5c08271e74
improve numeric stream analysis performance
2011-01-09 19:35:24 +02:00
3c3d01347e
improve all stream analysis performance
2011-01-09 02:51:41 +02:00
8de7beadc8
Highlighting: Sometimes highlighting returns empty fragments even thought there should be, closes #613 .
2011-01-08 03:19:49 +02:00
7364159d69
NPE for when passing null date to content builder, closes #610 .
2011-01-07 23:28:57 +02:00
286a6fac76
upgrade to jackson 1.7, also fix (really unused case) of getting compressed smile stored through REST with smile context type
2011-01-07 16:35:52 +02:00
b9be6d9ea7
Geo-distance in scripts (like custom_score), closes #607 .
2011-01-07 15:22:11 +02:00
8b8bad7572
build on aferreira improvements to handle empty parameters with no =, and improve component decoding logic
2011-01-07 14:34:34 +02:00
2d81acbbf2
Adding test for edge cases of RestUtils.decodeQueryString()
...
When writing tests for the fix of decodeQueryString() to
handle gracefully edge cases like:
&a
&a&b
it arises the question of what behavior is desirable
for these pathological cases (vs the regular p=v pairs).
This change just skips them which is consonant
to the preexisting code.
To be thorough, we add tests for a bunch of edge cases, like:
QUERY PARSED PARAMS
? {}
?& {}
?= { "": "" }
?a {}
?p=v&a { "p": "v" }
?p=v&a&p1=v1 { "p": "v", "p1": "v1" }
?a&b {}
2011-01-07 14:13:26 +02:00
154dcf007e
Make RestUtils.decodeQueryString() more robust in edge cases
...
The code of decodeQueryString() had some trouble with weird URLs:
(1) an input like "uri?param&p=v" causes an exception to be thrown
(2) an input like "uri?param1¶m2" causes an infinite loop
This could be verified against an ES server with requests like
curl -XGET localhost:9200/test/_analyze?t&text=this+is+a+test
# the exception stack trace shows up in logs
curl -XGET localhost:9200/test/_analyze?t1&t2&text=this+is+a+test
# never returns, never ends
This change fixes these issues.
2011-01-07 14:13:25 +02:00
edb075b611
Minor changes to RestUtils.decodeQueryString() tests
...
Testing
RestUtils.decodeQueryString("something", "something".indexOf('?') + 1, params);
is not really checking decoding of an empty query.
Instead, it is testing decoding of "something"
as a query (because "something.index('?')+1" evaluates to 0).
The parameter map is left empty because
"malformed" pairs like "something" are currently skipped.
Instead, this change modify this test to check the edge cases:
+ fromIndex >= queryString.length()
+ fromIndex < 0
2011-01-07 14:13:25 +02:00
fe0f9ebc9d
make lucene document creation allowed to be indexed twice
2011-01-07 14:09:39 +02:00
c7cb353956
fix serialization of count request
2011-01-06 20:44:36 +02:00
6a0a9ff374
parse query with offset
2011-01-06 16:13:15 +02:00
b9e808f755
improve version loading by going directly to the relevant reader
2011-01-06 11:50:16 +02:00
26c5f6c482
Allow ./bin/plugin to install plugins from the local filesystem / custom url, closes #595 .
2011-01-06 10:46:42 +02:00
cc5978393b
with the new stemmer analyzer, don't break when there is no language stopwords, but just create it without stopwords
2011-01-06 09:53:56 +02:00
4a45df88c6
Add Snowball stemmer as analyzer and filter
2011-01-06 09:40:49 +02:00
66d63055df
Add to node stats the number of times field cache was evicted due to memory constraints, closes #603 .
2011-01-05 22:37:15 +02:00
754b0d7a0f
Add `resident` field cache type, closes #602 .
2011-01-05 21:23:47 +02:00
0954df0623
change the default interval for gc checks to 1 second
2011-01-05 20:40:52 +02:00
31231531e1
Highlighting: The result of highlighting for a hit can contain data from another document, closes #600 .
2011-01-05 15:45:46 +02:00
6258915205
Bulk API: Add how long the bulk API took (in milliseconds) to the response, closes #599 .
2011-01-05 14:12:45 +02:00
29b308005a
try to be more predictable when doing pagination (still can return duplicates when going to different shards) by sorting on the shards results based on the shard id first
2011-01-04 23:42:44 +02:00
0faa05b3f2
add more integration versioning tests
2011-01-04 18:09:46 +02:00
9423378f27
add heavy concurrent updates to same doc, and make sure it has the same data on all replicas
2011-01-04 17:44:33 +02:00
45c1ab06b3
Versioning, closes #594 .
2011-01-04 04:04:30 +02:00
9335b3a9e6
Search: Add a `timed_out` element indicating if the search request timed out, closes #592 .
2011-01-03 15:25:24 +02:00
ff6d7254fe
optimize query parsing to use single field parser when executing on single field
2011-01-03 15:14:28 +02:00
2c8c6aa22a
add a uid scan benchmark to check perf
2011-01-03 14:34:38 +02:00
dc17d10b81
small refactor
2011-01-03 13:32:09 +02:00
5e029865a8
Des.config=/path/to/config/file doesn't replace $ES_HOME/elasticsearch.conf, just appends to it, closes #588 .
2011-01-03 12:53:50 +02:00
ccdbbef276
don't send leave request when the master node is null
2011-01-03 12:01:33 +02:00
d2a70b21cd
key (which is time) can be negative as well
2011-01-02 22:03:52 +02:00
07d361816e
Search: Date Histogram Facet, closes #591 .
2011-01-02 15:07:45 +02:00
d523104b91
add some sanity tests with joda for date time manipulation
2011-01-01 23:48:25 +02:00
b7b775228e
add some sanity tests with joda for date time manipulation
2011-01-01 23:40:46 +02:00
aec720218d
Histogram Facet: Improve perf / memory by defaulting to just providing counts with no totals, closes #587 .
2011-01-01 17:27:55 +02:00
45b93e2781
add histogram facet benchmark
2011-01-01 16:33:18 +02:00
e6b4834768
Highlighting: Allow to highlight on fields without term vectors, closes #585 .
2010-12-31 18:57:26 +02:00
7c959e7ec3
Master based operations (create index, delete index) to automatically retry on retryable cluster blocks (like recovery from gateway)
2010-12-31 14:09:45 +02:00
a92dbc537a
those two should not be final...
2010-12-31 00:23:04 +02:00
fd593acafe
make every static field final (immutable)
2010-12-31 00:22:31 +02:00
6b672e29f5
Remove infinite loop typo
2010-12-31 00:01:45 +02:00
82298d890c
Terms Facet: Performance and memory improvements when faceting numeric fields, closes #583 .
2010-12-30 19:45:36 +02:00
019359a896
add more scanning options on the field data cache
2010-12-30 17:08:32 +02:00
6c8b06c967
clean unused method
2010-12-30 16:45:12 +02:00
8e8a3a894f
upgrade testing framework jars, improve build scripts to share the testing jars
2010-12-30 16:08:38 +02:00
fea93b7fea
Delete By Query: Types are not serialized over network, causing them to be ignored, closes #580 .
2010-12-30 14:43:08 +02:00
789e1a343f
failing to load plugins, return an empty map
2010-12-30 13:16:09 +02:00
2e46add991
remove debugging dead code
2010-12-30 13:14:49 +02:00
51277a9486
no need to check for shard if tis null, we know its not
2010-12-30 13:12:08 +02:00
c4646a2b55
use close helper method to close streams
2010-12-30 13:11:20 +02:00
76d042f3c5
explicitly pass null values to methods where applicable
2010-12-30 13:09:11 +02:00
5a45e9c8bd
close stream if file not found (though a stream is not really opened in this case...)
2010-12-30 12:30:50 +02:00
20b6e0bdde
close stream if file not found (though a stream is not really opened in this case...)
2010-12-30 12:30:33 +02:00
4f8b4b405c
though not really needed (working on in memory byte array), close the stream opened
2010-12-30 12:27:46 +02:00
ed996c3e85
Non-data master nodes and non-master data nodes fail to store data, closes #579 .
2010-12-30 12:10:40 +02:00
e3322836b5
make settings prefix simpler and not bail when not prefixed with org.elasticsearch, allow to provide settings prefix for analyzer provider
2010-12-29 19:12:38 +02:00
f73a5e62d3
improve error message
2010-12-29 18:55:37 +02:00
c7fca8f0e0
refactor terms facet into string terms facets, paving the way for optimized ones for numeric values
2010-12-29 18:34:08 +02:00
e764f41341
await closing sockets on shutdown
2010-12-29 12:49:37 +02:00
f993c4b72b
start refactoring of facets to make them more pluggable
2010-12-29 12:49:02 +02:00
c8bfa455ad
REST Create Index: Not taking JSON index settings into account unless wrapped in settings.
2010-12-29 12:10:51 +02:00
d8aef57baa
add filter_cache_size to node stats
2010-12-28 16:57:55 +02:00
f908a73ec6
allow to construct transport client with settings builder
2010-12-27 20:11:28 +02:00
a84e0a49b7
change scope to _score
2010-12-27 19:19:17 +02:00
d73a459f1e
when parent mapping is set, fix setting routing to required
2010-12-27 18:41:28 +02:00
d560ceabe4
use high transport type when sending the cluster state
2010-12-27 16:50:55 +02:00
4245eb7395
add node level field cache size to node stats api
2010-12-26 12:18:15 +02:00
5dcba8a38c
revert back a stupid bug in setting compress to true always introduced in the last commit
2010-12-26 12:05:22 +02:00
49deb80d03
when compression is enabled, use the compressed byte buffer to write it to the translog (less data to write), and handle recovery by being able to parse also compressed source
2010-12-24 01:09:15 +02:00
f2eae5b605
Memory Index Store: Separate to two buffer size types, and fix bugs, closes #577 .
2010-12-23 23:21:32 +02:00
22fa91efa0
add reconnection code between nodes that are not masters
2010-12-23 17:26:55 +02:00
f6afc01fe5
add reconnection code between nodes that are not masters
2010-12-23 17:26:32 +02:00
ff5990daec
listener should be threaded
2010-12-23 17:24:23 +02:00
de7dd3c070
default to try and connect after network disconnection with fault detection
2010-12-23 16:51:09 +02:00
e313379ed5
only force shutdown on cached threadpool if its not the same as the execture service
2010-12-23 16:13:05 +02:00
42e8567477
interrupt joining the cluster thread if shutting down
2010-12-23 16:11:43 +02:00
5f25ae4f2f
check if the scheduled thread pool is not terminated when trying to terminate it
2010-12-23 15:18:22 +02:00
9f72a8d68b
use cached thread pool for notifying on disconnections
2010-12-23 14:22:40 +02:00
6dcc04b59c
Transport: Increase the default transport.tcp.connect_timeout from 1s to 30s, also add `network.tcp.connect_timeout` to conform with other common network settings, closes #576 .
2010-12-23 14:20:11 +02:00
5c338b7af2
add message to connect transport failure exception
2010-12-23 13:05:39 +02:00
5a58ed4274
when replacing a mapping, clear the aggregated mapping information and then readd it
2010-12-23 11:32:16 +02:00
473c2fa8f4
add reason for state change logging in index shard, add debug logging on ignore recovery in when handling cluster change in indices cluster
2010-12-23 10:56:37 +02:00
5ac42f2a4f
Excessive mapping parsing when cluster state changes, closes #575 .
2010-12-23 09:53:05 +02:00
f869951364
rename type method on MapperService to the more descriptive documentMapperWithAutoCreate
2010-12-23 08:43:17 +02:00
e0b2b4b4a6
add a warn when not using the custom FieldCache with eager reader based eviction
2010-12-23 07:58:43 +02:00
800b7f39bd
add toString to CompressedString to show when logging it
2010-12-23 07:45:04 +02:00
3b30930460
improve logging when trying to delete unallocated shard, check first if deleting is required, and only then delete it
2010-12-23 07:37:49 +02:00
ac73334c87
make id cache an index component, so a new instance will be created per index
2010-12-22 14:21:59 +02:00
bb0d7190e9
make id cache an index component, so a new instance will be created per index
2010-12-22 14:21:31 +02:00
1ee24ce60f
add an option to provide an OutputStream to a builder, allowing to generate directly into a stream provided by the user
2010-12-22 13:26:29 +02:00
eef3a95fa6
move the query parser cache to be index level and not node level
2010-12-22 12:31:30 +02:00
33339ae4b1
improve concurrency control in filter caching, reuse one if it was created already under concurrent request
2010-12-22 10:42:38 +02:00
d283c158d2
optimize for the more common case of single value facets when loading values
2010-12-22 09:04:35 +02:00
59ee8493ce
use public terms filter in all places
2010-12-21 23:43:05 +02:00
a1df3c637c
improve logic of when to load fields from source, only if they actually have mappings, otherwise, ignore them (as was the previous behavior)
2010-12-21 13:02:15 +02:00
3035254885
Search API: Allow to pass `&fields=` without any fields to cause only id and type to be returned, closes #572 .
2010-12-21 12:01:51 +02:00
a96b294de0
add multicast socket to execption message
2010-12-20 20:07:51 +02:00
efc95dc00a
Java Search API: Passing a the query as json fails, closes #568 .
2010-12-20 11:31:09 +02:00
789073639d
fix transport client compilation problem, the client cluster service was not used and removed
2010-12-19 05:50:07 +02:00
8dfc3acba0
when using dates in scripts, initialize the dates (DateTime) with UTC
2010-12-19 05:01:59 +02:00
96a2950ab5
Disable calling mlockall by default, can be enabled by setting bootstrap.mlockall to `true`, closes #567 .
2010-12-19 02:30:48 +02:00
415bb5d7f3
Java TransportClient: By default, don't sniff other nodes and use addresses provided as is, closes #565 .
2010-12-17 19:31:28 +02:00
005c6d9124
allow to set term index divisor
2010-12-16 21:56:59 +02:00
ff0eeb3d1a
improve getting setting as array, automatically support comma delimited values under the exact setting
2010-12-16 05:02:24 +02:00
f827b7cff5
fix settings
2010-12-16 00:48:26 +02:00
d6bab1a892
Tcp Transport: Connection pool between nodes and different connection types, closes #564 .
2010-12-15 20:11:10 +02:00
085066ed19
set, by default, the send and receive buffers to 32k on tcp
2010-12-15 05:54:09 +02:00
36be13ffb3
bring back the file chunk size in recovery to 100kb from 500kb
2010-12-15 01:28:21 +02:00
b6277ac14d
increase netty worker count to default to the updated value netty defaults to
2010-12-15 01:16:00 +02:00
bb178c9c1a
cleanup
2010-12-13 21:30:26 +02:00
d0eed79175
add log message on hte concurrent streams used
2010-12-13 00:08:57 +02:00
ceb7b90740
Recovery: Allow to control `concurrent_streams` (per node) during recovery
2010-12-13 00:03:04 +02:00
1c5477d4ed
Improve local node recovery process, removing using a block as indication for an index not being recovered. Moving to shard replication group level allocation (with the ability to distinguish between api based creation or recovery based creation). This also allows for throttling based control on initial local gateway based shard allocation.
2010-12-12 22:46:27 +02:00
bc04243a2b
Search Fields: If a field is not stored, automatically extract it from _source (without the need for _source prefix), closes #562 .
2010-12-12 05:00:24 +02:00
216b2ab912
Highlighting: Automatically use the field values extracted from _source if not stored explicitly in the mapping, closes #561 .
2010-12-12 04:40:25 +02:00
fe3f5d45de
remove global settings from the settings conf, a relic that is no longer needed...
2010-12-12 02:37:36 +02:00
dddca692b8
add search hit phase, which operates on a hit, and have all relevant hit phases implement and use it
2010-12-12 02:01:13 +02:00
be4b2e2de6
externalize script search lookup into a generalized lookup that can be used outside of scripts
2010-12-12 01:01:56 +02:00
a914865c45
Close API might cause index data to be wiped, closes #560 .
2010-12-10 22:19:45 +02:00
bc2dc9465a
add to the receive timeout exception the request id that it timed out on
2010-12-10 12:40:39 +02:00
2a7f964ca2
Geo search fails to find some docs when wrapping over IDL, closes #559 .
2010-12-10 08:39:10 +02:00
9d2fe162fa
better log message
2010-12-10 00:17:55 +02:00
3f6ed7e1de
improve update mapping on master, if we end up with the same mappings as the one the cluster state has, no need for a new cluster state...
2010-12-10 00:17:10 +02:00
167d35807c
Add terms/in query, alias terms filter to be in filter as well, closes #557 .
2010-12-09 07:25:05 +02:00
0828f18738
better child bench
2010-12-09 06:11:06 +02:00
e1495ff8be
add scrolling update test
2010-12-08 20:56:09 +02:00
34f3f3f79e
Aliased analyzers cause index deletion / cleanup failure, closes #555 .
2010-12-08 19:30:38 +02:00
f5a8c3881f
add debug logging on failed bug items executions
2010-12-08 01:21:21 +02:00
54437c1bd3
Parent / Child Support, closes #553 .
2010-12-08 00:16:05 +02:00
1a8017d17e
upgrade to jackson 1.6.3
2010-12-07 09:14:13 +02:00
22c27e4ff8
correct field name
2010-12-05 00:02:53 +02:00
d3cae103d5
wait for green status in test
2010-12-04 23:57:44 +02:00
d659bd0479
upgrade to jackson 1.6.2
2010-12-04 00:46:57 +02:00
a90684aaa5
Upgrade to Lucene 3.0.3, closes #549 .
2010-12-03 20:43:51 +02:00
d9610ed0a1
Search API: Optimize single shard search to `QUERY_AND_FETCH` from any other search type, closes #547 .
2010-12-01 03:59:03 +02:00
3ea19a514b
Index Setting: Add `index.refresh_interval` to simplify setting instead of "index.engine.robin.refresh_interval` (still works), closes #546 .
2010-12-01 03:38:41 +02:00
2b039bd0b0
change the default chunk size when recovering from the default 100kb to 500kb.
2010-11-29 14:33:49 +02:00
c0945831e8
Search Scroll API: Allow to provide the scroll id as the body of the scroll request, closes #545 .
2010-11-29 13:59:30 +02:00
103221e58a
remove the timed out message, can be misleading
2010-11-29 02:55:23 +02:00
84f97e96a5
Scrolling issues, fix releasing search context eagerly, they should not be released when scrolling, closes #136 .
2010-11-29 02:05:45 +02:00
6d9576c217
Reduce the number of concurrent recoveries per node to 2 from number_of_processors, closes #543 .
2010-11-28 16:35:39 +02:00
a222de1dbf
internal change to scope based collectors
2010-11-28 14:57:17 +02:00
8285ffe221
encode scroll id with base64 in url safe manner
2010-11-28 02:23:38 +02:00
cfa8c9aa79
Index / Index Template: Allow to define `_default_` mapping, closes #542 .
2010-11-27 23:30:18 +02:00
526f28f479
Create Index API: Allow to provide mappings. closes #541 .
2010-11-27 00:28:18 +02:00
151715dc2f
also merge lists
2010-11-27 00:05:11 +02:00
577f06fd43
add indices API to groovy client
2010-11-26 22:19:57 +02:00
73e5eb9e14
clean compile warnings
2010-11-26 16:03:25 +02:00
kimchy
Adriano Ferreira
harryf
Tim Dysinger