ae3df4f7c3
cjk phrases: hopefully the right fix this time for slack computation. lastpos-termcount correction was applied twice
Jean-Francois Dockes2019-07-06 13:52:51 +02:00
63cfcf0ad5
hldata generation during query processing: increase slack if position increases faster than term count (cjk)
Jean-Francois Dockes2019-07-06 08:28:05 +02:00
9ae775b095
snippets: allow a little more contiguous expansion of current snippet
Jean-Francois Dockes2019-07-06 08:26:42 +02:00
f877e7e459
Highlighting and snippets extraction: reworked to handle phrases properly. Use a compound position list instead of multiplying the OR groups inside a near clause
Jean-Francois Dockes2019-07-05 18:02:09 +02:00
00eb803f5d
Do not process hangul as words, but as ngrams. Same issues as with Katakana: word separation too hard
Jean-Francois Dockes2019-07-05 17:57:00 +02:00
75759ed481
GUI: snippets: dont recreate the window each time, allow displaying data for multiple documents. restable: update snippets when changing current row
Jean-Francois Dockes2019-07-03 13:46:38 +02:00
b79332580a
Regularise processing of hangul characters (there was a mixup of cjk/regular processing), and add a build-time option to either use cjk/ngram or regular term splitting for them
Jean-Francois Dockes2019-07-02 18:02:38 +02:00
3ccd2364b9
GUI snippets window: add options for the max list length and for sorting the snippets by page number
Jean-Francois Dockes2019-06-28 14:20:47 +02:00
0a460ea9c6
The container for temp files to be removed was a vector, but it needed stable member addresses. make it a list
Jean-Francois Dockes2019-06-27 11:12:01 +02:00
3f7d270691
GUI preview: improve operation when the index data is not up to date. Avoid erasing all the file index data in case the subsequent update fails (e.g. the file is locked). Improve the messages. Check for previous indexing error, and modify the message.
Jean-Francois Dockes2019-06-24 17:37:37 +02:00
4c2fd82d4e
pst: wait for pffexport and generate error if exit code is not 0
Jean-Francois Dockes2019-06-24 11:47:17 +02:00
ee8c5410bd
Avoid purging existing subdocuments on file indexing error (e.g.: maybe a file lock issue that will go away)
Jean-Francois Dockes2019-06-21 17:18:15 +02:00
e38e58c37a
In case the self-doc was not sent first by the handler, its udi was not recalculated, and it clobbered the last subdoc
Jean-Francois Dockes2019-06-16 13:46:00 +02:00
be214c4a5a
Take advantage of text storage when possible to display preview data for an unaccessible document
Jean-Francois Dockes2019-06-16 11:49:18 +02:00
9dcdb6e9a6
pdf: ocr function was broken for python3 in some cases (depending on how the ocr language was specified)
Jean-Francois Dockes2019-06-13 08:33:55 +02:00
b895980e95
PDF: fix the XMP metadata extraction code for python3 and other issues. Also get metadata from XML attributes
Jean-Francois Dockes2019-06-12 19:21:37 +02:00
f0944ae0b2
rclpst: indexing / searching mostly working with maybe issues in data charset conversions (check). Preview does not work, ipath needs conversion inside pffexport
Jean-Francois Dockes2019-05-28 18:39:37 +02:00
2c9bc17587
fix webengine reslist which had stopped working at some point in qt revs. Now working with 5.9-5.12 at least
Jean-Francois Dockes2019-05-27 17:04:28 +02:00
bc5ea83a3a
qt gui: fix highlighting for the table mode display. the query terms were fetched too early, before executing the query. Also share hiliter between restable and reslist and avoid allocating another DocSource for restable, share the one from reslist.
Jean-Francois Dockes2019-05-21 11:20:14 +02:00
8ddcc578ac
Reverted 34d43d1188 Keep only the main Snippet-producing makeabstract in rclquery, further formatting done in using modules This was just a bad idea. The common methods are also used by the python module
Jean-Francois Dockes2019-05-17 10:19:03 +02:00
a5810508ed
abstract: optimize the way we retrieve the wdfs by sorting the list of terms we query for. Big difference on very big docs
Jean-Francois Dockes2019-05-17 09:39:26 +02:00
fdb14e60ac
building abstract from stored text: limit count of terms explored to avoid taking forever on monster (multi mega-terms) documents
Jean-Francois Dockes2019-05-17 09:37:39 +02:00
780521ec6c
When checking if user input contains capital letters, take care of some lowercase letters which dont casefold to themselves
Jean-Francois Dockes2019-05-16 15:35:11 +02:00
Jean-Francois Dockes