF.Statement class
- No need to name subject, predicate, object in constructor any more.
Uris or string literals accepted in place of nodes in constructor
argument.
- __getattr__ method deleted, using new-style property() for
subject, predicate, object
RDF.Model class
- Deleted get_*_iterator methods
- Renamed sources() to get_sources(), targets() to get_targets(),
arcs() to get_predicates(). Retained old method names as aliases.
- Added get_sources_context(), get_targets_context(),
get_predicates_context() which return (node, context) tuples.
- Amended get_source(), get_sources() etc methods so that Uri and
string literals were acceptable as shortcut types for nodes.
- Added find_statements_context() which returns (statement, context)
tuples.
- Renamed context_remove_statements() to
remove_statements_with_context(). Retained old method name as an
alias.
- Added __delitem__ method so del[statement] and del[statement,
context] work.
- Added append() method so append(statement) and append(statement,
context) work. Suggest that append() is used in preference to
add_statement() for Python idiomatic usage.
- DEPRECATED serialise() in favour of as_stream().
- Added as_stream_context() to serialise (statement, context) tuples.
- Exception raised if len() is attempted on model with a non-countable
storage.
- DEPRECATED Model.add(), Model.add_typed_literal_statement() usages:
now that constructing Statements is easier, it's incongruous to have
these in the model. Additionally it saves reproducing various bits
of error checking that properly belong in the Node() constructor.
- Implemented __contains__() so "if statement in model" and
"if (statement, context) in model" work as expected using the more
efficient Redland C methods where possible.
RDF.Stream class
- DEPRECATED Stream.context_iter(). Use in preference the _context
variants of the appropriate Model methods. This leads to fewer
lines of code.
RDF.Uri class
- Amended constructor so strings or Uri instances can be passed as
the first argument, without named parameters being required, and
the right thing still happen.
Java Interface Changes
Additions for the new API calls as outlined above (such
as model/storage sync, parsing from strings etc.). New configure
argument --with-jdk to pick the JDK in order to
find the JNI headers. See the java API
page for details.
PHP, Ruby and Tcl Interfaces Changes
Updates in configuring and building these interfaces and new
--with-php, --with-ruby and --with-tcl
configure arguments. See the corresponding API pages and
installation documentation for details. Tcl now tries to guess the
includes directories.
Internal changes and fixes
- Added Sleepycat/BDB 4.0 and 4.1 support - the open API changed again
- Split the
librdf_model class into model interface
(librdf_model) and implementation class
(librdf_model_storage).
- Removed the old and outdated repat RDF parser.
- Removed the very old expat XML parser from the sources;
much newer or expat or libxml2 are very widespread.
- Added simple thread support from patches by Seth Ladd with
configuration option
--with-threads that locks key shared
classes to allow separate threads to work in the same memory
space. The current locked classes are URI and World.
- Deleting statements in the in-memory store now works.
- Getting the context of iterators returned from in-memory
sources, targets, arcs now works.
This version of Redland is a major update to 0.9.11 incorporating
a much more improved, complete and stable version 0.9.7 of the
Raptor parser
and a significant new feature, contexts. The
high-level language APIs received some updates and the start of two
new languages APIs were added: Ruby and
PHP.
Several incompatible API changes were made in this release that
effect the C interface and all the higher level language APIs. These
were made to support contexts and to fix some inconsistencies in the
interface about object ownership.
The persistent storage format was changed to support RDF typed
literals and this requires an upgrade of any existing
Berkeley/Sleepycat DB stores created by Redland 0.9.11 or earlier. A
utility redland-db-upgrade is provided that will create an
updated store from an existing one.
A Perl script utils/update-api-0912.pl
can help automate the API changes as far as possible or warn about those
that cannot be automatically updated.
New Feature - Contexts
This feature allows a Node to be given whenever a statement is
added to a model which can can be retrieved from any model query that
returns answers as an Iterator of Nodes or a Stream of Statements.
Both of these classes gained a new method get_context
that returns the original Node that was given when the statement
corresponding to the answer was added to the model.
The context node can also be used to add and remove sets of
statements to/from a model, and each statement with a given
context node can be listed as a stream of statements.
Adding this feature required substantial internal changes to these
two classes and the internal storage apis and implementations along
with moderate code changes at the application level, which are
described below.
This feature can be used for the following (not an exhaustive list):
- Enable true graph merging / updating / demerging - identify the
subgraphs with context nodes.
- Statement Identity - add each statement with a different context node
- Statement Provenance - use the context node as the subject of
other statements about the statement that is returned.
Iterator Class Changes
The Iterator
get_next method was split into get_object
always returning a pointer to a shared object and the
next method to advance the iterator.
The get_object method is generally called
current in the higher level language APIs.
(Iterator and Stream now have consistent method names; they
may be merged in future.)
The get_object method now always returns a pointer to
a shared object. If this object is needed outside the scope of an
iteration, it must be copied, which for Nodes is the copy constructor
librdf_new_node_from_node.
C example for Redland 0.9.11 and earlier:
while(!librdf_iterator_end(iterator)) {
node=(librdf_node*)librdf_iterator_get_next(iterator);
/* do something with the node */
}
Redland 0.9.12:
while(!librdf_iterator_end(iterator)) {
node=(librdf_node*)librdf_iterator_get_object(iterator);
/* do something with the node */
librdf_iterator_next(iterator);
}
A new method get_context was added returning a Node
for the context of the original Statement that generated the Iterator
Node.
C Iterator context example:
while(!librdf_iterator_end(iterator)) {
node=(librdf_node*)librdf_iterator_get_object(iterator);
context_node=(librdf_node*)librdf_iterator_get_context(iterator);
/* do something with the node and its context */
librdf_iterator_next(iterator);
}
Model Class Changes
The Model
add_statement method no longer takes ownership of the
passed in statement object. The caller now retains ownership and can
reuse the statement several times and has responsibility for freeing it
if need be (when it isn't shared).
C example for Redland 0.9.11 and earlier:
librdf_model_add_statement(model, statement);
Redland 0.9.12:
librdf_model_add_statement(model, statement);
librdf_free_statement(statement);
C streaming statement example for Redland 0.9.11 and earlier:
while(!librdf_stream_end(stream)) {
librdf_statement *statement=librdf_stream_next(stream);
/* This statement is new, so could be added directly */
librdf_model_add_statement(model, statement);
}
Redland 0.9.12:
while(!librdf_stream_end(stream)) {
librdf_statement *statement=librdf_stream_get_object(stream);
/* This statement is now shared so can still be just added */
librdf_model_add_statement(model, statement);
}
librdf_model_add_string_literal_statement lost the
xml_space argument.
A new method was added to add a Statement with
an RDF datatyped literals object:
int librdf_model_add_typed_literal_statement(librdf_model* model,
librdf_node* subject,
librdf_node* predicate,
char* string, char *xml_language, librdf_uri *datatype_uri);
This should be used in preference to
librdf_model_add_string_literal_statement which remains
in the API as a wrapper around the above.
New methods were added to add, remove and list Statements to contexts.
/* Add a single statement to the model with context */
int librdf_model_context_add_statement(librdf_model* model,
librdf_node* context, librdf_statement* statement);
/* Add all statements on the stream to the model with context */
int librdf_model_context_add_statements(librdf_model* model,
librdf_node* context, librdf_stream* stream);
/* Remove a single statement from the model with the given context */
int librdf_model_context_remove_statement(librdf_model* model,
librdf_node* context, librdf_statement* statement);
/* Remove all statements from the model with given context */
int librdf_model_context_remove_statements(librdf_model* model,
librdf_node* context);
/* List all statements in the model with the given context */
librdf_stream* librdf_model_context_serialize(librdf_model* model,
librdf_node* context);
Node Class Changes
The Node class gained a new constructor:
librdf_node* librdf_new_node_from_typed_literal(
librdf_world *world,
const char *string,
const char *xml_language,
librdf_uri* datatype_uri);
that should be used in preference to
librdf_new_node_from_literal which remains in the API as
a wrapper around the above.
librdf_node_new_nodefrom_literal lost the
xml_space argument.
Parser Class Changes
The Parser class has an updated
Raptor parser
version 0.9.7 - parsers named raptor for RDF/XML
(MIME Type application/rdf+xml)
and ntriples for N-Triples. See the
Raptor NEWS
for the detailed changes since 0.9.5 in the last release of Redland.
The Java SiRPAC parsers sirpac-stanford,
sirpac-w3c and the W3C LibWWW parser libwww
have been removed. Raptor replaces them.
Serializer Class Changes
The Serializer class
gained a new method:
int librdf_serializer_serialize_model_to_file(
librdf_serializer* serializer,
const char *name,
librdf_uri* base_uri,
librdf_model* model);
This writes the serialized model to a filename
(rather than a file handle as done by
librdf_serializer_serialize_model)
which is easier to use from higher level language APIs until a
cross-language I/O abstraction is available.
The class gained an RDF/XML serializer that can be called
either by the name rdfxml or the
MIME Type application/rdf+xml. The
examples in each language have been updated to demonstrate
using this.
Storage Class Changes
The built in storage factories (memory,
hashes) now both support contexts. These are enabled by
adding the storage option contexts='yes' to the options
string. The hashes storage additionally supports indexing predicates
as an extra hash which can be enabled with index-predicates='yes'.
This makes triple queries of the form
(subject unknown, predicate known, object unknown) fast.
C Example of creating a 0.9.12 storage that does contexts using
on-disk Berkeley/Sleepycat DB:
storage=librdf_new_storage(world,
"hashes",
name,
"hash-type='bdb',write='yes',new='yes',contexts='yes'");
Stream Class Changes
The Stream
next method was split into get_object
always returning a pointer to a shared Statement and the
next method to advance the iterator.
The get_object method is generally called
current in the higher level language APIs.
(Iterator and Stream now have consistent method names; they
may be merged in future.)
C example for Redland 0.9.11 and earlier:
while(!librdf_stream_end(stream)) {
statement=(librdf_statement*)librdf_stream_next(stream);
/* do something with the statement */
}
Redland 0.9.12:
while(!librdf_stream_end(stream)) {
statement=(librdf_statement*)librdf_stream_get_object(stream);
/* do something with the statement */
librdf_stream_next(stream);
}
A new method get_context was added returning a Node
for the context of the Statement.
C Stream context example:
while(!librdf_stream_end(stream)) {
statement=(librdf_statement*)librdf_stream_get_object(stream);
context_node=(librdf_node*)librdf_stream_get_context(iterator);
/* do something with the statement and its context node */
librdf_stream_next(stream);
}
Perl Interface Changes
The Perl interface
moved to the RDF::Redland package (the RDF
namespace was deprecated in 0.9.11).
The interface was updated to receive Redland error and warning
messages callbacks. These can be set using the
RDF::Redland::set_error_handler
and RDF::Redland::set_warning_handler subroutines
taking a subroutine argument. They are called with a single
scalar argument which is the Redland message.
The interface was updated to match the Interator and Stream
changes described above.
Perl iterator example for Redland 0.9.11 and earlier:
while(!$iterator->end) {
my $node=$iterator->next;
# do something with the node
}
Redland 0.9.12:
while(!$iterator->end) {
my $node=$iterator->current;
# do something with the node
$iterator->next;
}
Perl stream example for Redland 0.9.11 and earlier:
while(!$stream->end) {
my $statement=$stream->next;
# do something with the statement
}
Redland 0.9.12:
while(!$stream->end) {
my $statement=$stream->current;
# do something with the statement
$stream->next;
}
Perl example of using the Serializer class:
# Use any rdf/xml parser that is available
my $serializer=new RDF::Redland::Serializer("rdfxml");
$serializer->serialize_model_to_file("output.rdf", $uri, $model);
Python API Changes
The Python interface
gained pydoc comments, along with an
HTML derived version.
The interface was updated to receive Redland error and warning
messages callbacks as python exceptions. These can be caught
using the standard python try {} catch {} blocks.
The interface was updated to match the Interator and Stream
changes described above.
Python iterator example for Redland 0.9.11 and earlier:
while not iterator.end():
node=iterator.next()
# do something with the node
Redland 0.9.12:
while not iterator.end():
node=iterator.current()
# do something with the node
iterator.next()
Python stream example for Redland 0.9.11 and earlier:
while not stream.end():
statement=stream.next()
# do something with the statement
Redland 0.9.12:
while not stream.end():
statement=stream.current()
# do something with the statement
stream.next()
Python example of using the Serializer class:
# Use any rdf/xml parser that is available
serializer=RDF.Serializer()
serializer.serialize_model_to_file("output.rdf", model)
Java Interface Changes
The Java classes were updated to have
a finished() method that the user can call to cleanup
objects replacing the rather useless Java finalize()
method that gives no guarantees to when it is called.
The Java API is still experimental and may change further.
Release notes for 0.9.11 and earlier are in the
NEWS page or
ChangeLog