This is an exercise for the ICU Workshop (September 2000).
Day 2: September 12th 2000
Pre-requisite:
1. All the hardware and software requirements from Day 1.
2. Attended or fully understand Day 1 material.
3. Read through the ICU user's guide at
http://oss.software.ibm.com/icu/userguide/.
#Transformation Support
10:45am - 12:00pm
Alan Liu
Topics:
1. What is the Unicode normalization?
2. What kind of case mapping support is available in ICU?
3. What is Transliteration and how do I use a Transliterator on a document?
4. How do I add my own Transliterator?
INSTRUCTIONS
------------
This exercise was developed and tested on ICU release 1.6.0, Win32,
Microsoft Visual C++ 6.0. It should work on other ICU releases and
other platforms as well.
To install: Create a folder "translit" at:
<icu>/source/samples/translit
Within it, place the files:
translit.dsp
translit.dsw
main.cpp
util.cpp
util.h
Open the file "translit.dsw" in Microsoft Visual C++.
PROBLEMS
--------
Problem 0:
To start with, the program prints out a series of dates formatted in
Greek. Set up the program, build it, and run it.
Problem 1: Basic Transliterator (Easy)
The Greek text shows up almost entirely as Unicode escapes. These
are unreadable on a US machine. Use an existing system
transliterator to transliterate the Greek text to Latin so it can be
phonetically read on a US machine. If you don't know the names of
the system transliterators, use Transliterator::getAvailableID() and
Transliterator::countAvailableIDs(), or look directly in the index
table icu/data/translit_index.txt.
Problem 2: RuleBasedTransliterator (Medium)
Some of the text is still unreadable and shows up as Unicode escape
sequences. Create a RuleBasedTransliterator to change the
unreadable characters to close ASCII equivalents. For example, the
rule "\u00C0 > A;" will change an 'A' with a grave accent to a plain
'A'.
To save typing, use UnicodeSets to handle ranges of characters.
See the included file "U0080.pdf" for a table of the U+00C0 to U+00FF
Unicode block.
Problem 3: Transliterator subclassing; Normalizer (Difficult)
The rule-based approach is flexible and, in most cases, the best
choice for creating a new transliterator. Sometimes, however, a
more elegant algorithmic solution is available. Instead of typing
in a list of rules, you can write C++ code to accomplish the desired
transliteration.
Use a Normalizer to remove accents from characters. You will need
to convert each character to a sequence of base and combining
characters by applying a canonical denormalization transformation.
Then discard the combining characters (the accents etc.) leaving the
base character. Wrap this all up in a subclass of the
Transliterator class that overrides the pure virtual
handleTransliterate() method.
ANSWERS
-------
The exercise includes answers. These are in the "answers" directory,
and are numbered 1, 2, etc. In some cases new files that the user
needs to create are included in the answers directory.
If you get stuck and you want to move to the next step, copy the
answers file into the main directory in order to proceed. E.g.,
"main_1.cpp" contains the original "main.cpp" file. "main_2.cpp"
contains the "main.cpp" file after problem 1. Etc.
Have fun!
