Simple friendship table

rubyonrails-talk
1

Tables: Users (id, name, sex) Friendships (user_id, friend_id, status)

If user(1) and user(2) are friends shoud I add two rows into friendships table?

1, 2, "best friends" 2, 1, "best friends"

Or is it enough to have one of these rows? Select * from friendships where user_id = 2 or friend_id = 2 ???

But if I have only one friendship row and I need to list all friens of user(2) How can I do it? I don't know which one "user_id" or "friend_id" is needed "data".

Of course, I can check: IF search_id != friend_id print (friend_id).user.name ELSE print (user_id).user.name But it is slow and stupid..

2

Hi, you only need one entry per friendship (where, for extra functionality, the first id could always be the one who initiated the friendship). Friendships table could be something like have (uid1, uid2, status). Probably put first two as primary key.

To find ids of all friends of user 1 you would do :

(SELECT uid2 AS uid, status FROM friendships WHERE uid1 = 1) UNION (SELECT uid1,status FROM friendships WHERE uid2 = 1)

This would return a set of rows with columns (uid, status) for all friends of user 1. If you have:

Uid1,uid2,status 1,2,"best friend" 1,3,"partner" 3,2,"lover" 3,7,"barely met" 4,1,"enemy" 7,1,"buddy"

The above select would return:

Uid, status 2,"best friend" 3,"partner" 4,"enemy" 7,"buddy"

Regards

Binni

-----Oprindelig meddelelse-----

3

Just a small addition to my previous email. If you also want to see, from the result, whether user 1 initiated the friendship (i.e. where uid1=1), you could use the following SQL statement:

(SELECT uid2 AS uid, status, 1 AS initiated FROM friendships WHERE uid1 = 1) UNION (SELECT uid1,status,0 FROM friendships WHERE uid2 = 1)

This would give you the following result (using same data as below):

Uid, status, initiated 2,"best friend",1 3,"partner",1 4,"enemy",0 7,"buddy",0

Where initiated=1 are the ones with uid1=1 and initiated=0 are the ones with uid2=1

-----Oprindelig meddelelse-----

4

Brynjolfur Thorvardsson wrote in post #1037658:

(SELECT uid2 AS uid, status FROM friendships WHERE uid1 = 1) UNION (SELECT uid1,status FROM friendships WHERE uid2 = 1)

It works fine, but it is slow :frowning:

Only one entry per friendship (1 500 000 rows in database)

SELECT with UNION: 1.2108 sek

Dublicate entryes ( total 3 000 000 rows in database)

SELECT without UNION: 0.0007 sek

5

(SELECT uid2 AS uid, status FROM friendships WHERE uid1 = 1) UNION (SELECT uid1,status FROM friendships WHERE uid2 = 1)

How to write same in Rails 3 Active Record Query format?

6

It would probably run a lot faster if you had a view and separate indexes on the two uid columns. I haven't got access to my MySQL at the moment, so i can't check the syntax for creating a view but you should be able to find that yourself. Something like:

CREATE VIEW friendshipview AS (<the union select statement>)

Creating separate indexes on the columns used in the select statements should speed things up quite a bit, an index should speed up by a factor of 10^2 or 3. If you have a very active database you might also consider rebuilding the index regularly, with OPTIMIZE TABLE <table name>.

-----Oprindelig meddelelse-----

7

A simple way would be to just call <model>.find_by_sql(<sql query>)

On the other hand, if you create a view in MySQL to handle the query, say the view is called "friends", then you should be able to create a model like this:

app/models/friend.rb

class Friend < ActiveRecord::Base end

Obviously, depending on the view and tables behind it, you might want to make your new model read only. That can apparently be done like this (I haven't actually tested this!)

class Post < ActiveRecord::Base

  def readonly?     return true   end

end

There are probably lots of other ways of wrapping the result as an object, but basically I don't think you can avoid passing the SQL UNION statemenet as is. There might be a way of calling two separate SQL statements, store the results in separate variables and then adding the two:

@var1 = @friends.where(<select 1>) @var2 = @friends.where(<select 2>)

@friendsfound = @var1 + @var2

Obviously, the returned column names would have to match, and I'm not at all sure this would do what it looks like doing!

-----Oprindelig meddelelse-----

8

First setup the model relationships so that you are not messing about with id values all the time, something like this railscast [1]. Then if you want all the friends of a user it is simply. user.friends + user.inverse_friends

Colin

[1] #163 Self-Referential Association - RailsCasts

9

Brynjolfur Thorvardsson wrote in post #1037995:

@var1 = @friends.where(<select 1>) @var2 = @friends.where(<select 2>)

@friendsfound = @var1 + @var2

tmp1 = Friendships.includes(:users).select("uid1 as user_id", status).where("uid2 = (?)", @user.id) tmp2 = Friendships.includes(:users).select("uid2 as user_id", status).where("uid1 = (?)", @user.id) @friends = tmp1 + tmp2

Tested, it works, but the only problem is ordering.

How to order:

.order("FIELD( friendships.status, 'lover', 'best friend', 'partner', 'enemy' ), users.name")

10

Because you are adding the two collections together you will have to sort it afterwards. Have a look at Array#sort and sort_by.

Colin