Jean-Francois Dockes
c6e228b7c6
Prepared windows port by removing a number of spurious reference to unix-specific interfaces, and using some xapian posix adaptor includes
2015-08-19 14:41:10 +02:00
Jean-Francois Dockes
e6c0ca403d
recollindex: do not retry files which previously failed to be indexed, except if they were changed since, or option -k is set
2015-04-23 10:37:37 +02:00
Jean-Francois Dockes
cd892ee1a7
usemtime config variable needs to be global
2014-12-09 15:58:23 +01:00
Jean-Francois Dockes
d630cbbaec
Delete RCL_USE_XATTR configure/compile time variable, it was not
...
useful. Add configuration variable to use mtime instead of ctime for update
detection. Useful on a system where xattrs would be modified but not
indexed, to avoid excessive reindexing.
2014-12-09 11:15:17 +01:00
Jean-Francois Dockes
9346c4db9c
fsindexer: avoid generating an empty bogus document when getting a subdoc error after the base doc has been seen
2014-10-29 08:18:20 +01:00
Jean-Francois Dockes
bba7875a10
Defining a topdir under a skippedPath is allowed and usefully works with normal indexing. This change fixes a bug which prevented it from working with the real-time monitor and recollindex -i
2014-10-10 10:17:57 +02:00
Jean-Francois Dockes
368ebf2f86
All docs get containerfilename field, indexed exclusively with prefix (only a field search containerfilename:someterm can match).
2014-07-23 19:30:41 +02:00
Jean-Francois Dockes
088f1f9a9f
try out modification to use container file name for display, but not search, purposes when the subdocument file name is not set
2014-07-16 17:05:26 +02:00
Jean-Francois Dockes
cc052a31f8
don't add the top container file name to the subdocs terms
2014-06-12 14:17:18 +02:00
Jean-Francois Dockes
e867f855ad
get rid of numerous probably inocuous valgrind/helgrind messages by ensuring that actual string copies are passed between threads, without refcount/shared data magic
2014-05-05 19:01:58 +02:00
Jean-Francois Dockes
56a56500c1
Handle partial indexing of document restricted to metadata from extended attributes
2013-10-04 10:57:11 +02:00
Jean-Francois Dockes
f49264a3f5
message
2013-09-27 13:04:23 +02:00
Jean-Francois Dockes
9cb5371820
get rid of the OFFTPC nonsense and cast to long long use lld to print off_t
2013-09-26 14:35:50 +02:00
medoc
641acd3d68
move the execution of external metadata-gathering commands from fsindexer to internfile for consistency of handling with filter-generated metadata
2013-09-06 11:51:00 +02:00
Jean-Francois Dockes
2b80c77c23
Add possibility to display a list of sub-documents for a given result
2013-04-24 16:33:53 +02:00
Jean-Francois Dockes
906e58feff
added code to purge obsolete messages when a compound document (esp. mbox) is shortened and a partial update is performed (no general purge). Else the orphan docs remained in the index potentially forever (needed actual reindex of the file by a full pass to go away)
2013-04-22 11:32:49 +02:00
Jean-Francois Dockes
f05cae7344
arrange so that a default first indexing run for a given user runs a quick shallow pass, so that queries return some results quickly after indexing starts, avoiding user frustration
2013-04-18 14:29:32 +02:00
Jean-Francois Dockes
a74d885309
use st_ctime always to detect file changes
2013-03-31 08:36:11 +02:00
Jean-Francois Dockes
3736c00a3b
more small fnmatch() fixes
2013-03-27 16:56:54 +01:00
Jean-Francois Dockes
3c80e51940
simplified temp file handling for compressed documents and, for querying, implemented caching for last file uncompressed
2013-03-06 18:52:57 +01:00
Jean-Francois Dockes
d3631b5ddf
cleaned up processing of metadata from diverse origins (doc,extattrs,localfields)
2013-01-29 14:33:57 +01:00
Jean-Francois Dockes
e8afc66d8f
fix compile issue with idxthreads enabled
2013-01-28 09:45:33 +01:00
Jean-Francois Dockes
4272939c3d
add possibility to import tags from external app, e.g. tmsu
2013-01-22 18:30:18 +01:00
Jean-Francois Dockes
7ee94dbf3e
traces
2012-12-28 16:31:10 +01:00
Jean-Francois Dockes
a0e273ddd4
Drain the job pipeline in fsindexer::purgefiles like was done in fsindexer::indexfiles as there is no warranty that the latter will be called before a db::close for each iteration of the monitor loop
2012-12-15 09:21:50 +01:00
Jean-Francois Dockes
72927b3fc8
additional cleanup after switch from .beagle to .recollweb
2012-12-14 13:20:09 +01:00
Jean-Francois Dockes
4afde68f9e
more config isolation. Still crashing
2012-11-30 16:45:02 +01:00
Jean-Francois Dockes
9146929f5e
implemented configuration for thread structure. Still crashes, but not often...
2012-11-29 17:26:48 +01:00
Jean-Francois Dockes
f252050247
ensure that indexing threads operations are transparent to high level callers such as the real time monitor
2012-11-26 15:40:31 +01:00
Jean-Francois Dockes
9079258f91
Observe followSymlinks in indexfiles()
2012-11-26 09:14:51 +01:00
Jean-Francois Dockes
cc79ee3f73
threads: added recoll_threadinit calls to block signals
2012-11-23 13:36:05 +01:00
Jean-Francois Dockes
ee7d0f2ee7
1st parallel multithreaded version of indexing which can do my home without crashing... Let's checkpoint
2012-11-01 11:19:48 +01:00
Jean-Francois Dockes
e0bc65bfdd
small mods inocuous or auxiliary to case/diac sensitivity but which can live in main branch
2012-09-13 12:25:01 +02:00
Jean-Francois Dockes
ee9dbda9fc
comments doc and formatting
2012-08-24 10:26:16 +02:00
Jean-Francois Dockes
8b34610dde
Cleaned up file name handling. Fixes that file names were sometimes indexed split, sometimes not. They now always are both, with different prefixes. Forces reindex
2012-04-13 09:18:08 +02:00
Jean-Francois Dockes
9f402d33cb
got rid of unused csguess module
2012-04-06 15:14:01 +02:00
Jean-Francois Dockes
85166c93b2
Changed the way we handle document sizes. The fbytes field should now be in most cases the most "natural" document size. pcbytes holds the top external container size and dbytes the text size
2012-03-07 15:39:30 +01:00
Jean-Francois Dockes
9bc2fc8958
Experimented with multithreading the indexing pipeline. Left undef'd as 15%-30% improvement of indexing time does not seem worth the complexity
2012-02-21 17:09:02 +01:00
Jean-Francois Dockes
fccc0ac09c
Get rid of the GUI indexing thread. All indexing now done by recollindex, with start/stop from the GUI
2012-02-07 17:07:07 +01:00
Jean-Francois Dockes
516863b5d6
GUI: perform up to date check before previewing a subdoc. This is for example to avoid showing the wrong message if a mail folder has been compacted
2012-01-20 17:48:55 +01:00
Jean-Francois Dockes
38e0957962
const string cleanup
2011-10-01 16:39:38 +02:00
Jean-Francois Dockes
e0aa67f0dc
let dir go through indexfiles() (name will be indexed, non recursive)
2011-09-30 08:44:50 +02:00
"Jean-Francois Dockes ext:(%22)
88685d2e64
search/index: fixed a number of bad conversions to properly deal with text documents bigger than 2GB
2011-07-12 08:28:09 -07:00
Jean-Francois Dockes
b6c73ecdeb
debug: improve consistency of log messages about up to date/processed files
2011-06-04 10:18:46 +02:00
Jean-Francois Dockes
478fb84ec4
index: add recollindex -f option to ignore skippedPaths/Names when using recollindex -i. Allows arbitrary (non-recoll) path handling for skipped subtree
2011-05-15 22:19:55 +02:00
Jean-Francois Dockes
55f124725f
Fix problems that occurred when multiple threads were trying to read/convert files at the same time (ie: indexing and previewing threads in the GUI calling internfile()). Either get rid of or lock-protect all shared data, eliminate misc initialization possible conflicts by using static initializers. Hopefuly closes issue #51
2011-04-28 10:58:33 +02:00
Jean-Francois Dockes
b28eaf23fb
Got rid of all the old RCS id strings
2011-04-27 08:22:17 +02:00
Jean-Francois Dockes
f4c1c3678d
indexing: an error on an archive member could crash or block the indexing because of the unclean way the ipath was passed in/out of internfile(). Closes issue #55
2011-04-25 16:41:43 +02:00
Jean-Francois Dockes
52fa332239
1.15.2
2011-02-14 20:30:34 +01:00
Jean-Francois Dockes
6c03417195
Move locafields parsing code from fsindexer to rclconfig for possible reuse
2010-11-22 15:56:14 +01:00
