Redland RDF Application Framework - To Do list

New bug and feature requests should be reported at the Redland Issue Tracker. This list is mostly historical.

Bugs, Features, Requests For Enhancement

In no real order. Help wanted.

'triples' query language implementation seems to have rotted as of 1.0.0
Docs: hard to find/know of raptor parser and serializer features
RFE: model class method to load into model with context
RFE: model class method to a remove/load into model for some context
Docs: it was hard to find the legal (parser) feature URIs in the documentation, or what values to use.
Documentation: describe the redland features available per class
Update the RDF/RDFS terms/labels for the latest RDF Schema revision
In-memory storage uses more memory than needed
Defensive code - check object types are correct by adding magic fields to start of all objects
Document the storage constructor hash options under the main class
RFE: make add_statement method on model take URIs
Add arcs_in/arcs_out model methods
Add container class for manipulating RDF containers - Seq, Bag, Alt, ...
If LGPL/Apache license, include compatible code OR add separate package?
Provide any necessary support for RDF Schema (fast rdf:type lookups)
Write some better example code
Write a quickstart / howto document
Maybe generate part of librdf.h automatically to make a 1-header file system.
Fix the URI hash mechanism - currently stores URI/pointer pairs in hash - wrong
The library does not do file locking on e.g. DB files
For IRIX (and others?), try also linking with -lpthreads when testing for working Berkeley DB
String encoding - needs iconv posix function and/or Unicode library. Some utility functions have been written for UTF-8 to/from ISO Latin-1.

(Excluding issues near FIXMEs in the source)

Done

The most recent changes are at the top, with the first release version where appropriate.

Write a C++ interface - redland++ under development and three other external ones are also available (1.0.9).
Consider an AVL-tree in-memory store (libavl) - done with 'trees' storage (1.0.8)
Initialise the default genid base in librdf_new_world with something more fine-grained than seconds. Add some microseconds digits (1.0.3)
RFE: consider how to add parser guessing (via raptor) into the redland parser API - use the 'guess' parser name (1.0.3)
RFE: add user mappings from URIs to namespace aliases to the serializer (Raptor 1.4.7)
RFE: The parser should be able to adjust error and warning actions, stopping immediately or ignoring them by features. Document the defaults. (1.0.2)
Add #ifdef WIN32 around parts including the config file and create a win32 build (1.0.1 - third party)
RFE: Add node.as_sha1 easy digest methods for literal nodes (1.0.1)
RFE: return serialized models to syntax as strings (1.0.0)
Removed the useless parser error/warning callbacks with a proper logging API that can be accessed by the application to and dealt with as desired. (0.9.17)
Allow raptor parser features (raptor_set_feature) to be set via librdf_parser_set_feature (0.9.17)
rdf_uri.h and rdf_node.h exported functions returning rdf_digest* which is a private structure, now made internal (0.9.16)
Added some defensive code checks for NULL object parameters and abort when debugging (0.9.15)
The serializer code crashed in PHP (0.9.15)
raptor-uri-test in redland when compiled against system raptor failed to find raptor.h; fixed in raptor 1.1.0.
Added librdf_model_get_storage method so storage object needn't be kept around when dealing with model (0.9.15).
Fixed serializing URI http://xmlns.com/foaf/0.1/mbox_sha1 to a property element to breaks the namespace/URI after 'sha1' as the maximal legal XML name (0.9.15)
Altered rdfproc to by default append to stores on parsing (0.9.15)
Docs: note in language docs that 'make install' of the main library is generally needed before any of the language builds will work (0.9.15)
BUG: Blank nodes with illegal nodeIDs (as far as RDF/XML are concerned) can be serialized. Fixed by the change below; all internal nodeIDs are now serializable in RDF/XML or N-Triples (0.9.14)
Rewrite incoming rdf:nodeID attribute values into same form as generated blank node IDs, ie. rredland-start-timerunique-sequence-counter (0.9.14)
example2.c is rather useless, replaced with a simple string parsing example (0.9.13)
Replaced *_seriali[sz]e methods with *_to_stream in order to reduce confusion versus serializer class (0.9.13)
BDB 4.1 causes problems - now supported (0.9.13)
perl - if you didn't clone a node before trying to re-use it the whole thing would segfault (0.9.13)
swig 1.3.14 may fail to work for some languages (perl,python ok) (0.9.13)
New method for syncing of models to store - librdf_model_sync (0.9.13)
Add proper error/warning reporting - world class has this now (0.9.13)
Ensure non-file: URIs do not cause crashes - fixed by Raptor URI class (0.9.13)
Add URI retrieval via external www libraries - done via raptor for retrieving URI content for parsing. Not by a new module. (0.9.13)
RFE: Add is_literal etc. methods to node class (0.9.13)
storage hash-type="memory" with no name (not really needed) crashed (0.9.13)
Initial thread support (0.9.13)
Repat removed - not up-to-date, maintained (0.9.13)
Crashed when contexts are attempted to be used when not enabled with contexts='yes' (0.9.13)
parse_as_stream methods assumed the URI object lived throughout the parsing. They may not, and had to be copied on the first call - fixed for raptor only (0.9.13)
make librdf_statement_set_subject* free any existing node before setting (0.9.12)
Pass RDF parser error messages to user code, scripting languages (0.9.12)
Add a swig test in autogen.sh, abort if it is too old (0.9.12)
swig 1.3.14 changed its default to shadow and failed to work out of the box. Add -noproxy to the swig line for 1.3.14+ (0.9.12)
Hacked around repat parser failing to pass bnode ids when as statement objects (0.9.12)
Added relative URI parsing code from Repat (via raptor) (0.9.12)
Handle xml:lang when the rdf parsers provide it (0.9.12)
Add query language interfaces (0.9.12)
Add support for blank nodes (were anonymous resources) (0.9.12)
Removed all node types except for resource, literal, blank, li (statement (0.9.12))
Use libxml rather than expat as default xml parser (at least for raptor) Reasons: uses iconv to handle more content encodings and provides better error reporting (0.9.12)
Link XML Libraries (-lexpat etc.) before libwww libraries to ensure correct XML library is used (0.9.11)
Python interface to parsers now works (0.9.11)
Multiple startup/shutdowns of entire library now works (0.9.11)
Make the library work as a shared library (0.9.11)
Model size method implemented (0.9.11)
Added a Java interface (0.9.10)
librdf_hash_from_string() - \ escaping doesn't work properly (0.9.9)
SiRPAC wrappers handle multi-line literal objects (0.9.9)
Added Tcl interface (0.9.8)
Add module - Raptor RDF parser - internal (0.9.8)
Make API docbook sgml generate 1 big HTML file (0.9.6)
Python interface (0.9.6)
Add proper error/warning reporting for RDF/XML parsers (0.9.6)
Perl interface (0.9.4)
module - storage implemented as hashes
made get methods in all hash impls. use iterators (side effecting - yuck)
made BDB V2 and V3 use btree and handle duplicates
removed GDBM from default code since doesn't handle duplicates
removed references to factories from simple user code
module - rdf parser - via libwww RDF parser
annotate the functions with gnome-doc / kernel-doc style comments (after JavaDoc) with care taken to make sure it is clear who owns the memory of objects passed in or returned by functions
module - rdf parser - via Java RDF API / SiRPAC
module - in memory storage
module - storage factory
module - model
module - memory allocation
change all functions to start with librdf_
a decent source-code based documentation system - GNOME / Linux kernel gnome-doc / kernel-doc programs based on the Javadoc formatted comments style, but for C
wrap the code with #ifdef etc. to prevent exposing internals to user code
module - berkeley DB hash
change functions to all start with RDF_ - new and free functions
decide on the license(s) - GPL/LGPL and BSDish ???
iterator interface and module for returning results

Decided to do differently, not to address

I reserve the right to change my mind again!

Docs: note that compiling redland may also build raptor and rasqal, if they are needed (1.0.9 no longer relevant)
RFE: model.parseString(format, "string"), model.parseURI(format, URI, baseURI=URI) creating and destroying temporary parser objects for format. Means no need to care about parser class if you just want simple things. Similarly for model.serialize() and Serializer class, however take care with serialise() method. Better as model.load(), model.save() or write()? Addressed by the guess parser in raptor 1.4.7+.
Make API docbook sgml generate decent PostScript / PDF output (1.0.3)
Split the header files into internal / user ones - now most headers are user ones (1.0.1)
Write stringbuffer for easy creation of growing in-memory strings - now being considered inside raptor.
Provide a way to represent reified statements
Redland does not work with GDBM hashes. BDB is much better (0.9.12)
Redland does not work with BerkeleyDB V1. Not fixing since it works with V2,V3 and V4 (0.9.12)
RFE: make Node and URI be the same class, change all methods using Nodes to take URI OR URI or Literal as appropriate. The RDF abstract syntax has nodes that are (URI-refs or Literals or ...) so this design matches it well (0.9.12)
These will not be made into modules at present, the internal RDF parser will use expat or libxml as necessary.
- Add module - xml dom parser - external
- Add module - xml dom parser - libxml
- Add module - xml sax parser - libxml
- Add module - xml parser - sax
- Add module - xml parser - dom
- Modularise the library so some parts can be removed [already done]
This is not needed as a module, the internal RDF parser will do this.
- module - xml namespaces