TODO - external/liblouis-github - Git at Google

 This file describes the TODO items for the liblouis project.  	-*- org -*-

 When a task is done and accepted, consider moving it into the NEWS
 file, with a bit of extra info.

 * 2.6
 ** Investigate another crash/patch reported by brailcom: http://www.freelists.org/post/liblouis-liblouisxml/Another-crash-in-Python-hyphenation
 ** Document the changes to LOUIS_TABLEPATH
 ** Extend and document the scripting language
 http://www.freelists.org/post/liblouis-liblouisxml/Very-preliminary-documentation-of-scripting-language

 ** configure.ac does not check for nose now.
 Instead it is done with a try except in the python script. Need to be
 added to configure.ac so that debian packages can automatically pick
 up dependencies.

 ** Mapping between table filenames, human readable names, and translatable strings, needs to be accessible from bindings.
 Will remove the need of redefining translatable names for the tables
 in orca, nvda, blaster etc.

 * near term
 ** (google issue 9) bindings should provide variable to pick up table location at runtime.

 ** Fix the problem that LOUIS_TABLEPATH always looks in the standard PATH
 even if that was not in the environment var

 ** Add more test harness data for languages.

 ** [mh] compBrlLeftCursor is wrong, provide test case for correct behaviour
 Once corrected, consider merging compBrlLeftCursor and
 compBrlAtCursor, since neither provide the needed behaviour, but the
 corrected version should. also see
 https://bugzilla.gnome.org/show_bug.cgi?id=592421

 ** fix bug described by squash_space.c

 ** Fix cursor position problems when capsigns are used.
 Originally reported at:
 https://bugzilla.gnome.org/show_bug.cgi?id=651217 testcase added in
 en-GB-g2_harness.txt, search for 651217. The bug can be seen with
 other tables too.

 ** Esperanto table should not be blacklisted, work out whats wrong and make sure it is usable.

 ** [mh] According to the harness, Danish table isn't producing correct output.
 From memory of Swedish, the actual output is correct, but the harness
 might be wrong, best to check with Danish users.

 * unallocated
 ** (google issue 16) infinite loop in lou_backtranslate.

 ** (google issue 6) italword opcode not documented

 ** (google issue 4) problem with contraction cursor position and compBrlAtCursor.

 ** Add java bindings
 It would be nice to have some canonical java bindings. There are
 several potential candidates:
 - Bindings by Michael Whapples
 - Minimal jna bindings by SBS
 - port jni bindings from utdml
 - new jna bindings by Bert Frees

 ** Enhance the API to handle pre-hyphenated text
 This basically just means to port the code which is in the java
 bindings to C so that it can be used from other bindings

 ** Add readline support to all the tools

 ** Use portable malloc from gnulib
 to get rid of the windows #ifdefs

 ** Enhance translation table compiler to issue warnings
 [jb]: It should be an error to define the same single-cell dot pattern
 for two different characters. I am considering issuing an error
 message and rejecting the table if this happens.

 [mh]: It would also be very helpful if we could issue a warning when a
 character has been defined as two or more braille representations.
 Could we have these as warnings, not errors please.

 ** followup to above enhancement, either at the terminal or when called by
 bindings, we should be able to give more useful feedback, i.e. could
 not translate because table not found, or table found but has errors,
 or characters undefined, etc. also see:
 http://www.nvda-project.org/ticket/2448

 ** Optimize for use with large tables
 When used with dictionary based tables liblouis is very slow. The
 issue is probably that the hash key is not very well suited for this
 use case and there will be tons of collisions, making the lookup
 essentially linear.

 There was a discussion about this on the mailing list
 (http://www.freelists.org/post/liblouis-liblouisxml/Improved-hash-function-for-tables).

 ** apply the the patch by Igor B. Poretsky

 ** apply the jptest_patch
	This file describes the TODO items for the liblouis project. -- org --

	When a task is done and accepted, consider moving it into the NEWS
	file, with a bit of extra info.

	* 2.6
	** Investigate another crash/patch reported by brailcom: http://www.freelists.org/post/liblouis-liblouisxml/Another-crash-in-Python-hyphenation
	** Document the changes to LOUIS_TABLEPATH
	** Extend and document the scripting language
	http://www.freelists.org/post/liblouis-liblouisxml/Very-preliminary-documentation-of-scripting-language

	** configure.ac does not check for nose now.
	Instead it is done with a try except in the python script. Need to be
	added to configure.ac so that debian packages can automatically pick
	up dependencies.

	** Mapping between table filenames, human readable names, and translatable strings, needs to be accessible from bindings.
	Will remove the need of redefining translatable names for the tables
	in orca, nvda, blaster etc.

	* near term
	** (google issue 9) bindings should provide variable to pick up table location at runtime.

	** Fix the problem that LOUIS_TABLEPATH always looks in the standard PATH
	even if that was not in the environment var

	** Add more test harness data for languages.

	** [mh] compBrlLeftCursor is wrong, provide test case for correct behaviour
	Once corrected, consider merging compBrlLeftCursor and
	compBrlAtCursor, since neither provide the needed behaviour, but the
	corrected version should. also see
	https://bugzilla.gnome.org/show_bug.cgi?id=592421

	** fix bug described by squash_space.c

	** Fix cursor position problems when capsigns are used.
	Originally reported at:
	https://bugzilla.gnome.org/show_bug.cgi?id=651217 testcase added in
	en-GB-g2_harness.txt, search for 651217. The bug can be seen with
	other tables too.

	** Esperanto table should not be blacklisted, work out whats wrong and make sure it is usable.

	** [mh] According to the harness, Danish table isn't producing correct output.
	From memory of Swedish, the actual output is correct, but the harness
	might be wrong, best to check with Danish users.

	* unallocated
	** (google issue 16) infinite loop in lou_backtranslate.

	** (google issue 6) italword opcode not documented

	** (google issue 4) problem with contraction cursor position and compBrlAtCursor.

	** Add java bindings
	It would be nice to have some canonical java bindings. There are
	several potential candidates:
	- Bindings by Michael Whapples
	- Minimal jna bindings by SBS
	- port jni bindings from utdml
	- new jna bindings by Bert Frees

	** Enhance the API to handle pre-hyphenated text
	This basically just means to port the code which is in the java
	bindings to C so that it can be used from other bindings

	** Add readline support to all the tools

	** Use portable malloc from gnulib
	to get rid of the windows #ifdefs

	** Enhance translation table compiler to issue warnings
	[jb]: It should be an error to define the same single-cell dot pattern
	for two different characters. I am considering issuing an error
	message and rejecting the table if this happens.

	[mh]: It would also be very helpful if we could issue a warning when a
	character has been defined as two or more braille representations.
	Could we have these as warnings, not errors please.

	** followup to above enhancement, either at the terminal or when called by
	bindings, we should be able to give more useful feedback, i.e. could
	not translate because table not found, or table found but has errors,
	or characters undefined, etc. also see:
	http://www.nvda-project.org/ticket/2448

	** Optimize for use with large tables
	When used with dictionary based tables liblouis is very slow. The
	issue is probably that the hash key is not very well suited for this
	use case and there will be tons of collisions, making the lookup
	essentially linear.

	There was a discussion about this on the mailing list
	(http://www.freelists.org/post/liblouis-liblouisxml/Improved-hash-function-for-tables).

	** apply the the patch by Igor B. Poretsky

	** apply the jptest_patch