We are about to release Xaira version 1.20, and as usual we are making
available a 'candidate' release in advance. The files are at
http://www.oucs.ox.ac.uk/rts/xaira/Download/
Most of the changes in this release are bug fixes and performance
improvements connected with the (still unfinished) conversion of BNC to xml.
A brief list follows:
1. Introduced API version control to make sure the client and remote server
are running the same version of the API.
2. Fixed Windows executable to remove dependence on getaddrinfo, which is
not in Windows 2000.
3. Fixed bug that prevented sorting on non-w-tagged words in w-tagged
corpora (eg in headers).
4. Changed default log path to /var/log/xaira.
5. Improved efficiency of sort in client.
6. Show correct attribute frequencies in attribute list when a class of a
partition is active.
7. Improved efficiency of scrolling, fixed bug with scrolling to end of long
list of hits.
8. Fixed bug where words in multiword hit were being counted as collocates.
9. Fixed display of columns in collocate window when a lemma scheme or
colligation is used.
10. Fixed sort in xml display.
11. Fixed bug when collocation window was an xml element.
12. Fixed 32-bit rounding error when thinning to a random set.
13. Brought java source up to date.
14. Added simple checks for Xerces and ICU headers in configuration.
Known remaining problems:
1. Not next links in query builder are broken.
2. Words break incorrectly after letter p in page mode. This is an ICU bug.
We have tested Xaira with the forthcoming ICU 3.6 release, which also adds
Unicode 5 support, and the p bug is indeed fixed, but in the alpha there is
a new bug that occasionally loses the final line of a paragraph in page
mode. So 1.20 will continue to use ICU 3.4.
available a 'candidate' release in advance. The files are at
http://www.oucs.ox.ac.uk/rts/xaira/Download/
Most of the changes in this release are bug fixes and performance
improvements connected with the (still unfinished) conversion of BNC to xml.
A brief list follows:
1. Introduced API version control to make sure the client and remote server
are running the same version of the API.
2. Fixed Windows executable to remove dependence on getaddrinfo, which is
not in Windows 2000.
3. Fixed bug that prevented sorting on non-w-tagged words in w-tagged
corpora (eg in headers).
4. Changed default log path to /var/log/xaira.
5. Improved efficiency of sort in client.
6. Show correct attribute frequencies in attribute list when a class of a
partition is active.
7. Improved efficiency of scrolling, fixed bug with scrolling to end of long
list of hits.
8. Fixed bug where words in multiword hit were being counted as collocates.
9. Fixed display of columns in collocate window when a lemma scheme or
colligation is used.
10. Fixed sort in xml display.
11. Fixed bug when collocation window was an xml element.
12. Fixed 32-bit rounding error when thinning to a random set.
13. Brought java source up to date.
14. Added simple checks for Xerces and ICU headers in configuration.
Known remaining problems:
1. Not next links in query builder are broken.
2. Words break incorrectly after letter p in page mode. This is an ICU bug.
We have tested Xaira with the forthcoming ICU 3.6 release, which also adds
Unicode 5 support, and the p bug is indeed fixed, but in the alpha there is
a new bug that occasionally loses the final line of a paragraph in page
mode. So 1.20 will continue to use ICU 3.4.