Commit Graph

85 Commits (9bdee5c71c0057d095818dddb0cd15fcf6953399)

Author SHA1 Message Date
orbiter 2c595a6a47 added new methods to count the number of objects in RWIs. lots of refactoring was necessary to introduce new Rating class and to unify naming of methods
14 years ago
orbiter 75df87832c refactoring/better naming of methods and classes
14 years ago
sixcooler 5f8a5ca32d - not doing merge-jobs while short on Memory
14 years ago
sixcooler 07f5954570 try better handling of corrupt blobs
14 years ago
orbiter 1f300217f8 more protection for the cleanup thread
14 years ago
orbiter d13103a0a7 changed the way how the index cache is flushed: do not flush when a put was made because that could cause that many put calls synchronize for a long time when the dump or a merge is performed. Instead a watchdog thread is doing the dump and therefore puts cannot block any more which is good when a put happens during a search result preparation.
14 years ago
orbiter 1912d0cccc changed handling of RowSet element retrieval: until today all elements had been copied from the underlying byte[] arrays into a new Entry object that again had a copy of a portion of that byte[] in its own bye[]. There was an option to just refer to the underlying byte[] with a pointer but that was almost never used. This commit now changes an interface to the Row class where it is now necessary to tell if a copy is always required. Fortunately the copy is only needed in very rare cases. That means that this change should cause much less memory allocation; it is expected that this happens especially during search situations.
14 years ago
orbiter 6d2e252bcf fix for:
14 years ago
orbiter b666a929e7 fixed Semaphore handling in case of interruptions
14 years ago
orbiter 267290a821 removed the semaphores from the cache dump process because I believe some of the semaphores may be lost somewhere which then causes that the cache is never flushed and then the peer dies from a OOM. The re-introduced synchronization may not be the best solution but should ensure that the caches are flushed.
14 years ago
orbiter f803da8aae code cleanup
14 years ago
orbiter 4bea3f9714 hack to reduce resource contention caused by massive UTF8 decodings which use java.nio resources:
14 years ago
orbiter 10e2f588f8 - enhanced ybr ranking computation
14 years ago
orbiter 3ed4a09368 small features, some bug fixes and performance hacks
14 years ago
orbiter b45701d20f this is a re-implementation of the YaCy Block Rank feature
14 years ago
orbiter 123375bfba added a new yacy protocol servlet 'idx'. This returns an index to one of the data entities that is stored in YaCy.
14 years ago
orbiter deda54d684 - relaxed matching of string-search (this is now case-insensitive)
14 years ago
orbiter 1989ebc24b removed more warnings
14 years ago
orbiter a07a1a8b1e removed type cast warnings
14 years ago
orbiter a564230c48 more enhancements against blocked threads occurred in seed age evaluation (blocks httpd in some cases)
14 years ago
orbiter 3820525464 more memory protection: auto-flush of caches in case of memory shortage
14 years ago
low012 3b40b98256 *) set SVN properties
14 years ago
orbiter 619b561a4a enhanced secondary search: index abstracts decompression is now much faster and does not cause strong CPU load after several searches with more than one word
14 years ago
orbiter cb1f49d0f2 replaced all 'new String' with default encoding (missing) or UTF-8 encoding with a String generation method that uses a pre-defined Charset constant for UTF-8. This avoids a cache-lookup for the Charset object using String hashing of the String 'UTF-8'.
14 years ago
orbiter f8d0454c53 small bug fixes and experiments with search speed enhancement
14 years ago
orbiter 0ce17d823a - fixed bug in ordering
14 years ago
orbiter 5e45ded8e2 - removed locks from WordReference
14 years ago
orbiter d84b4a072e healing for some OOM problems
14 years ago
orbiter fe93caac5a added flags and administration options to show advanced search and to show search result attributes (for each search result)
14 years ago
orbiter feefe17568 npe assert fix
14 years ago
orbiter 387db84087 maybe found bug in non-working index dumper
14 years ago
orbiter e753027c43 fix for http://forum.yacy-websuche.de/viewtopic.php?p=21439#p21439
14 years ago
orbiter bf4ef1513e - fix for map view
14 years ago
orbiter db99db4be9 some redesign of the search-fail-response mechanism:
14 years ago
orbiter 7cd9d9d22a - enhanced DidYouMean computation using a faster count on index entries; this causes that results can be ranked better
15 years ago
orbiter 24f1cba7b2 performance hacks:
15 years ago
orbiter e54cb7fb0c more bugfixes (also for latest commit)
15 years ago
orbiter be6b48311c misc bugfixes
15 years ago
orbiter 39f409a7bb performance hacks
15 years ago
orbiter 64860dc1bb enhanced search event logging (to be used for further improvements)
15 years ago
orbiter 0f276dd63f - MapHeap now implements Map<byte[], Map<String, String>>
15 years ago
orbiter 6388a58fc7 better memory management and slightly less (in total and temporary) RAM allocation:
15 years ago
orbiter 5924a0d851 - enhanced concurrency in database index access for multicore
15 years ago
orbiter 37b8827a7a - removed the UPnP library sources from sbbi and added the jar library again. The library was included to get support for fedora releases, but after this time the fact that the sbbi cannot be part of fedora should be re-discussed. If this will still not be possible, then we may integrate the sbbi UPnP package using reflection.
15 years ago
orbiter fbf021bb50 redesign of index abstract processing - currently disabled until enough peers have fix in SVN 6928
15 years ago
orbiter 3a1cebb598 bugfixes
15 years ago
orbiter 51332b787d reverted SVN 6869 as discussed with dulcedo in car after LinuxTag:
15 years ago
orbiter b03caaa57a better handling of OOM situations
15 years ago
orbiter 439b44be9e removed exit from computation in ReferenceContainerArray.get merge method
15 years ago
orbiter 455a763d7c performance hacks
15 years ago