Marc Cromme [Mon, 7 May 2007 13:08:26 +0000 (13:08 +0000)]
remove now unnecessary ICU bug experimenting file, useful content moved into icu_I18N.c
Marc Cromme [Mon, 7 May 2007 12:52:04 +0000 (12:52 +0000)]
pretty-formatted all ICU code and removed dead code sections
Marc Cromme [Mon, 7 May 2007 12:18:34 +0000 (12:18 +0000)]
updated ICU casemap wrappers to use dynamic buffers, all ICU tests succeed
Marc Cromme [Mon, 7 May 2007 09:31:36 +0000 (09:31 +0000)]
moved working ICU sorting into YAZ unittest test_icu_I18N.c
commented casemapping out for the time beeing, need to integrate with new dynamic ICU buffers
Marc Cromme [Mon, 7 May 2007 08:42:45 +0000 (08:42 +0000)]
updatet error reporting to only report when strings are actually
sorted wrongly
Marc Cromme [Mon, 7 May 2007 08:15:34 +0000 (08:15 +0000)]
corrected error handeling in UErrorCode icu_utf16_from_utf8() to
mirror the error handeling in UErrorCode icu_utf16_from_utf8_cstr();
Marc Cromme [Mon, 7 May 2007 08:02:03 +0000 (08:02 +0000)]
unnecessary comments and print statements removed
Marc Cromme [Mon, 7 May 2007 07:58:31 +0000 (07:58 +0000)]
ICU sorting works correct now. Had forgotten to pass on the correct length of the destination buffer. Dynamic destination bugger resizing works as well.
Marc Cromme [Thu, 3 May 2007 11:53:12 +0000 (11:53 +0000)]
buffer stuff ok now, correct resizing
added printout of sort keys to see what get's wrong here ..
Marc Cromme [Thu, 3 May 2007 11:35:33 +0000 (11:35 +0000)]
changed error handeling, which had nasty side effects
Marc Cromme [Thu, 3 May 2007 09:36:33 +0000 (09:36 +0000)]
tweaking .. using dynamic allocated buffers. Now sorting fails again, but using static buffers as in icu_bug.c it works ..
Jakub Skoczen [Wed, 2 May 2007 19:32:13 +0000 (19:32 +0000)]
Minor changes to make it easier for for the server side scipt to init session.
Marc Cromme [Wed, 2 May 2007 14:03:03 +0000 (14:03 +0000)]
added ICU experiment which sorts correctly given all locales tried.
Marc Cromme [Wed, 2 May 2007 14:01:36 +0000 (14:01 +0000)]
tweaking, still no good results with danish sorting
Marc Cromme [Tue, 1 May 2007 13:27:32 +0000 (13:27 +0000)]
Added some more locales which fail. Something is very rotten in the kingdom of Denmark! Need to find out what wents wrong ...
Marc Cromme [Tue, 1 May 2007 13:16:09 +0000 (13:16 +0000)]
Added sorting test for ICU - only used in test_icu_I18N.c so far.
English and german sorting tests perform fine (including german special characters), but sorting of danish special characters fails. Very suspect. Needs more investigation! See test_icu_I18N_sortmap() in test_icu_I18N.c for details
Marc Cromme [Tue, 1 May 2007 08:17:05 +0000 (08:17 +0000)]
moved ICU helper function declarations from icu_I18N.h header file to icu_I18N.c source file
Marc Cromme [Tue, 1 May 2007 08:10:26 +0000 (08:10 +0000)]
cleaned ICU case folding/mapping tests
Adam Dickmeiss [Tue, 1 May 2007 07:58:43 +0000 (07:58 +0000)]
Fixed compilation of test test_icu_I18N (syntax error).
Sebastian Hammer [Tue, 1 May 2007 05:04:53 +0000 (05:04 +0000)]
Handle situation where IDF becomes 0 because all records contain a term (occurs
frequently when records result from a search).
This actually suggests that there may be a better technique than IDF for balancing
our TF, but I'll be darned if I know what it is.
Sebastian Hammer [Tue, 1 May 2007 05:02:54 +0000 (05:02 +0000)]
Handle records with null-value for string sortkey
Sebastian Hammer [Mon, 30 Apr 2007 14:29:48 +0000 (14:29 +0000)]
Added Paratext to demo
Sebastian Hammer [Mon, 30 Apr 2007 14:29:12 +0000 (14:29 +0000)]
Added new resources. Configuration changes
Sebastian Hammer [Mon, 30 Apr 2007 14:28:09 +0000 (14:28 +0000)]
Varous display changes to MK demo
Marc Cromme [Mon, 30 Apr 2007 13:56:52 +0000 (13:56 +0000)]
checked in test for ICU uppercase lowercase, title and foldcase char mapping
Sebastian Hammer [Mon, 30 Apr 2007 03:35:57 +0000 (03:35 +0000)]
Hardcoded motd in index.html. Cleared in onFormSubmit. This should be read from a file
or something, but what?
Marc Cromme [Fri, 27 Apr 2007 14:31:14 +0000 (14:31 +0000)]
added autoconf checks on ICU internationalization libs and established first test framework on ICU. Tests are still empty, to be added later.
Marc Cromme [Fri, 27 Apr 2007 12:17:04 +0000 (12:17 +0000)]
Factorized char normalization code out to normalize7bit.[hc] .
Removed unnecessary tests and assert
on service conf_metadata, as those are now
satisfied on service creation time (in config.c in function
conf_metadata_assign()) once under startup, and not every time a new
record is to be parsed.
Removed unneeded logging when service->merge == Metadata_merge_no .
Replaced manual record allocation with call to record constructor.
Replaced loop to find conf_metadata field_id with call to
conf_service_metadata_field_id().
Replaced manual record_metadata allocation with call to
record_metadata_create(se->nmem).
All these changes have been made with safety guards following this idom:
#if 0
//oldcode
#else
/newcode
#endif
such that they can be reverted easy if need might be. The #if guards will of
course disapear, when we have confirmed that the code behaves exaactly the
same as before.
Marc Cromme [Fri, 27 Apr 2007 11:56:41 +0000 (11:56 +0000)]
added struct record_metadata * record_metadata_create(NMEM nmem)
Marc Cromme [Fri, 27 Apr 2007 10:27:35 +0000 (10:27 +0000)]
in conf_metadata_assign() enforcing that merge_range is always
type_year _and_ enforcing that type_year is always range_merge, thus
setting contrains correctly at beginning of a server session setting
processing.
Removing these then unnecessary tests from ingest_record(), where
there will produce warnings logs nobody can do anything about at that
time, and where they are run for each ingested record.
Less code, less performed if statements, no unnecessary warning logs.
Marc Cromme [Fri, 27 Apr 2007 09:38:13 +0000 (09:38 +0000)]
enforcing that years type metadata is always range merged
Marc Cromme [Thu, 26 Apr 2007 21:33:32 +0000 (21:33 +0000)]
started facturizing 7 bit ascii normalizing functions out of logic.c into normalization7bit.[hc]
later to be replaced by ICU normalization
still many bits need to be identified and moved
Marc Cromme [Thu, 26 Apr 2007 21:31:05 +0000 (21:31 +0000)]
added setting sortkey on clusters, needs still nice wrapping in cluster interfaces
Marc Cromme [Thu, 26 Apr 2007 12:12:19 +0000 (12:12 +0000)]
using more of the record functions, adding record_metadata creator
Marc Cromme [Thu, 26 Apr 2007 11:41:26 +0000 (11:41 +0000)]
using record creator functions in production code in ingest_record()
Marc Cromme [Thu, 26 Apr 2007 11:03:54 +0000 (11:03 +0000)]
renamed local variables to more expressive names
Marc Cromme [Thu, 26 Apr 2007 10:19:05 +0000 (10:19 +0000)]
renaming position to field_id
Marc Cromme [Wed, 25 Apr 2007 13:57:49 +0000 (13:57 +0000)]
simplified to one title string sorting only, still crashes
Marc Cromme [Wed, 25 Apr 2007 13:28:55 +0000 (13:28 +0000)]
comed to the point where I can start testing sorting of records in the recordslist.
But there is a nasty segfault in reclist_sort(list, sort_parms) which needs to be sorted out before proceeding - prpbably a global parameter referenced someplace. just checking in to save the work so far, continuing tomorrow
Marc Cromme [Wed, 25 Apr 2007 13:16:13 +0000 (13:16 +0000)]
added test_reclists to CVS ignores
Marc Cromme [Wed, 25 Apr 2007 13:14:46 +0000 (13:14 +0000)]
forgotten to commit Makefile.am, which contains instructions for new test_reclists.c
Marc Cromme [Wed, 25 Apr 2007 13:09:17 +0000 (13:09 +0000)]
added interface for adding sort parameters for record lists, use like this
reclist_sortparms_insert(nmem, &sort_parms, service, "title", 1)
adde unit test of the same
Marc Cromme [Wed, 25 Apr 2007 09:23:03 +0000 (09:23 +0000)]
breaking too long source code lines, otherwise no changes
Marc Cromme [Wed, 25 Apr 2007 08:55:01 +0000 (08:55 +0000)]
broken too long source code lines, otherwise no changes
Marc Cromme [Wed, 25 Apr 2007 08:07:47 +0000 (08:07 +0000)]
function reclist_insert() does not change mergekey, so parameter signature changed to const char *merge_key
Marc Cromme [Wed, 25 Apr 2007 07:00:33 +0000 (07:00 +0000)]
removed unnecessary header includes
Marc Cromme [Tue, 24 Apr 2007 22:19:19 +0000 (22:19 +0000)]
protected from dereferencing of empty pointer
Marc Cromme [Tue, 24 Apr 2007 22:17:05 +0000 (22:17 +0000)]
changed prototype of
union data_types * data_types_assign(NMEM nmem,
union data_types * data1,
union data_types data2)
to
union data_types * data_types_assign(NMEM nmem,
union data_types ** data1,
union data_types data2)
to make sure that side effects are transmitted outside.
added tests for side effects.
added code for proper initialization of
record->metadata[i] and record->sortkeys[i] - valgrind my your friend!
Marc Cromme [Tue, 24 Apr 2007 13:50:07 +0000 (13:50 +0000)]
adding additional checks for correct side conditions on next metadata in record
Adam Dickmeiss [Tue, 24 Apr 2007 12:30:09 +0000 (12:30 +0000)]
Added edu.xml; based on edu.pz.
Adam Dickmeiss [Tue, 24 Apr 2007 08:03:03 +0000 (08:03 +0000)]
Functions connection_destroy, connection_release defined in connectin.h.
Adam Dickmeiss [Tue, 24 Apr 2007 07:04:36 +0000 (07:04 +0000)]
Annotate a little
Sebastian Hammer [Tue, 24 Apr 2007 04:37:58 +0000 (04:37 +0000)]
Added descriptive text about dynamic settings. Todo, create links from
places marked XXX
Adam Dickmeiss [Mon, 23 Apr 2007 21:05:23 +0000 (21:05 +0000)]
Refactor stuff from logic.c: connection stuff in connection.[ch],
client stuff to client.[ch]. host definitions in host.h. Global
parameters to parameters.h.,
Marc Cromme [Mon, 23 Apr 2007 12:33:00 +0000 (12:33 +0000)]
finalized helper utilities for record class, and added unit tests
Marc Cromme [Mon, 23 Apr 2007 09:06:20 +0000 (09:06 +0000)]
reverting to exit call, as it might be dangerous to have a dictionary which is not in order. Still, this place needs attention, as sending a wrong setting bt WS effectively kills the server.
Marc Cromme [Mon, 23 Apr 2007 08:56:52 +0000 (08:56 +0000)]
uupps - removed an exit() too fast without thought, reverting
Marc Cromme [Mon, 23 Apr 2007 08:48:50 +0000 (08:48 +0000)]
more records constructor work, no yet finished
Marc Cromme [Mon, 23 Apr 2007 08:15:22 +0000 (08:15 +0000)]
removed some of the most obvious exit() statements, which are now with the dynamic settings system a very easy way to kill the srver: just send a malformed settings file, and it's dead. Still many other exit() needs attention.
Adam Dickmeiss [Mon, 23 Apr 2007 08:06:21 +0000 (08:06 +0000)]
Allow any number of worker threads for sel_thread. Added optional parameter
work_destroy handler for sel_thread_create. This will only be called
sel_thread_destroy, in the case of non-empty work queues. Three resolver
threads now running - seems to make resolving many.xml slightly faster.
Adam Dickmeiss [Mon, 23 Apr 2007 07:55:23 +0000 (07:55 +0000)]
Saving seven keystrokes
Adam Dickmeiss [Mon, 23 Apr 2007 07:34:48 +0000 (07:34 +0000)]
Prefix resolved message
Adam Dickmeiss [Mon, 23 Apr 2007 07:29:34 +0000 (07:29 +0000)]
Rename test functions a bit
Adam Dickmeiss [Mon, 23 Apr 2007 07:26:15 +0000 (07:26 +0000)]
Added these target settings based on their .pz-equivalents.
Adam Dickmeiss [Mon, 23 Apr 2007 07:03:06 +0000 (07:03 +0000)]
Added ID for partintro
Adam Dickmeiss [Sun, 22 Apr 2007 16:41:42 +0000 (16:41 +0000)]
Connections now wait for hosts to be resolved. Delayed connect attempt
is handled by connect_resolver_host. All connection with bad addresses
(unresolved ones) are destroyed with connection_destroy. Connect
failures leave the connections in the open. If that's to be changed
connection_destroy should be called if connection_connect returns -1.
Sebastian Hammer [Sun, 22 Apr 2007 15:19:36 +0000 (15:19 +0000)]
Notes on implementation status of a couple of settings
Adam Dickmeiss [Sun, 22 Apr 2007 15:07:10 +0000 (15:07 +0000)]
Per-session databases and configuration. This was implemented
by Sebastian.
Adam Dickmeiss [Sun, 22 Apr 2007 15:00:56 +0000 (15:00 +0000)]
Removed load_simpletargets
Adam Dickmeiss [Sat, 21 Apr 2007 12:00:54 +0000 (12:00 +0000)]
Perform getaddrinfo in separate thread for a given struct host.
Sebastian Hammer [Fri, 20 Apr 2007 16:37:35 +0000 (16:37 +0000)]
Changed new code to use session_setting_oneval
Sebastian Hammer [Fri, 20 Apr 2007 16:21:19 +0000 (16:21 +0000)]
Made load of XSLT normalization stylesheet dynamic
Sebastian Hammer [Fri, 20 Apr 2007 15:36:48 +0000 (15:36 +0000)]
*** empty log message ***
Marc Cromme [Fri, 20 Apr 2007 14:37:17 +0000 (14:37 +0000)]
started record constructors. still unfnished, but yet only used in test_relevance.c
Sebastian Hammer [Fri, 20 Apr 2007 14:05:23 +0000 (14:05 +0000)]
Collapsing charset into native encoding. It simplifies implementation, and the
two are closely related anyway.
Jakub Skoczen [Fri, 20 Apr 2007 13:56:20 +0000 (13:56 +0000)]
problems with ajaxError catching all exceptions solved
pz2.js will now throw an exception if the protocol version is not supported
smarter error handling in general
Marc Cromme [Fri, 20 Apr 2007 13:03:22 +0000 (13:03 +0000)]
splitting record specific definition out from pazpar2.h header to record.h header, thus preparing for implementation of record constructors in the coming record.c file
Adam Dickmeiss [Fri, 20 Apr 2007 11:56:37 +0000 (11:56 +0000)]
Minor
Adam Dickmeiss [Fri, 20 Apr 2007 11:44:58 +0000 (11:44 +0000)]
Select thread system now passes a simple test using IOCHANSs.
Adam Dickmeiss [Fri, 20 Apr 2007 11:43:43 +0000 (11:43 +0000)]
Fix hard-coded 30 second limit. Unfortunately we have a lot of
these select loops around - all more or less buggy.
Adam Dickmeiss [Fri, 20 Apr 2007 11:20:16 +0000 (11:20 +0000)]
Linking with threads
Marc Cromme [Fri, 20 Apr 2007 11:00:29 +0000 (11:00 +0000)]
added metadata and sortkey assign functions (to conf_service objects)
in parse_service, seperated code in three groups: xml parsing, parsed value logic, and assignment to conf_service, using the new creator and assign functions.
Adam Dickmeiss [Fri, 20 Apr 2007 10:15:19 +0000 (10:15 +0000)]
sel_thread_result reads from pipe
Adam Dickmeiss [Fri, 20 Apr 2007 10:06:52 +0000 (10:06 +0000)]
Started work on select thread system.
Sebastian Hammer [Fri, 20 Apr 2007 04:32:33 +0000 (04:32 +0000)]
Allows definition/override of pz:cclmap:* settings via settings ws.
CCL parser allocated/deallocated on the fly at query time
Sebastian Hammer [Fri, 20 Apr 2007 04:08:14 +0000 (04:08 +0000)]
Removed query from session structure. It is no longer needed, since a pquery
is associated with each client.
Marc Cromme [Thu, 19 Apr 2007 19:44:31 +0000 (19:44 +0000)]
added ignore of session*.xml files
Marc Cromme [Thu, 19 Apr 2007 19:42:30 +0000 (19:42 +0000)]
added conf_service creator inclusive metadata and sortkey assigners, added unittests on the same as well. creators for record construction still needed
Marc Cromme [Thu, 19 Apr 2007 19:40:22 +0000 (19:40 +0000)]
restricting to non-test files
Adam Dickmeiss [Thu, 19 Apr 2007 16:07:20 +0000 (16:07 +0000)]
Refactor access to global variable channel_list. There were a few
places where this was accessed and always in the same way.
Adam Dickmeiss [Thu, 19 Apr 2007 16:06:10 +0000 (16:06 +0000)]
Tiny example update
Adam Dickmeiss [Thu, 19 Apr 2007 15:31:23 +0000 (15:31 +0000)]
Removed redundant declarations
Marc Cromme [Thu, 19 Apr 2007 11:57:53 +0000 (11:57 +0000)]
started adding conf_service and conf-metadata constructors. not finished yet, just checked in for file moving. Do not use yet
Adam Dickmeiss [Wed, 18 Apr 2007 19:50:06 +0000 (19:50 +0000)]
Ignore test_relevance program and log file
Sebastian Hammer [Wed, 18 Apr 2007 19:45:09 +0000 (19:45 +0000)]
Fixed bug 854. Using nmem_split() to split phrase terms on word boundaries.
I love nmem_split. If I were not a family man, I would ask her to marry me.
Sebastian Hammer [Wed, 18 Apr 2007 16:11:41 +0000 (16:11 +0000)]
Removed 'Masterkey' from implementation name
Marc Cromme [Wed, 18 Apr 2007 15:09:51 +0000 (15:09 +0000)]
pretty-formattting comments , now not exeeding 80 chars
Adam Dickmeiss [Wed, 18 Apr 2007 13:41:03 +0000 (13:41 +0000)]
Make Pazpar2 configure prefer static YAZ - when locally built
Sebastian Hammer [Wed, 18 Apr 2007 12:41:38 +0000 (12:41 +0000)]
I think this addresses Bug 1070.
Marc Cromme [Wed, 18 Apr 2007 07:41:03 +0000 (07:41 +0000)]
encapsulating all record fields in a xsl:for-each to prevent emitting empty
<md-title> elements. Now all emitted elements have always content, and the final web client JS does not break any more.