abe7c4fb56
Do not register types with encoding/gob
...
Registering the same type multiple times can lead to a panic. While the
panic is handled in Save, the handling prevents saving the cache.
While doing this seems like a good idea (since then the caller wouldn't
have to register the types themselves), the need to deal with duplicate
entries causes issues with Load. If calling Load to initialize a cache
right after startup, encoding/gob will have no information about the
types written in the cache, and thus won't be able to load the cache.
There are also issues w.r.t. how the type is registered in gob, and
if they are registered in a different form inside go-cache than they
were outside, a panic happens due to different paths being generated:
https://github.com/golang/go/blob/master/src/encoding/gob/type.go#L857
Make sure that it's the caller's respansibility to register types with
encoding/gob, to avoid issues with gob's Register and so that they can
both Save and Load without having to Save beforehand in the same instance.
2014-12-11 17:41:46 +09:00
7d1d6d6ae9
Note about needing explicit synchronization if you want to use the returned items map and its cache at the same time
2013-08-08 15:10:54 -04:00
3ff6a5407e
Reference 'go doc' in the README
2013-08-08 15:02:23 -04:00
c383ebd2ba
Remove go doc-style reference from README since go doc now shows public methods from (private) embedded structs
2013-08-08 15:01:48 -04:00
e91e36d111
ItemCount() is equivalent to len(c.Items()), not len(c.Items)
2013-08-08 14:54:03 -04:00
fd89281dfd
s/field/fields/
2013-08-08 14:53:24 -04:00
f8cc27acbc
Export Item wrapping type, and add an Items() method to cache
2013-08-08 14:53:02 -04:00
baa5a64da4
Revert 3986bff for now
2013-06-30 22:05:40 -04:00
2fb27e8369
Change the names of the MutexMap benchmarks to RWMutex to clarify the changes to the map benchmarks and the cache itself
2013-06-30 20:40:26 -04:00
155ab21e5d
Use an RWMutex in the cache-equivalent map tests now that the cache uses one
2013-06-30 20:31:46 -04:00
3986bff69b
Don't keep the read lock after traversing the items map (before saving to the io.Writer) in Save
2013-06-30 20:20:38 -04:00
5388b25b3b
Use an RWMutex instead of a Mutex, making Get a read operation only (this slightly changes the eviction behavior: lookup doesn't completely expunge an expired item, but the janitor still will at the next cleanup.) Also, use the same RWMutex in Load and Save (thanks, Alan Shreve)
2013-06-30 20:11:57 -04:00
1fc39f1402
Add ItemCount method description to README reference
2013-04-18 15:26:16 -04:00
1140511315
Condense the reference in the README by removing the duplicate descriptions and providing one general one for all of them
2013-04-18 15:24:40 -04:00
739fb494dc
Update LICENSE file to 2012-2013
2013-04-18 14:35:10 -04:00
c9f5754801
Add O(1) ItemCount method to retrieve items currently in the cache map (whether expired and not cleaned up or not)
2013-04-18 14:32:01 -04:00
d5d03c28d4
Add specialized increment and decrement methods
2013-04-18 14:24:30 -04:00
b5601e904d
Clarify licensing on contributed code
2012-10-08 12:34:42 +02:00
525c0ccc26
Add CONTRIBUTORS file and 'and the go-cache contributors' bit to the copyright line in LICENSE
2012-10-08 12:31:23 +02:00
e30c8eff7a
Merge pull request #4 from temoto/patch-1
...
Attempt to close the file in SaveFile/LoadFile, and return either the serialization error, or the Close() error.
2012-10-08 03:27:04 -07:00
a768a84319
Save/LoadFile must close file
2012-09-19 03:25:42 +04:00
72916cf4df
Update README to reflect API change: IncrementFloat now only works on float32 and float64, and the use of Decrement should be encouraged over Increment (with a negative number) since Increment converting a negative signed integer to an unsigned integer probably isn't the intended behavior
2012-08-17 13:48:55 +02:00
b3a957a46c
Handle float64 and int64 Increment/Decrement separately, and don't use Increment(-n) in Decrement to support uints
2012-08-17 13:39:02 +02:00
9cc10f6f2f
Use a type switch instead, and Use unexported NewSharded in test
2012-08-17 11:35:20 +02:00
13225a8ae4
Unexport the shardedcache interface and implementation until both the API and implementation are solid
2012-08-17 11:32:24 +02:00
15cda21ff4
Add an interface, embed the mutex, and unexport some unnecessarily exported fields
2012-07-04 19:34:37 +01:00
bf97a87cc7
Cleaner bucket selection
2012-06-22 09:33:27 +01:00
52c269d8ae
Testing a sharded cache. Could be useful for massively parallel applications
2012-06-22 09:24:09 +01:00
0f0584a805
Fix BenchmarkCacheSetDeleteSingleLock 'cheating' by unintentionally only locking once in the whole benchmark, rather than once per set and delete
2012-06-22 06:52:32 +01:00
0a962bf9eb
Keep the same tense in docstrings, and remove a little ambiguity
2012-06-22 04:24:48 +01:00
8f6294df58
Remove unused Version const
2012-06-22 04:10:33 +01:00
2b93e5eceb
Stay below 80 characters in width in the README as well
2012-06-22 04:04:21 +01:00
026efae39f
Stay below 80 chars in width
2012-06-22 03:56:12 +01:00
724bd0e963
Move comment explaining the manual mutex unlocking down to the first place where we actually do it
2012-06-22 03:51:34 +01:00
aa7f52c169
Be clearer about the defaults, <80 width for the janitor comment, and remove 'default cleanup interval' -- there can be only one
2012-06-22 03:50:10 +01:00
8495026156
c can be collected. It isn't necessarily (right after)
2012-06-22 03:44:04 +01:00
23661b37bd
Point out why using go-cache might be desirable
2012-06-22 03:11:31 +01:00
803ceeaf1a
Use runtime.numCPU() workers in concurrent benchmarks, remove pointless benchmarks for racy maps, and add replacement benchmarks using mutexed maps
2012-06-22 03:06:06 +01:00
3bd539b94d
Merge pull request #2 from dustin/master
...
Limit BenchmarkCacheGetConcurrent goroutine count
2012-06-21 18:50:17 -07:00
fbcdb10061
Don't overparallelize the concurrent test.
2012-06-21 17:36:24 -07:00
3c46230312
use an integer example for the second to make it more obvious that the cache stores anything, not just strings
2012-05-10 11:58:45 -05:00
c1bb5e2a35
The .gitignore file is no longer necessary
2012-04-18 09:40:25 +02:00
74f0b20353
Don't fmt.Println warning if gob.Register fails--just return an err
2012-02-23 04:16:42 +01:00
84d15102eb
Add test for concurrent cache.Get
2012-02-21 18:46:25 +01:00
8805e79189
Properly handle Gob panic
2012-02-19 01:21:07 +01:00
91bd4334f1
Don't use defer for mutex unlocking (it currently adds ~200ns)
2012-02-17 01:40:55 +01:00
13b338b204
Modify error on Gob panic
2012-02-17 00:22:46 +01:00
dfeb3cfa2b
No reason to use *sync.Mutex
2012-02-17 00:20:20 +01:00
39b4f413a1
Don't format in t.Error()
2012-02-13 21:37:17 +01:00
6660155e6c
gofmt
2012-02-12 01:54:16 +01:00
Diogo Franco (Kovensky)
Patrick Mylund Nielsen
Patrick Mylund Nielsen
Sergey Shepelev
Dustin Sallings