Commit Graph

18 Commits (56ff9d5fd4287203d1e95a7c65961efe0524cbe1)

Author SHA1 Message Date
orbiter 455a763d7c performance hacks
15 years ago
orbiter 90c3e5d6f6 - cleanup, removed unused imports
15 years ago
orbiter 93ea0a4789 enhanced remove operation in search consequences (which are triggered when the snippet fetch proves that the word has disappeared from the page that was stored in the index)
15 years ago
orbiter 8b8107b2a3 reduced IO-load and synchronization/blocking
15 years ago
orbiter ed07046870 flush only when > 3000 RWIs present + code cleanup
15 years ago
orbiter 1a8a134e0c continuing String-hash - to - byte[]-hash redesign that was started in SVN 6775 and continued in SVN 6790
15 years ago
orbiter 650be3599f added a time-out to the RWI cache to flush the cache if it has not been written for ten minutes. This additional dump criteria is necessary because some data sources repeat their vocabulary and may cause that the number of words in a RWI does not increase while the number of references in the RWI set increases. Now the RWI Buffer is flushed every 10 minutes or later if at that time already a dump is ongoing.
15 years ago
orbiter 1e8e79b9ef redesign of reference hash (URL-hash) parameter hand-over:
15 years ago
orbiter 749ffbd642 - added another catch case for the index dump and index merge process that should cause non-blocking behavior in case that index dump and/or index merge caused any unexpected exception.
15 years ago
orbiter 31e29a8831 - removed synchronization during index dump and index cleaning
15 years ago
orbiter eeca2ded92 fix for http://forum.yacy-websuche.de/viewtopic.php?p=18500#p18500
15 years ago
orbiter 37245430c3 fix for NPE during DHT RWI selection
15 years ago
orbiter 362b7a929b added extensive memory protection logic to avoid out of memory errors that may be caused by the RowCollection memory allocation function
15 years ago
orbiter 8281e29963 - more configuration for profiling graph (number of events)
15 years ago
orbiter 4a5100789f replaced _all_ size() == 0 with isEmpty() and all size() > 0 with !isEmpty(). The isEmpty() method is much faster in some cases, especially when used to access badly balanced hashtables where an size() operation becomes a large iteration.
15 years ago
orbiter 3528b970d6 - refactoring
16 years ago
orbiter bea3b99aff moved table and util classes
16 years ago
orbiter 1e4f8b56ed accumulated classes from different packages into the new rwi package
16 years ago