The opcodes letsign and begcomp/endcomp are stopgaps to the larger issue in
that LibLouis' concept of mode is inadequate. Just as emphasis, mode is
something that should be defined in the tables and not hardcoded several
different ways. LibLouis should do this in one consistent way.
One problem that we keep running into here is the inability to switch
modes/grades in the middle of a translation. While this may not be an issue
for screenreaders, it is an issue more complex translating we here at APH
require.
I have dealt with similar issues when writing compilers and such, and the
solutions were not pretty but doable. I have ideas on upgrading LibLouis and
will organize and post them.
MRG
mgray@xxxxxxx
________________________________________
From: liblouis-liblouisxml-bounce@xxxxxxxxxxxxx
[liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] on behalf of Davy Kager
[DavyKager@xxxxxxxxxx]
Sent: Monday, February 22, 2016 3:58 AM
To: 'Christian Egli'; liblouis-liblouisxml
Subject: [liblouis-liblouisxml] Re: Analysis of regressions in ueb_update branch
Hi Christian and all,
@egli: very clear, thanks much for the info!
In particular it seems we have two broken opcodes:
1. letsign, when used in the context of mixing letters and numbers, e.g.
blah.t2t. I believe Mike added a new opcode for this, along with numericmode,
but since I reverted numericmode we may need to revert some other stuff.
2. begcomp/endcomp. Specifically, if the text between begcomp and endcomp
contains capitals, then after endcomp Mike's code inserts end indicators for
these capitals.I just noticed that begcomp/endcomp isn't in the user docs, but
I believe the idea is to suppress all indicators for text that contains some
text defined by compbrl.
There is also a problem in Finnish:
The test specifies a phrase of words but doesn't emphasize the closing period.
So Mike's code adds endemphword to indicate that the period is not emphasized.
Arguably this is correct, but the test expects this to be ignored. For now I
marked them as xfails. I think there are two solutions:
1. Indicate the period as emphasized.
2. Add functionality similar to capsmodechars that says "don't consider
non-letters at the end of a word when determining emphasis."
Davy
-----Oorspronkelijk bericht-----
Van: Christian Egli [mailto:christian.egli@xxxxxx]
Verzonden: vrijdag 19 februari 2016 12:42
Aan: liblouis-liblouisxml <liblouis-liblouisxml@xxxxxxxxxxxxx>; Davy Kager
<DavyKager@xxxxxxxxxx>
Onderwerp: Analysis of regressions in ueb_update branch
Hi all, hi Davy
I did some work on the regressions in the ueb_update branch. I cleaned the
error output and unchecked the tests that were marked as expected failures
previously but now actually work (150).
There are now 15 failing (yaml) test files with 248 failing test cases.
I think that these can be classified into a couple of main classes of
issues:
- Problems with in-word capitalization
ar-ar-g1_harness has a test for 'MPower' which should apparently be
translated as '⠠⠠⠍⠏⠕⠺⠑⠗' where as it comes out as '⠠⠠⠍⠏⠠⠄⠕⠺⠑⠗'. It
appears that the end of the capitalized letters is indicated in the
received output. Maybe the expected string is actually wrong.
Similarly in ta-ta-g1_harness where 'GHz,' is expected as '⠠⠠⠛⠓⠵⠂' but
actually comes out as '⠠⠠⠛⠓⠠⠄⠵⠂'.
In en-ueb-g2-dictionary_harness.txt.bak there is a test for 'Ac' which
is expected as '⠰⠠⠁⠉' but actually comes out as '⠠⠁⠉'. I don't speak
ueb but the actual result seems more plausible to me.
- Problems with symbols in ueb
en-ueb-symbols_harness has many tests for single symbols. At the
moment I fail to see a common pattern there. For example
'"' translates to '⠐' instead of '⠠⠶'.
'`' translates to '⠈' instead of '⠨⠡'. and
'│' (BOX DRAWINGS LIGHT VERTICAL) translates to '⠄⡳⠭⠆⠢⠴⠆⠄' instead of '⠸'
- Mixed numbers and letters
da-dk-g26-dictionary_harness for example has a test for "1kg" and
expects "⠼⠁⠠⠅⠛" as a translation. However the resulting translation is
"⠼⠁⠅⠛". It appears that the indicator to indicate the end of the
number is missing.
- Capitalization and backtranslation
In one test It appears that some of backtranslation of capitalized
braille is broken. Single letter capitals are translated correctly but
stuff like '⠨⠨⠍⠧⠹⠕⠱' is not translated back to 'MVGYOSZ'. See
tests/yaml/hu-hu-g1_backward.yaml:1179
- Emphasis
fi_harness has failures in emphasis. It appears that the start
indicator is doubled and an end indicator is added for single word
emphasis where it shouldn't be (yaml/fi_harness.yaml:74). For
multiword emphasis the closing indicator seems to be added at the and
instead of at the beginning. Maybe just a problem in the table?
- Single word capitalization
no_harness seems to have a problem with a single letter capital: 'I'
is translated as '⠠⠄⠠⠊⠠⠄' instead of as '⠠⠊'. This is right after an
emphasis word
Hope that helps
Christian
--
Christian Egli
Swiss Library for the Blind, Visually Impaired and Print Disabled Grubenstrasse
12, CH-8045 Zürich, Switzerland
-----
Jetzt kostenlos bestellen: Eidgenoessische und kantonale Abstimmungsunterlagen
fuer den 28. Februar zum Hoeren auf CD. Bestellungen an medienverlag@xxxxxx
DISCLAIMER:
De informatie verzonden met dit e-mail bericht is uitsluitend bestemd voor de
geadresseerde. Indien u niet de beoogde geadresseerde bent, verzoeken wij u
vriendelijk dit aan de afzender te melden (of via:
info@xxxxxxxxxx<mailto:info@xxxxxxxxxx>) en het origineel en eventuele kopieën
te verwijderen.
The information sent in this e-mail is solely intended for the individual or
company to whom it is addressed. If you received this message in error, please
notify the sender immediately (or mail to
info@xxxxxxxxxx<mailto:info@xxxxxxxxxx>) and delete the original message and
possible copies.
?��u�?�*m���~�^������h�yhiحjwe�y,��k�7����z�(��m����&�謢�
��u� �*m���~�^������h�yhiحjwe�y,��k�7����z�(��m����&��謢�
