fc07e62f10
Moved read-size logic from issuing instance / Msg39Request ot the processing instance
...
Each instance should be resposnsible for limiting processing sizes according to disk speed, cpu load, cpu speed and memory size. Having the issuing instacne do that doesn't make a lot of sense.
2016-08-01 16:14:34 +02:00
af4e70a051
Revert "Copied over changes from broken 'nomerge' branch"
...
This reverts commit 9e48989e93
2016-08-01 14:52:50 +02:00
9e48989e93
Copied over changes from broken 'nomerge' branch
2016-08-01 14:52:09 +02:00
e3336521fb
Removed Msg39::m_stackBuf
...
Misnamed member - it hasn't been on the stack for a long time. m_lists is now a regular array allocated with new[]
2016-07-19 16:19:45 +02:00
80cd8d78ba
Remvoed Msg39Request::m_reserved* members
2016-07-11 17:16:48 +02:00
a30d24ed13
Removed Msg39Request::m_makeReply
2016-07-11 17:16:48 +02:00
e315dda7a0
Msg39 member renames
...
m_tt -> m_toptree
m_r -> m_msg39req
2016-07-11 17:10:53 +02:00
16f9aff736
Simplify Msg39::getDocIds()
2016-07-11 16:40:04 +02:00
3b6c9e9a5e
Close of more of Msg39
2016-07-11 15:57:23 +02:00
096856e9cb
Close of more of Msg39
2016-07-11 15:48:02 +02:00
83070b1078
Close of more of Msg39
2016-07-11 15:33:40 +02:00
51017cd683
#include cleanup
2016-07-11 14:57:14 +02:00
ad81c9518c
Close of more of Msg39
2016-07-11 14:48:20 +02:00
0472b96dc8
Made Msg39Request::reset() non-inline
2016-07-11 14:26:43 +02:00
8d54c24991
Close of more of Msg39
2016-07-11 12:36:29 +02:00
d634c7bff4
Renamed Msg39::m_tmpq to m_query
2016-07-11 12:31:03 +02:00
0865ef7f09
Removed superfluous Msg39::m_docIdStart/m_docIdEnd
2016-07-07 16:02:39 +02:00
de2b46a09e
Split Posdb.* source files
...
Posdb.* split into Posdb.* (the DB) and PosdbTable.* (ranking)
2016-07-04 15:06:55 +02:00
c42c46cb14
Support fx_qid. Carried all the way through to msg39request
2016-06-17 14:03:31 +02:00
8dab0db467
Remove extra semicolons
2016-05-19 18:37:26 +02:00
a746eb91e5
Support enabling/disabling high-freq-term-cache via CGI parameter
2016-05-13 16:07:58 +02:00
be9840d1c0
Replaced Threads.* with a jobscheduler
...
Threads were being created and destroyed which can be expensive. The
thread-per-job model has been changed to a job scheduler that manages the job
queues and threads in pools. The submission of a job now specifies start/finish
routines, state, and as precisely what kind of job it is. The job scheduler then
takes care of the rest. it is hidden how many queues and pools there are.
2016-04-29 14:27:27 +02:00
d95976ac03
IndexList cleanup
...
Removed most uses of IndexList. Removed unused member functions from IndexList.
2016-04-12 16:25:10 +02:00
e57fb2f72c
Removed explicit m_buf[0] from Msg39Request/Msg39Reply
...
This also fixes the sizeof/(...) versus &m_buf-this confusion which is currently
harmless but a bug in the upcoming encoding changes.
2016-04-04 12:36:32 +02:00
12265b4ed9
Removed double assignment of m_getDocIdScoringInfo
2016-03-11 11:36:39 +01:00
ab0b9d03ea
Standardize header guards
2016-03-08 22:16:02 +01:00
035a2f3e59
Send back in results an estimate of how much of the desired index scope was actually searched.
...
Dead shards and deadline-cutoff docid splits subtract from this. Unavailable
summaries have no influence on the estimate.
2016-03-01 14:14:58 +01:00
92d29e62e5
Removed unused macro MAX_MSG39_REQUEST_SIZE in Msg39.h
2016-03-01 11:52:24 +01:00
b69c90df10
Remove m_boolFlag variable
2016-02-28 13:03:02 +01:00
b5d7b6e26c
Remove more facet codes
2016-02-25 16:54:56 +01:00
19918297cd
Remove more sections codes
2016-02-24 13:42:56 +01:00
bd4ad2feb3
Do not docid ranges if the deadline would be crossed
2016-02-22 17:45:06 +01:00
5ecdd52131
Remove unused Msg39::doDocIdSplitLoop
2016-02-17 16:08:13 +01:00
80b8f1bb91
Changed Msg39::m_timeout (currently unused) to milliseconds
2016-02-15 12:12:06 +01:00
abd465f5a4
Removed unused members Msg39::m_nowUTC and Msg40::m_nowUTC
2016-02-15 11:40:22 +01:00
60d94b8ba1
Removed Msg39Request::m_useSeoResultsCache and related code
2016-02-05 14:33:31 +01:00
df810c7d10
Cleanup of Msg39 (private members, friend declarations, etc)
2016-02-04 17:44:32 +01:00
2946ed654e
Removed even more unused memebrs from Msg39
2016-02-04 17:17:29 +01:00
c472a824fb
Remvoed 'hack' members of Msg39
...
Apparently they are used by Matt's secret edition, there is no need to us to
carry over complexity only useful to non-public code.
2016-02-04 17:09:58 +01:00
9705538f32
Removed unused Msg39 members m_topScore50/m_topDocId50
2016-02-04 16:58:22 +01:00
46f8453c92
Remove unused m_fastIntersection
2016-01-27 00:02:09 +01:00
d49cc62223
Remove unused m_useNewAlgo
2016-01-26 23:57:21 +01:00
1177eaee35
Remove unused variable m_useMinAlgo
2016-01-26 23:34:56 +01:00
f1a52d3d15
Removed effectively constant-value Msg39::m_restrictPosdbForQuery (always false)
2016-01-21 15:20:27 +01:00
2480728703
Removed outdated comments and commented-out code
2016-01-21 14:21:11 +01:00
725df6b6e5
Removed unused member Msg39::m_seoDebug
2015-12-29 16:17:44 +01:00
dd21abf042
Clear Msg39Request in reset() so we don't send uninitialized bytes over the network
2015-11-13 11:04:09 +01:00
eaeb641680
Remove uncompiled/unused files
2015-11-05 12:01:47 +01:00
c8cf0e5440
fix some mem leaks from allowing really big queries.
...
added a max query term control to search controls to
limit users doing really big queries. but default it
very high to 1M.
2015-07-13 23:17:53 -06:00
f088e734f6
allow up to 3000 query terms. really we can allow
...
much more since we are mostly dynamically allocating,
only a few smaller arrays use the 3000 on the stack.
2015-07-10 19:02:30 -06:00
