Skip to content

Bazist/HArray

Repository files navigation

HArray

Build status

Probably, this is most optimized Trie structure in the World ! Thats all what you need know about this project :)

HArrayInt - Key\Value In Memory structure for 32bit keys

HArray - Key\Value In Memory structure for keys with variety size

HArrayChar - Wrapper for HArray with interfaces: char* key, uint32 keyLen, char* value, uint32 valueLen

HArrayUniqueIntValueList - Wrapper for HArray with inteterfaces: uint32* key, uint32 keyLen, List<uint32> listOfUniqueValues



Overview

  • High optimized Trie structure implemented in more than 8K lines
  • Tested on 1 Billion keys succesfully
  • Keys with variable length inside one instance of the container
  • Without any Stop World events such as Rebuild/Rehashing on Insert key.
  • Without any Hash Functions, the container has adpative algorithm for different nature of keys
  • Scan by Prefix/Scan by Range functionality as bonus
  • All Keys are sorted. Ability to set your Custom Order of Keys
  • Predictable behaviour even in worst case: smoothly consuming memory, almost constantly latency on insert/lookup
  • Prefix Compression helps reduce memory when keys have the same prefix: urls, file paths, words etc.
  • Serialize/Deserialize from/to file at a speed close to the max speed of the hard drive
  • Fair Delete operation with smoothly dismantling each Key. Dead fibres are used for insert new keys, so structure perfect works in massive insert/delete scenarios.

Build Library and Benchmarks

Prerquisites

The following need to be installed and configured on development box:

  • C++ development tools.
  • CMake.

Build on Linux and Mac

mkdir build
cmake -Bbuild
cd build && make

To build release version of the library and benchmarks run instead:

cmake -Bbuild -DCMAKE_BUILD_TYPE=Release

The library (libharray.a) and benchmark application (HArrayBenchmark) will be created in build folder.

Build on Windows

md build
cmake -Bbuild
msbuild build\HArray.sln

To build release version run:

msbuild build\HArray.sln /property:Configuration=Release

The benchmark application HArrayBenchmark.exe together with static library libharray.lib will be in build\Debug or build\Release folder.

Why we love Trie ? Because it has much more functionality and stability than Hashtables and much more faster than Binary Trees. Let's compare properties:

alt tag


Trie ? I heard about Trees and Hastables but don't know anything about Trie

Examples

Initialize container

#include "HArray.h"

HArray ha;
ha.init(); //ha.init(24) set your custom capacity for big datasets

Insert a key

uint32 key[] = { 10, 20, 30, 40 };
uint32 keyLen = sizeof(key) / 4; //in key segments
uint32 value = 1;

ha.insert(key, keyLen, value);

Get value by key. Will return 0 if key is not found

uint32 value;
if(ha.getValueByKey(key, keyLen, value))
{
   printf("%d", value)
}
else
{
   //key is not found
}

Get all keys by range from min key to max key.

HArrayPair pairs[5];
uint32 pairsSize = 5;

uint32 minKey[] = { 10, 10 };
uint32 minKeyLen = sizeof(minKey) / 4; //in key segments

uint32 maxKey[] = { 20, 20 };
uint32 maxKeyLen = sizeof(maxKey) / 4; //in key segments

uint32 count = ha.getKeysAndValuesByRange(pairs, pairsSize, minKey, minKeyLen, maxKey, maxKeyLen);
for (uint32 i = 0; i < count; i++)
{
   uint32* key = pairs[i].Key;
   uint32 keyLen = pairs[i].KeyLen;

   uint32 value = pairs[i].Value;
   
   //here your code
}

Scan all keys through visitor

bool visitor(uint32* key, uint32 keyLen, uint32 value, uchar8 valueType, void* pData)
{
   //handle founded key here
   // ...

   //return true to continue scaning or false otherwise
   return true;
}

//Start scanning

void* pData = 0; //useful data in context

ha.scanKeysAndValues(&visitor, pData);

Serialize/Deserialize containter from/to file

ha.saveToFile("c:\\dump");

ha.loadFromFile("c:\\dump");

Check if container has part of key

uint32 key[] = { 10, 20, 30 };
uint32 keyLen = sizeof(key) / 4; //in key segments

if(ha.hasPartKey(key, keyLen))
{
   //code here
}

Set specific comparator to redefine order of keys in the container.

float key[] = { 10.0, 20.0, 30.0 };
uint32 keyLen = sizeof(key) / 4; //in key segments

uint32 value = 1;

//Set float comparator for right sorting
//Another options: setStrComparator, setInt32Comparator, setUInt32Comparator 
//or define your custom comparator through setCustomComparator
ha.setFloatComparator();

ha.insert((uint32*)key, keyLen, value);

Delete Key and Value from container

ha.delValueByKey(key, keyLen);

How to Run

git clone github.com/Bazist/HArray.git
cd HArray/HArray
make
./HArray

License

The code is licensed under the GNU General Public License v3.0, see the LICENSE file for details.

Releases

No releases published

Packages

No packages published

Languages