Part of my application is private messages between 2 users.
I have a converstations table that has user1 and user2 foreign keys.
Now I wanna have an eloquent function in the User model where I can get his conversations. But a user could be user1 or user2 in any given conversation.
This function for example will only account to conversations where I happen to be user1.
public function conversations()
{
return $this->hasMany('App\Conversation', 'user1');
}
So how can I do it?
A suggestion into your approach would be to store the users of a Conversation in a pivot table. So a Conversation can have more than, in your case, two users. This will lead to easier queries and more expension options. In that case you would just have to do the following:
return $this->belongsToMany(Conversation::class);
Check the docs for more info.
Addition to comment
If you really need this table to have those two columns people would normally do the following:
return $this->hasMany(Conversation::class, 'user1')->orWhere('user2', $this->id);
However since this is an orWhere it could get you incorrect records. This query is executed:
select * from "conversations" where "conversations"."user1" = 1 and "conversations"."user1" is not null or "user2" = 1
But there I have found another way. There is a noConstraints method on the Relation abstract class , from which all relation classes extend, that allows you to modify the whole query:
use Illuminate\Database\Eloquent\Relations\HasMany;
return HasMany::noConstraints(function () {
return $this->hasMany(Conversation::class)
->where('user1', $this->id)
->orWhere('user2', $this->id);
});
This will result in the query what you are looking for:
select * from "conversations" where "user1" = 1 or "user2" = 1
I am not sure how this will effect eagerloading and such.
Still I suggest having a pivot table to keep your database/application flexible.
public function conversations()
{
if(test_for_user1_or_user2())
{
return $this->hasMany(converstations ::class,'user1_id');
}
return $this->hasMany(converstations ::class,'user2_id');
}
Related
I have a Pivot table thats used to join two other tables that have many relations per hotel_id. Is there a way I can eagerload the relationship that pulls the results for both tables in one relationship? The raw SQL query, works correctly but when using belongsToMany the order is off.
Amenities Pivot Table
id
hotel_id
distance_id
type_id
Distance Table
id
name
Type Table
id
name
RAW Query (This works fine)
SELECT * FROM amenities a
LEFT JOIN distance d ON a.distance_id = d.id
LEFT JOIN type t ON a.type_id = t.id WHERE a.hotel_id = ?
My "Hotels" Model is using belongsToMany like so
public function distance() {
return $this->belongsToMany('Distance', 'amenities', 'hotel_id', 'distance_id');
}
public function type() {
return $this->belongsToMany('Type', 'amenities', 'hotel_id', 'type_id');
}
This outputs the collection, but they are not grouped correctly. I need to loop these into select fields side by side as entered in the pivot table, so a user can select a "type" and the "distance", but the order is off when using the collection. The raw query above outputs correctly.
Hotels::where('id','=','200')->with('distance', 'type')->take(5)->get();
Ok Solved it. So apparently you can use orderBy on your pivot table. Incase anyone else has this issue this is what I did on both relationships.
public function distance() {
return $this->belongsToMany('Distance', 'amenities', 'hotel_id', 'distance_id')->withPivot('id')->orderBy('pivot_id','desc');
}
public function type() {
return $this->belongsToMany('Type', 'amenities', 'hotel_id', 'type_id')->withPivot('id')->orderBy('pivot_id','desc');
}
It's not really a great practice to include other query building steps in the relationship methods on your models. The relationship method should just define the relationship, nothing else. A cleaner method is to apply eager load constraints. (scroll down a bit) Consider the following.
Hotels::where('id', 200)->with(array(
'distance' => function ($query)
{
$query->withPivot('id')->orderBy('pivot_id','desc');
},
'type' => function ($query)
{
$query->withPivot('id')->orderBy('pivot_id','desc');
},
))->take(5)->get();
If you find that you are eagerly loading this relationship in this way often, consider using scopes to keep things DRY. The end result will allow you to do something like this.
Hotels::where('id', 200)->withOrderedDistance()->withOrderedType()->take(5)->get();
P.S. Your models should be singular. Hotel, not Hotels. The model represents a single record.
Solved by using ->withPivot('id')->orderBy('pivot_id','desc');
Posted answer in the question.
So I am having some trouble figuring out how to do a feed style mysql call, and I don't know if its an eloquent issue or a mysql issue. I am sure it is possible in both and I am just in need of some help.
So I have a user and they go to their feed page, on this page it shows stuff from their friends (friends votes, friends comments, friends status updates). So say I have tom, tim and taylor as my friends and I need to get all of their votes, comments, status updates. How do I go about this? I have a list of all the friends by Id number, and I have tables for each of the events (votes, comments, status updates) that have the Id stored in it to link back to the user. So how can I get all of that information at once so that I can display it in a feed in the form of.
Tim commented "Cool"
Taylor Said "Woot first status update~!"
Taylor Voted on "Best competition ever"
Edit #damiani
So after doing the model changes I have code like this, and it does return the correct rows
$friends_votes = $user->friends()->join('votes', 'votes.userId', '=', 'friend.friendId')->orderBy('created_at', 'DESC')->get(['votes.*']);
$friends_comments = $user->friends()->join('comments', 'comments.userId', '=', 'friend.friendId')->orderBy('created_at', 'DESC')->get(['comments.*']);
$friends_status = $user->friends()->join('status', 'status.userId', '=', 'friend.friendId')->orderBy('created_at', 'DESC')->get(['status.*']);
But I would like them all to happen at once, this is because mysql sorting thousands of records in order is 100x faster then php taking 3 lists, merging them and then doing it. Any ideas?
I'm sure there are other ways to accomplish this, but one solution would be to use join through the Query Builder.
If you have tables set up something like this:
users
id
...
friends
id
user_id
friend_id
...
votes, comments and status_updates (3 tables)
id
user_id
....
In your User model:
class User extends Eloquent {
public function friends()
{
return $this->hasMany('Friend');
}
}
In your Friend model:
class Friend extends Eloquent {
public function user()
{
return $this->belongsTo('User');
}
}
Then, to gather all the votes for the friends of the user with the id of 1, you could run this query:
$user = User::find(1);
$friends_votes = $user->friends()
->with('user') // bring along details of the friend
->join('votes', 'votes.user_id', '=', 'friends.friend_id')
->get(['votes.*']); // exclude extra details from friends table
Run the same join for the comments and status_updates tables. If you would like votes, comments, and status_updates to be in one chronological list, you can merge the resulting three collections into one and then sort the merged collection.
Edit
To get votes, comments, and status updates in one query, you could build up each query and then union the results. Unfortunately, this doesn't seem to work if we use the Eloquent hasMany relationship (see comments for this question for a discussion of that problem) so we have to modify to queries to use where instead:
$friends_votes =
DB::table('friends')->where('friends.user_id','1')
->join('votes', 'votes.user_id', '=', 'friends.friend_id');
$friends_comments =
DB::table('friends')->where('friends.user_id','1')
->join('comments', 'comments.user_id', '=', 'friends.friend_id');
$friends_status_updates =
DB::table('status_updates')->where('status_updates.user_id','1')
->join('friends', 'status_updates.user_id', '=', 'friends.friend_id');
$friends_events =
$friends_votes
->union($friends_comments)
->union($friends_status_updates)
->get();
At this point, though, our query is getting a bit hairy, so a polymorphic relationship with and an extra table (like DefiniteIntegral suggests below) might be a better idea.
Probably not what you want to hear, but a "feeds" table would be a great middleman for this sort of transaction, giving you a denormalized way of pivoting to all these data with a polymorphic relationship.
You could build it like this:
<?php
Schema::create('feeds', function($table) {
$table->increments('id');
$table->timestamps();
$table->unsignedInteger('user_id');
$table->foreign('user_id')->references('id')->on('users')->onDelete('cascade');
$table->morphs('target');
});
Build the feed model like so:
<?php
class Feed extends Eloquent
{
protected $fillable = ['user_id', 'target_type', 'target_id'];
public function user()
{
return $this->belongsTo('User');
}
public function target()
{
return $this->morphTo();
}
}
Then keep it up to date with something like:
<?php
Vote::created(function(Vote $vote) {
$target_type = 'Vote';
$target_id = $vote->id;
$user_id = $vote->user_id;
Feed::create(compact('target_type', 'target_id', 'user_id'));
});
You could make the above much more generic/robust—this is just for demonstration purposes.
At this point, your feed items are really easy to retrieve all at once:
<?php
Feed::whereIn('user_id', $my_friend_ids)
->with('user', 'target')
->orderBy('created_at', 'desc')
->get();
I have two tables 1)users
{ id, password }
2)expertise { id, expertise}
the relationship I have is
Models
Expertise.php
function User()
{
$this->hasOne('Expertise');
}
User.php
function Expertise()
{
$this->hasOne('User');
}
So how can I query using Eloquent to get the first 10 users with a certain expertise?
I want to join users.id = expertise.id and get the first 10 people with a specified expertise (Where clause).
Beginner to laravel, I've checked other sources but was not successful
Right now you are having a problem with the way that you modeled your data. If you have a one-to-one relationship the best practice to model it is to have one entity store the id of the other. The Laravel convention for this is to have a column named <model>_id:
Users
| id | password |
Expertises
| id | expertise | user_id |
Then in your models you can do this:
Models
Expertise.php
class Expertise extends Eloquent
{
public function User()
{
// because expertise has a column user_id
// expertise belongs to user
return $this->belongsTo('User');
}
}
User.php
class User extends Eloquent
{
public function Expertise()
{
// because expertise is the one with the column
// user_id, user has one expertise
return $this->hasOne('Expertise');
}
}
The Query
After you have all this set up, to be able to query the first 10 users with a certain expertise you can do this.
$users = User::whereHas('Expertise', function($q)
{
$q->where('expertise', '=', <expertise you are looking for>)
})
->take(10)
->get();
To get a further reading in querying relationships in Laravel please take a look at this:
Laravel - Querying Relationships
Keep in mind
keep in mind that the tables name must be plural, if not then you should specify the name of the table inside the model:
protected $table = 'expertise';
I have been trying to do some queries and getting a count on related tables using eloquent.
Tables:
requests
contact (belongs to requests)
history (belongs to contact)
As such X number of requests each have Y number of contacts which in term each have Z number of histories
Using sql I can do something like this to get all the counts.
SELECT
id,
(
SELECT count(contact.id)
FROM contact
WHERE contact.requests_id = requests.id
) AS n_contact,
(
SELECT count(history.id)
FROM contact INNER JOIN history ON (history.contact_id = contact.id)
WHERE contact.requests_id = requests.id
) AS n_history
FROM requests;
But I am a bit lost when using eloquent to build queries. If for instance I was selecting all contacts for a given request at what point would I join/count the history? Or do I need to add in some accessor's into the relevant Models for these 3 tables?
public function getAllContacts($id) {
return Requests::where('requests.id', '=', $id)
->join('requests', 'contact.requests_id', '=', 'requests.id')
->select('contact.*', 'requests.name');
->get();
}
Thanks in advance for any help.
You can use helper relation for this, if you'd like to use Eloquent instead of manual joins:
// Request model
public function contactsCount()
{
return $this->hasOne('Contact')->selectRaw('request_id, count(*) as aggregate')->groupBy('request_id');
}
public function getContactsCountAttribute()
{
if ( ! array_key_exists('contactsCount', $this->relations)) $this->load('contactsCount');
return $this->getRelation('contactsCount')->aggregate;
}
The same would go for Contact model towards History model.
For counting far relation (Request -> History) you can use hasManyThrough relation with a little adjustment.
This way you can eager load those aggregates for multiple models without n+1 issue, nice and easy:
$requests = Request::with('contactsCount', 'contacts.historyCount')->get();
// 1 query for reuqests, 1 query for contacts count and 2 queries for contacts.historyCount
// example output
$requests->first()->contactsCount; // 17
$requests->first()->contacts->first()->historyCount; // 5
/* Make Relation in the Request Model */
public function contacts()
{
return $this->hasMany('App\Model\Contact', 'request_id', 'id');
}
/* use withCount() to get the total numner of contacts */
public function getAllContacts($id) {
return Requests::with('contacts')
->withCount('contacts')
->find($id);
}
I have a Group model that belongsToMany Contacts, and Contact belongsToMany Groups, usually a group could contain about 300,000 contact, and a contact could belong to 2 or 3 groups, and I want a function that retrieves all "unique" contacts in a number of given groups, so I tried
$contacts = Contact::With(array('groups'=>function($query) use ($groups)
{
$query->whereIn('groups.id' , $groups);
}
))->get();
It worked but the problem is that I discovered the sql generated uses where contact_id IN ( ALL CONTACTS' IDs). So, if I'm retrieving 100,000 contacts it will have 100,000 of contact ids in the where in clause.
Then I ended up using this
foreach($groups as $group)
{
$contacts = array_unique(array_merge($contacts , $group->contacts()->lists('email'))) ;
}
But I'm still concerned about performance. I don't know how lists works and if it is a good idea to array_unique and array_merge on arrays containing hundreds of thousands of email strings?
You said:
usually a group could contain about 300,000 contact, and a contact
could belong to 2 or 3 groups
According to this your relationship should be many-to-many and both Group and Contact should use belongsToMany to make relationship between them using a pivot table contact_group:
// Group model
public function contacts()
{
return $this->belongsToMany('Contact');
}
// Contact model
public function groups()
{
return $this->belongsToMany('Group');
}
The contact_group pivot table could be like this:
id | group_id | contact_id
If you have a setup like this then you'll be able to do this:
$groups = Group::has('contacts')
->with('contacts')
->whereIn('id', [1,2,3]) // group ids, could be any field
->get();
$uniqueContacts = $groups->map(function($group) {
return $group->contacts->lists('email');
})->flatten()->toBase()->unique();
// You got all unique contacts in $uniqueContacts
dd($uniqueContacts); // all unique contacts
Don't know how fast it'll work on 100,000 contacts but should work better.
Update: Could be other (maybe better) ways by joining DB tables and querying from DB scope.