Raptor RDF Parser Toolkit - Release Notes

Raptor 1.4.2 Changes

Make raptor_xml_escape_string fail correctly when given bad UTF-8 to escape.

Raptor 1.4.1 Changes

Fixed a buffer overrun in decoding a URI scheme in raptor_uri constructors such as raptor_new_uri.

Fixed a crash in RSS enclosures crash when the url attribute seen on a non-<enclosure> element

raptor_xml_escape_string return value has changed to be an int, returning <0 on failure. This allows the empty string encoding an empty string case to work and be distinguished from an error.

Raptor 1.4.0 Changes

A release with the major new feature of providing serializing of RDF triples to syntaxes. It also added a new support class for I/O streams and had other minor fixes.

Added a Raptor Serializer class (raptor_serializer) with similar style to Parser (raptor_parser). Two serializers are provided, for RDF/XML and N-Triples. The serializing can be done to files, C FILE* or to strings. The raptor_iostream class that provides this also allows writing to any other form by creating a custom iostream.

The new raptor_serializer class functions added are: raptor_serializers_enumerate, raptor_serializer_syntax_name_check, raptor_new_serializer, raptor_free_serializer, raptor_serialize_start, raptor_serialize_start_to_filename, raptor_serialize_start_to_string, raptor_serialize_start_to_file_handle, raptor_serialize_statement, raptor_serialize_end, raptor_serializer_get_iostream, raptor_serializer_set_error_handler, raptor_serializer_set_warning_handler and raptor_serializer_get_locator

Added a Raptor I/O stream abstraction in raptor_iostream class to support serializing of RDF to multiple output streams such as to filenames, to C standard I/O FILE* handles and to strings especially for cross-language use. A raptor_iostream_handler can be used to construct a user-defined iostream.

The new raptor_iostream class functions added are: raptor_new_iostream_from_handler, raptor_new_iostream_to_sink, raptor_new_iostream_to_filename, raptor_new_iostream_to_file_handle, raptor_new_iostream_to_string, raptor_free_iostream, raptor_iostream_write_bytes, raptor_iostream_write_byte, raptor_iostream_write_end, raptor_iostream_write_string, raptor_iostream_write_counted_string, raptor_iostream_get_bytes_written_count, raptor_iostream_write_decimal and raptor_iostream_format_hexadecimal.

The rapper utility was modified to use serializer class so that the output formats supported are now N-Triples (-o ntriples) - the default, and RDF/XML (-o rdfxml).

Raptor now exports more static namespace URI strings for general application use: raptor_xml_namespace_uri, raptor_rdf_namespace_uri, raptor_rdf_schema_namespace_uri, raptor_xmlschema_datatypes_namespace_uri, raptor_owl_namespace_uri, and the length raptor_rdf_namespace_uri_len.

The raptor_stringbuffer class gained a new method raptor_stringbuffer_copy_to_string which allows efficient copy-out of a constructed string.

The raptor_www class gained a new method raptor_www_fetch_to_string to allow retrieving of web content as a single string.

RSS tag soup parser gained support for generating triples for enclosures, after a patch from Suzan Foster. Changes made include correcting the enclosures namespace and tidying some memory leaks.

Raptor 1.3.3 Changes

A release with major improvements along with several minor fixes.

Raptor's License was changed from LGPL 2.1/MPL 1.1 to LGPL 2.1/Apache 2

Thanks to Chris Pointon for several patches to make Raptor easier to build under Win32 which were applied, with some slight modifications.

Increased WWW content retrieval buffer size from 256 bytes to 4K since this was causing problems for even moderate size documents.

After testing raptor on a very large RDF/XML file with many rdf:ID values, the check for duplicate values was found to be inefficient in memory and slow. The implementation was improved to be more memory efficient and a new parser feature check_rdf_id was added to disable checking (default is enabled).

Added a new Unicode NFC checker to replace the functionality formally available by calling the GNOME glib function g_utf8_normalize. This new checker is done via several tables and adds approximately 50K to the object size of the library when compiled on x86. This code and tables can be disabled with configure option --disable-nfc-check causing all checks to succeed.

Fix the exporting of raptor_xml_literal_datatype_uri_string and raptor_xml_literal_datatype_uri_string_len as constants for use by applications. Previously raptor.h wasn't doing this correctly.

Added raptor_calloc_memory for allocating zeroed memory inside raptor, for use by applications passing memory in/out of raptor.

Added a new configure option --enable-parsers to allow the selection of the required RDF parsers from any of those supported (RDF/XML, Turtle, N-Triples, RSS tag soup).

Reorganised the sources to split parsing support from RDF/XML to support compiling without this parser.

Updated the RSS Tag Soup parser to start to handle the Atom 0.3 currently being standardised by the IETF Atom Publishing Format and Protocol working group.

Altered the Turtle parser to work with large source documents that exceeded bison limits. Thanks to Geoff Chappell for providing a fix for this.

Rewrote the URI parsing to create an internal structure and improved the relative URI resolving in preparation for future work such as potentially supporting URI canonicalisation such as proposed to be used by Atom.

Raptor 1.3.2 Changes

A release with some minor fixes.

Added a new configure option --with-expat-source=DIR to allow the use of external expat source trees in either the old or newer directory structure style. (Patch from Mark Smith).

Added raptor_alloc_memory for handlers that need to allocate memory in the same heap as raptor uses for raptor_free_memory. This is mostly useful for allocating memory that is freed by raptor in error, ID and statement handlers on win32 which has separate heaps for different DLLs.

A bug was fixed where errors which happened when fetching WWW content were always printed to stderr. They are now passed to the main error routines which allows applications to retrieve them.

Accessor functions were added for parts of the public raptor_locator structure which makes it possible to get structured error information from language bindings via Redland (Patch from Edd Dumbill). The new functions are:

The Unicode Normal Form C (NFC) checking via the GNOME glib library function g_utf8_normalize is broken, comparing the data it says is failed against other NFC checkers. It is also slower than need be since it is doing full normalizing rather than just checking for NFC, and adds a rather large dependency for just one function. A new portable checker will be added in a later release.

Raptor 1.3.1 Changes

A release primarily to fix some win32 and portability issues.

raptor.h now includes stdarg.h

Corrected the raptor_print_statement declaration in raptor.h for the argument statement to have one less 'const' which matches the actual code.

Made several portability fixes for compiling natively on win32 which doesn't quite do POSIX or C99.

Changed the support for file: URIs and converting to and from filenames. It now %-escapes spaces and % characters on conversion to and from filenames with raptor_uri_uri_string_to_filename, raptor_uri_uri_string_to_filename_fragment and raptor_uri_filename_to_uri_string. For Win32, more tests were added and the format of URIs supported corrected to use the file:///c: form rather than file://c|/

URIs that resolve to directories now return an error when lstat is available to check.

Parser Changes

The Turtle parser was updated to only allow language with non-datatyped literals, allow a '_' immediately after a ':' in qnames and to make a bare ':' qname work correctly.

The Turtle parser was fixed to re-initialise correctly when performing multiple parsings. The other parsers already did this correctly.

Added a warning to the RDF/XML parser for unknown rdf:parseType values, when parsing in lax mode - which is the default. It now tells the user when the parsing is working as 'Literal' mode by finding an unknown value. This is controlled by a new parser feature warn_other_parsetypes which is default set true in lax mode. Parser modes are controlled by the raptor_set_parser_strict method.

Raptor 1.3.0 Changes

A release primarily to provide support for the new Rasqal RDF query library but with some new features and fixes.

Parser Changes

Added a new constructor raptor_new_parser_for_content to guess the parser to use from hints in URIs or content, using a new utility function raptor_guess_parser_name.

Additional checks were added to the RDF/XML parser for RDF-namespaced names in element and attributes and if they are forbidden giving an error otherwise if unknown, giving a warning.

The Turtle parser was updated to correct the collections syntax, allow '-' in names and QNames and to add integer literals. This parser now correctly uses raptor_generate_id when a blank identifier name is needed.

Completed parser feature support by adding raptor_get_feature, raptor_feature_from_uri, and raptor_features_enumerate to get values and enable discovery of supported features at run time. raptor_set_feature was changed to give return a success value

Added a new method raptor_get_mime_type to get the MIME type of the syntax for a parser

raptor_parse_uri_with_connection (which is called by raptor_parse_uri) now sets the HTTP Accept: header to the MIME type of the parser in WWW requests using the new raptor_www_set_http_accept.

rapper changes

Added options -f/--feature for setting features and -g/--guess for guessing syntax from some content or identifiers. See rapper(1) for all rapper options.

Utility function changes

Added raptor_syntax_name_check to check for valid syntax language names.

Added raptor_free_memory to free memory returned by raptor functions.

Added Unicode utility functions raptor_unicode_char_to_utf8 and raptor_utf8_to_unicode_char.

Exported URI string raptor_xml_literal_datatype_uri_string.

Deprecated raptor_print_statement_detailed always intended to be internal.

WWW Class changes

Added support to set the HTTP Accept: header for curl and libxml2 when retrieving HTTP content by the new raptor_www_set_http_accept method.

New classes - Sequence and Stringbuffer

Added a utility class raptor_sequence providing simple sequences that can handle stacks and queues

Added a utility class raptor_stringbuffer for constructing strings from substrings appended or prepended.

Raptor 0.9.0 - Raptor 1.2.0 Changes

Release notes for 1.2.0 and earlier are in the NEWS page or ChangeLog


Copyright 2003-2004 Dave Beckett, Institute for Learning and Research Technology, University of Bristol