AuroraMiddleware/scuffed-code
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

ICU-1465 note on japanese collation levels and performance

Browse Source 
X-SVN-Rev: 7214
This commit is contained in:
Markus Scherer 2001-11-30 18:11:35 +00:00
parent 1b95ca6012
commit 1121202bea
1 changed files with 23 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
icu4c/readme.html
View File

@ -301,7 +301,29 @@
<h3>Collation improvements</h3>
<p>The performance of Japanese Katakana collation is improved, and the
Japanese collation is changed for conformance with the JIS X 4061 standard.</p>
Japanese collation is changed for conformance with the JIS X 4061 standard.
The improvement is in the handling of the length and iteration marks, making
the processing of regular letters faster.</p>
<p>The JIS X 4061 standard specifies a 5-level sorting algorithm. Sorting
with all five levels according to JIS is
achieved in ICU 2.0 with the &quot;identical&quot; strength. The fifth level
distinguishes regular character codes from compatibility variants.</p>
<p>There is special code to handle the fourth (quarternary) level of the JIS
standard, which distinguishes between Hiragana and Katakana letters. In ICU
2.0 string comparisons (like ucol_strcoll), when using the
&quot;shifted&quot; option, this is slow because it
generates complete sort keys for both strings. This is not an issue if the
&quot;shifted&quot; option is not used, or if the string comparison is done
with fewer levels.</p>
<p>
Quarternary strength, without the &quot;shifted&quot; option, is the default for Japanese collation in ICU 2.0.</p>
<p>Three-level sorting (tertiary strength) and lower &mdash; if sufficient &mdash; is
faster even with &quot;shifted&quot; on (for string comparisons: <em>much</em>
faster in this case).</p>
<h3>License Change (for ICU 1.8.1 and up)</h3>