L2J Server

Hi.

Just wanted to give u some infos about all that FastMap/TIntHashs.

We start with: FastMap
A FastMap is nothing else then an regular HashMap/THashMap, but has a big difference... all entries are LINKED!
Thats also the reason why FastMap iterations are much faster then iterating over an HashMap or using forEach on THashMap.

Whats "setShared()" on FastMap?
This flag does 2 things:
1: The linked entries never loose it`s links to the prev/next entry which results in an "save" iteration while removing/adding values to an map.
2: It will self synchronized when adding/removing values, like "synchronized(map){map.put(key, value);}"
Ofc this synchronisation slows down the FastMap because it must enter the thread monitor when adding/removing values.

Well Trove is not far away form the Java originals (in basics) but as u know it use primitive datatypes as key/value. So far u need to sync every access to the map if u don`t want urgend exceptions or fails. ReentrantReadWriteLock is ur friend again.

When to use Trove, Java, Javolution? Less = better.
As u can see Trove is ONLY suitable for primitives!
So if u have to iterate more often or do not have an primitive as a key use Javolution.

And now about the lists....
FastList is an LINKED list. If u know the final size of the list or now that u have to add more then > 8 entrys an ArrayList is faster (even with expand)!!! Specialy those cases where u fill into FastList and use toArray()!!!

Hopefully this will help u somehow.

Thanks very interesting

So what you're basically saying is.. change knownlist FastList's to ArrayList?
(since there's a big chance u have to add more than 8 )

Interesting thread btw.

big thx link for my friend


If u dont read api docs of javolution http://javolution.org/target/site/apido ... stMap.html



FastMap is use better if too many values(more than 10k) if values size is lower need use HashMap

VISTALL wrote:big thx link for my friend


If u dont read api docs of javolution http://javolution.org/target/site/apido ... stMap.html



FastMap is use better if too many values(more than 10k) if values size is lower need use HashMap

That depends on the functionallity your data structure is intended to do. Each data structure performs in a different way based on the operation you are performing to it ( insertion, search, removal ).

Here's an example: The SkillsTable. It gets loaded whenever the server gets started or you call //reload skills. But it gets read very often , much more often than it gets inserted. So in that case, a data structure with small search complexity would be preffered over a data structure with small insertion complexity, since in total we got more "reads" than "insertions.

Issle wrote:

VISTALL wrote:big thx link for my friend


If u dont read api docs of javolution http://javolution.org/target/site/apido ... stMap.html



FastMap is use better if too many values(more than 10k) if values size is lower need use HashMap

That depends on the functionallity your data structure is intended to do. Each data structure performs in a different way based on the operation you are performing to it ( insertion, search, removal ).

Here's an example: The SkillsTable. It gets loaded whenever the server gets started or you call //reload skills. But it gets read very often , much more often than it gets inserted. So in that case, a data structure with small search complexity would be preffered over a data structure with small insertion complexity, since in total we got more "reads" than "insertions.

yeah is data is static(put only at start is not really difference), but if is ad L2World and Map of moving characters is use good FastMap

Trove 3.0.0 Final Stable has been released today

Changelog:

--- 3.0.0 ---
  - !!! Major restructuring of classes and packages !!!
    Interfaces have been broken out to allow for alternate implementations.
    The package changes will require code migration, although it should mainly
    consist of simply updating import statements. However, because everything is
    in new packages, both Trove 2.x and Trove 3.x can co-exist in a classpath
    (to hopefully ease the migration burden).
  - Broke out classes that allow custom hashing for performance reasons.
    New collections are:
       - TCustomHashMap
       - TCustomHashSet
       - TObject*CustomHashMap (ex: TObjectIntCustomHashMap)
    IdentityHashingStrategy is available for identity-based hash maps.
  - [ 3152276 ] Add sum() method to T*List
  - [ 3159432 ] THashMap.entrySet().remove() throws on non-Entries
  - [ 3153005 ] fix for null keys in Object maps
  - Benchmarks for comparing different libraries. Currently java.util, Colt,
    Trove 2 and Trove 3.
  - Fixes to serialVersionUID's and where Serializable is marked. For example,
    Serializable has been removed from the interfaces for various collections
    (to better match how it's done in java.util).
  - Many bug fixes, particularly to TLinkedList.

Download link: http://sourceforge.net/projects/trove4j ... p/download

Trove 3.0.1 is now available for downloads. This is mainly a bug fix release with a number of minor features.

Changelog:

--- 3.0.1 ---
Bugs Fixed:
- fix for ClassCastException in equals(T,T) when comparing with REMOVED
- [3408129] Benchmarks don't run correctly
- [3394094] Args to gnu.trove.generator.Generator not basedir-relative
- Re-introduce TLinkableAdapter class
New Features:
- [3412967] Need TMap interface
- .zip and .tgz files now expand into a sub-directory
- Documentation for generator patterns and replicated blocks
- Maven artifact generation now a bit more automatic

Download Link: http://sourceforge.net/projects/trove4j ... p/download

Trove 3.0.2 has been released. This release contains a couple of significant bug fixes and is highly recommended for all 3.x users.

Changelog:

Bug Fixed:
■ [3429703] TIntObjectHashMap losing entry
■ [3431135] ArrayIndexOutOfBoundsException at TIntIntHashMap.keys
■ [3432212] entrySet() does not unwrap in cased of no_entry_value or no_entrykey
■ [3432175] Decorator entrySet() -keySet() impl - Minor problem with the way null keys are unwarped

Download Link: http://sourceforge.net/projects/trove4j ... p/download

Trove 3.0.3 has been released. This release is a minor bugfix release recommended for all users.

Bugs Fixed:
- [3445639] ArrayIndexOutOfBoundsException from TIntHashSet.retainAll
- [3460395] no_entry_value is set always to 0 in map default constructor

Download Link: http://sourceforge.net/projects/trove4j ... ove/3.0.3/

MELERIX wrote:Trove 3.0.3 has been released. This release is a minor bugfix release recommended for all users.

Bugs Fixed:
- [3445639] ArrayIndexOutOfBoundsException from TIntHashSet.retainAll
- [3460395] no_entry_value is set always to 0 in map default constructor

Download Link: http://sourceforge.net/projects/trove4j ... ove/3.0.3/

Anyone using 3.0.3 have any issues? or should we update it in our lib?

3.0.3 give some errors with L2J (you can see them in eclipse), I guess is due some parts of code require update.

MELERIX wrote:3.0.3 give some errors with L2J (you can see them in eclipse), I guess is due some parts of code require update.

Thanks will look into it and keep you posted

Trove 3.0.2

values
public V[] values()

• Returns the values of the map as an array of int values. Changes to the array of values will not be reflected in the map nor vice-versa.

Specified by:

1. values in interface TIntObjectMap<V>

Returns:

• the values of the map as an array of int values.

Trove 3.0.3

values
public java.lang.Object[] values()

• Returns the values of the map as an Object array. Note that the array returned is typed as an Object[] and may not be cast to a more specific type. See #values(V[]) for a version which allows specifically typed arrays. Changes to the array of values will not be reflected in the map nor vice-versa.

Specified by:

1. values in interface TIntObjectMap<V>

Returns:

• the values of the map as an array of int values.

They seriously don't have what todo..