This article tries to describe the practical techniques of searching by a HAS_MANY relation using ActiveRecord of Yii 2.0.
In Yii 1.1, we sometimes got lost trying to search by a HAS_MANY relation. The wiki article Drills : Search by a HAS_MANY relation in Yii 1.1 was an effort to offer some practical techniques of searching by a HAS_MANY relation in Yii 1.1.
Now, we have Yii 2. Let's examine what we can do with the new ActiveRecord of Yii 2 when we have to solve the same tasks of searching by a HAS_MANY relation.
Relation
Two entities are sometimes connected with a relation of 1:N. Or we may say that 1:N is the only possible relation between 2 entities as long as we are in the RDB world. 1:1 relation is just a particular kind of 1:N where N is always assumed to be 1 at the maximum. And N:N relation can be considered as a combination of two 1:N relations.
Yii supports this 1:N relation in ActiveRecord as HAS_ONE and HAS_MANY relations. 1:N relation seen from the side of N is HAS_ONE, and from the side of 1 it is HAS_MANY.
Now, let's construct an example of 1:N relation.
Example of HAS_MANY
Think about blog posts and their authors. It's a relation of "An Author has many Posts" and "A Post has an Author" at the same time. This pair of relations is represented with the following code:
/** * Author model * @property integer $id * @property string $name the author's name ... */ class Author extends \yii\db\ActiveRecord { ... /** * @return \yii\db\ActiveQuery */ public function getPosts() { return $this->hasMany(Post::className(), ['author_id' => 'id']); } ...
/** * Post model * @property integer $id * @property integer $author_id FK to the author's id * @property string $title the title of the post ... */ class Post extends \yii\db\ActiveRecord { ... /** * @return \yii\db\ActiveQuery */ public function getAuthor() { return $this->hasOne(Author::className(), ['id' => 'author_id']); } ...
Note that a relation in Yii 2 is a getter method returning ActiveQuery.
We are going to solve the possible use cases of searching with this example.
Task #1
Show all posts that has a word in post title
I want to start with an easy one. Let's retrieve all the posts that has a certain word in their title.
// get all the posts that has a keyword in their title $posts = Post::find() ->where(['like', 'title', $searchword]) ->all(); // show the results foreach($posts as $post) { echo "Title = {$post->title}\n"; }
Ah, it was too easy. And no relation is involved in this task.
OK, let's move on to the next.
Task #2
Show all posts that has a certain word in post title, with their authors
Now we have to retrieve the authors' name, too.
// get all the posts that has a keyword in their title, with their authors $posts = Post::find() ->where(['like', 'title', $searchword]) ->with('author') ->all(); // show the results foreach($posts as $post) { echo "Title = {$post->title}\n"; echo "Author = {$post->author->name}\n"; }
Well, you may say "You didn't have to do it. Just adding a line to echo $post->author->name to the code of the 1st task would have been enough."
Yes, you are right. Post model has the author relation, so all you have to do to retrieve its Author is just accessing $post->author. How convenient is it! It's so-called lazy loading approach.
But the lazy loading approach has a drawback in this case, because you have to execute one query for retrieving an array of Posts, and every one query per each Post for retrieving its author. It will end up in 1+N queries to be executed.
Here we are retrieveing Posts and their Authors at the same time by using yii\db\ActiveQuery::with(). It enables you to do the job with only 2 queries. Just after the 1st query has retrieved the main models, the 2nd one will get all of the related models at once. This is so-called eager loading approach.
The eager loading is preferrable in this particular case, because it's more effective. But, you have to note, it is not always so.
Task #2-B
Show all posts with their authors that has a certain word in the author's name
This is a variation of the task #2. You have to search by an attribute in the related model, not in the main model.
// get all the posts that has a keyword in their author's name, with their authors $posts = Post::find() ->joinWith('author') ->where(['like', 'author.name', $searchword]) ->all(); // show the results foreach($posts as $post) { echo "Title = {$post->title}\n"; echo "Author = {$post->author->name}\n"; }
We join the author table using yii\db\ActiveQuery::joinWith(), because we have to search the main model by an attribute of the related model. Note that we may need to disambiguate the column name with the table name for the parameter of where.
The more important thing to note is that the joining tables using joinWith will not reduce the number of queries in the eager loading scenario. Yii will still use 2 queries, one for the main model and the other for the related model. Yii 2 will always use the separated queries for the main model and the related models.
So, you would not be able to make the code more efficient by replacing with with joinWith in the answer for the task #1.
We join the related tables using joinWith not because we expect good performance, but because we have to access a column of the related table for searching the main table.
And joinWith will cause ActiveRecord to choose the eager loading approach by default, but you may suppress the loading of the related models by specifying the optional parameter $eagerLoading as false.
Huh? Who said "It's just a HAS_ONE relation. Yeah, it's simple. I know."?
Yes, you are right, definitely. Let's move on to the next task where we will deal with HAS_MANY.
Task #3
Show all authors who have a post that has a certain word in its title
Now we will retrieve the authors, not the posts.
// get all the authors who have a post that has a certain word in its title $authors = Author::find() ->joinWith('posts', false) // will not get the related models ->where(['like', 'post.title', $searchword]) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; }
We join the posts relation using joinWith because we need post.title in the searching, while we set $leagerLoading to false because we don't need the posts to be retrieved.
Task #4
Show all authors who have a post that has a certain word in its title, with all of their posts
// get all the authors who have a post that has a certain word in its title, with all of their posts $authors = Author::find() ->joinWith('posts') // will eagerly load the related models ->where(['like', 'post.title', $searchword]) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; foreach($author->posts as $post) { echo "Title = {$post->title}\n"; } }
Don't forget that the query for the main model and those for the related models are separated. We add a condition using post.title in where part in order to filter the main models, but it doesn't affect the loading of the related models. The related models are loaded without any condition.
Well, then, what about the next task? It's a variation of the above.
Task #4-B
Show all authors who have a post that has a certain word in its title, with all of their relevant posts
In fact, this is a little more confusing than the above.
One solution is this:
// get all the authors who have a post that has a certain word in its title $authors = Author::find() ->joinWith('posts', false) // won't get posts ->where(['like', 'post.title', $searchword]) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; // get all the posts that have a certain word in its title $posts = $author->getPosts()->where(['like', 'post.title', $searchword])->all(); foreach($posts as $post) { echo "Title = {$post->title}\n"; } }
We join the posts relation using joinWith because we need post.title in the searching, while we set $leagerLoading to false because we don't want to load the posts at that point. And then we perform the lazy loading of the related posts with on-the-fly conditions for each author afterwards.
You could specify an on-the-fly condition to the relation when you eagerly load the related models:
// get all the authors who have a post that has a certain word in its title $authors = Author::find() ->joinWith('posts', false) // this is for the main models ->where(['like', 'post.title', $searchword]) ->with([ // this is for the related models 'posts' => function($query) use ($searchword) { $query->where(['like', 'post.title', $searchword]); }, ]) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; foreach($posts as $post) { echo "Title = {$post->title}\n"; } }
And finally you can do it like the following:
// get all the authors who have a post that has a certain word in its title $authors = Author::find() ->joinWith([ 'posts' => function($query) use ($searchword) { $query->where(['like', 'post.title', $searchword]); }, ]) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; foreach($posts as $post) { echo "Title = {$post->title}\n"; } }
In this way, you can use a customized relation for both the searching of the main models and the loading of the related models.
Let's move on to the next. It's a bit tough, though.
Task #5
Show top 5 authors in the order of the name who has at least one post that has a certain word in post title, with his/her all posts
OK, so we need to add "LIMIT" and "ORDER". Going to try with this one.
Trial
$authors = Author::find() ->joinWith('posts') ->where(['like', 'post.title', $searchword]) ->orderBy(['author.name' => SORT_ASC]) ->limit(5) ->all(); // show the results foreach($authors as $author) { echo "Author = {$author->name}\n"; foreach($author->posts as $post) { echo "Title = {$post->title}\n"; } }
You will notice that there is something strange in the output results. For example, the results may look like the following:
[search word = foo]
Author = Andy
Post = Don't use foo
Post = Use yoo for foo
Post = Don't use bar
Post = Use yar for bar
Author = Ben
Post = foo is great
Post = I love foo
Post = I also love bar
Author = Charlie
Post = What's foo?
Post = What's bar?
[end]
We could get only 3 authors, instead of 5 that is what we wanted.
Why did it happen? It was because the SQL used to retrieve the main models had been something like the following:
SELECT * from author LEFT JOIN post on author.id = post.author_id WHERE post.title LIKE '%foo%' ORDER BY author.name LIMIT 5
Notice that LIMIT 5 is applied to the number of the rows of an virtual table that is the result of joining author table with post table. The resulted rows may have less unique authors than the limit, because an author may have many posts.
In Yii 1.1, we could use GROUP BY trick, though it was MySQL specific. It can also be used in Yii 2:
Solution #1
$authors = Author::find() ->joinWith('posts') ->where(['like', 'post.title', $searchword]) ->groupBy('author.id') ->orderBy(['author.name' => SORT_ASC]) ->limit(5) ->all();
In this way we could remove the redundant rows by grouping the results by author's ID.
You may say "Why don't you use DISTINCT? It might be cleaner and more compatible than a dirty trick." Seems fairly reasonable. Let's give it a try:
Solution #2
$authors = Author::find() ->distinct() ->joinWith('posts') ->where(['like', 'post.title', $searchword]) ->orderBy(['author.name' => SORT_ASC]) ->limit(5) ->all();
Huh, done!
We are going to the last task, feeling a bit confused by the unexpected easiness.
Task #6
Show top 5 authors in the order of name who has at least one post that has a certain word in post title, with his/her relevant posts
Now it's a simple task. You just have to apply the same filter both for the searching of the main models and the loading the related models.
Example of a Solution
$authors = Author::find() ->distinct() ->joinWith([ 'posts' => function($query) use ($searchword) { $query->where(['like', 'post.title', $searchword]); }, ]) ->orderBy(['author.name' => SORT_ASC]) ->limit(5) ->all();
Thus we finish the drills in an anticlimactic end.
Summary
Fantastic, isn't it?
Reference
Please refer to the database sections of the guide.