#graph db

I am pretty excited about Google's semantic search -  connecting dots between metadata is something I have been working on for last few months.

An article in a German publication mentions (according to Google translator) that sones GraphDB is up for sale:

The administrator of sones GmbH Hartig, Dr. Oliver lawyer, said that the graph database of insolvent sones GmbH will be sold.

Anyone interested?

NoSQL分支Graph Database中. Neo4j算是比較熱門的. Von曾經花時間瞭解Neo4j , Allegro , FlockDB 等幾款.. 最後覺得Neo4j無論是在支援度,文件,速度... 等 都對Ruby使用者比較友善. 如果架構確定只需要一層的關係搜尋,並且不易更改. 可以考慮使用FlockDB. 或像EventBrite 使用 Graph on Mongo 在Ruby要使用Neo4j非常簡單. 只需要兩個步驟 : 1. 裝好 jruby. 2. gem install neo4j  

裝好之後, 我們便可以開始寫些簡單的Code. 以一個簡單的圖形關係如下 :

  在這圖形中. 若是以MySQL該如何做關係搜索? 頭大吧! 在Neo4j中可是非常簡單的.   定義資料欄位

首先我們先定義幾個資料欄位

class Member include Neo4j::NodeMixin property :member_id property :name index :member_id index :name end

在Neo4j中若單純使用關係而不做特定搜尋,是不需要index的. 一旦我們需要對某個欄位資料做搜尋,好比搜尋會員名字有含John的. 則在定義欄位時需要再加上index.  

增加及關係建立 定義好資料欄位後便是實際的增加資料和關係.

Von用一些最簡單的新增方法讓大家都能看懂.

Neo4j::Transaction.run do one=Member.new(:member_id=>1,:name=>"Member One") two=Member.new(:member_id=>2,:name=>"Member Two") three=Member.new(:member_id=>3,:name=>"Member Three") four=Member.new(:member_id=>4,:name=>"Member Four") five=Member.new(:member_id=>5,:name=>"Member Five") six=Member.new(:member_id=>6,:name=>"Member Six") seven=Member.new(:member_id=>7,:name=>"Member Seven") one.outgoing(:has_friend) << two one.outgoing(:has_friend) << three one.outgoing(:has_friend) << four two.outgoing(:has_friend) << three two.outgoing(:has_friend) << six three.outgoing(:has_friend) << four four.outgoing(:has_friend) << five four.outgoing(:has_friend) << six five.outgoing(:has_friend) << six six.outgoing(:has_friend) << seven end

搜尋 現在資料建好了. 再來就是做搜尋. Neo4j搜尋的方式是使用Lucene.

簡單的搜尋方式可以是

Member.find(:name=>"John")

或是

Member.find("name like John")

而所產生出來的結果與ActiveRecord一樣會是陣列. 如果沒有結果則是空陣列. 至此. 我們已經可以搜尋出一些結果. 下面是完整的程式碼 :   原始碼 :

require 'rubygems' require 'neo4j' puts "start" class Member include Neo4j::NodeMixin property :member_id property :name index :member_id index :name def add_friend(name) self.outgoing(:has_friend) << Neo4j::Node.new(:name => name) end end puts "object" Neo4j::Transaction.run do one=Member.new(:member_id=>1,:name=>"Member One") two=Member.new(:member_id=>2,:name=>"Member Two") three=Member.new(:member_id=>3,:name=>"Member Three") four=Member.new(:member_id=>4,:name=>"Member Four") five=Member.new(:member_id=>5,:name=>"Member Five") six=Member.new(:member_id=>6,:name=>"Member Six") seven=Member.new(:member_id=>7,:name=>"Member Seven") one.outgoing(:has_friend) << two one.outgoing(:has_friend) << three one.outgoing(:has_friend) << four two.outgoing(:has_friend) << three two.outgoing(:has_friend) << six three.outgoing(:has_friend) << four four.outgoing(:has_friend) << five four.outgoing(:has_friend) << six five.outgoing(:has_friend) << six six.outgoing(:has_friend) << seven end puts "rels" Neo4j::Transaction.run do one=Member.find(:member_id=>1) if one && one.size>0 one.first.outgoing(:has_friend).depth(:all).each{|friend| puts friend.name} else puts "no member found" end end puts "search"

    --Von所寫關於Neo4j的其他文章 Neo4j 進階搜尋 Neo4j HA觀念及實作