13 KiB
Nesting Models
Objectives
Currently, we only have two unrelated models: people and locations. Let's add the functionality so that whenever we have JSON that represents a person, it will contain a home property, which is the associated location object:
{
"id":4,
"name":"Bob",
"age":25,
"home": {
"id":2,
"street": "123 Fake Street",
"city": "Aurora",
"state": "NY"
}
}
Also, when we have a location, it will contain an array of people objects:
{
"id":2,
"street": "123 Fake Street",
"city": "Aurora",
"state": "NY",
"inhabitants":[
{
"id":4,
"name":"Bob",
"age":25
},
{
"id":7,
"name":"Sally",
"age":74
},
]
}
This will be a one-to-many relationship.
Add home_id to people
First, go into psql and add a home_id INT column to the people table:
ALTER TABLE people ADD COLUMN home_id INT;
Now we can give people homes like so:
UPDATE people SET home_id = 2 WHERE id = 4;
The previous code will find the row with an id of 4 and set the home_id column to 2.
Update People SQL
The first thing we want to do is to update the SQL code in the People class so that each row from the people table is joined onto a corresponding row from the locations table. Once we have the additional columns for each row in the people table, we can use that information to create a home for each Person object. Try the following in psql:
SELECT * FROM people JOIN locations ON people.home_id = locations.id;
The above SQL statement will only show rows where the people have a value in the home_id column. This is because Postgres is attempting to create rows by matching rows in the people table with rows in the locations table, based on whether people.home_id = locations.id. When it encounters a row from the people table where the home_id column is NULL, it can't find any rows from the locations table to match it with, since no rows from locations have an id column of NULL. Since no matches can be found, it doesn't include that row from people. We need to perform the join and then add any missing rows from people:
SELECT * FROM people LEFT JOIN locations ON people.home_id = locations.id;
Now that we have all the people rows, we're ready to plug that into our php. Alter the pg_query code in our People::all() function:
$results = pg_query("SELECT * FROM people LEFT JOIN locations ON people.home_id = locations.id");
If you view http://localhost:8888/people, you'll see something funky. Some people have 0 ids, and other rows have unexpected id columns. The reason for this can be discovered by looking back in psql at our previous LEFT JOIN statement results. You'll notice there are two id columns, one for the people table and one for locations table. When PHP attempts to convert a row into an object, it reads the people table's id column and creates an id property for the object. It then goes through, adding name, age, and home_id properties. It then reaches the id column for the locations table and overwrites the id property on the $row_object with the value from the id column of the locations table.
To fix this, we can alter our SQL statement to rename one or both of the id columns. Let's rename the id column from the locations table.
SELECT
people.*,
locations.id AS location_id,
locations.street,
locations.city,
locations.state
FROM people
LEFT JOIN locations
ON people.home_id = locations.id;
Now if we change our SQL statement in PHP, we'll see better results:
$results = pg_query("SELECT
people.*,
locations.id AS location_id,
locations.street,
locations.city,
locations.state
FROM people
LEFT JOIN locations
ON people.home_id = locations.id;");
Give People Homes
If we put a var_dump inside the while loop in our People::all() function, we can see $row_object now contains information about each person's location (if you're using Postman to view http://localhost:8888/people, you might need to switch to the Raw view).
while($row_object){
var_dump($row_object); //print $row_object so we can see what it looks like now
$new_person = new Person(
$row_object->id,
$row_object->name,
$row_object->age
);
$people[] = $new_person;
$row_object = pg_fetch_object($results);
}
Now lets use the additional information on $row_object to create Location objects and add them to $new_person where necessary. First, let's include the Location model at the top of models/person.php.
include_once __DIR__ . '/location.php';
Now, inside the People::all() while loop, we'll add the logic to create a location and add it to $new_person:
while($row_object){
$new_person = new Person(
$row_object->id,
$row_object->name,
$row_object->age
);
if($row_object->location_id){ //test if location_id is truthy
$new_location = new Location( //create a location from the row data
intval($row_object->location_id), //turn the string into an int
$row_object->street,
$row_object->city,
$row_object->state
);
$new_person->home = $new_location; //attach $new_location to $new_person->home
}
$people[] = $new_person;
$row_object = pg_fetch_object($results);
}
Here we test to see if $row_object->location_id is truthy. If it has a value -- a string representation of the id column of the locations table -- then the block of code belonging to the if statement will run, creating $new_location and attaching it to the home property of $new_person.
Give Locations Inhabitants
Again, let's start with some SQL to get all of our locations, with their respective inhabitants:
SELECT * FROM locations JOIN people ON locations.id = people.home_id;
But this only gets the locations that have people associated with them. If there aren't any rows in the people table that have the home_id column set to the id of a specific row in the locations table, then the row won't show up. It's the same issue we had previously with some of the rows in the people column not showing up because they didn't have a home_id. We need to perform the join like normal, and then add in any missing location rows:
SELECT * FROM locations LEFT JOIN people ON locations.id = people.home_id;
Now we have the additional missing rows from the locations table.
Currently, our Locations:all() should look something like this:
static function all(){
//create an empty array
$locations = array();
//query the database
$results = pg_query("SELECT * FROM locations");
$row_object = pg_fetch_object($results);
while($row_object){
$new_location = new Location( //create a new location
$row_object->id,
$row_object->street,
$row_object->city,
$row_object->state
);
$locations[] = $new_location; //push new person object onto $$locations array
$row_object = pg_fetch_object($results);
}
return $locations;
}
Let's adjust the code to include our new SQL code with the JOIN:
$results = pg_query("SELECT * FROM locations LEFT JOIN people ON locations.id = people.home_id");
If you visit http://localhost:8888/locations, you'll notice that we now have duplicate locations whenever a location has more than one person associated with it. This makes sense because we have extra rows for those locations when we perform the SQL query in psql. We'll fix that soon.
You'll also notice that the id properties for the various locations are incorrect. This is the same problem that we encountered before when we were creating People:all(). If we look at the SQL statement in psql, we'll notice that there are two id columns. The second id column overwrites the values of the first one, even though the correct id is the first one. Let's alias the second column:
SELECT
locations.*,
people.id AS person_id,
people.name,
people.age
FROM locations
LEFT JOIN people
ON locations.id = people.home_id;
We can plug this into our PHP code:
$results = pg_query("SELECT
locations.*,
people.id AS person_id,
people.name,
people.age
FROM locations
LEFT JOIN people
ON locations.id = people.home_id;");
Now our id properties are correct.
Next, let's deal with our duplicate locations. Currently, our while loop looks like this:
$row_object = pg_fetch_object($results);
while($row_object){
$new_location = new Location(
$row_object->id,
$row_object->street,
$row_object->city,
$row_object->state
);
$locations[] = $new_location;
$row_object = pg_fetch_object($results);
}
The issue here is that our SQL produces multiple rows for the same location, if it has more than one person associated with it. This is okay, because we need multiple rows to exist in order to have data for each person associated with a given location. What we need to do is, each time a duplicate location row is encountered, skip the process for creating the location:
$row_object = pg_fetch_object($results);
$last_location_id = null; //keep track of the last location id encountered
while($row_object){
//if the current row's id is different from the last row's id, create a new location
//otherwise skip the process of creating a new location
if($row_object->id !== $last_location_id){
$new_location = new Location(
$row_object->id,
$row_object->street,
$row_object->city,
$row_object->state
);
$locations[] = $new_location;
$last_location_id = $row_object->id; //update the $last_location_id to be the current row's id
}
$row_object = pg_fetch_object($results);
}
We start off by initializing a $last_location_id variable. Then, each time we loop through to the next row, we check to see if the new row's id is the same as the last row's id. If it is, we don't create a new Location object.
With that working, let's add people to locations when necessary. This is similar to, but not the same as, the last section when we created Location objects and added them to people. First, whenever we create a new Location object, we want to automatically give it an inhabitants property which is set to an array:
class Location {
public $id;
public $street;
public $city;
public $state;
public function __construct($id, $street, $city, $state) {
$this->id = $id;
$this->street = $street;
$this->city = $city;
$this->state = $state;
$this->inhabitants = [];
}
}
Now include the People model at the top of models/location.php:
include_once __DIR__ . '/person.php';
Next, update the while loop in Locations::all():
while($row_object){
if($row_object->id !== $last_location_id){
$new_location = new Location(
$row_object->id,
$row_object->street,
$row_object->city,
$row_object->state
);
$locations[] = $new_location;
$last_location_id = $row_object->id;
}
//test if person_id is truthy
if($row_object->person_id){
//create a person from the row data
$new_person = new Person(
intval($row_object->person_id), //turn the string into an int
$row_object->name,
$row_object->age
);
//add the new person as an inhabitant of the last element of the locations array
$locations_length = count($locations); //count() returns the # of elements in an array
$last_index_of_locations = $locations_length-1;
$most_recently_added_location = $locations[$last_index_of_locations];
$most_recently_added_location->inhabitants[] = $new_person;
}
$row_object = pg_fetch_object($results);
}
This is pretty similar to when we added homes to people, except for the last section. It finds last element in the $locations array and adds the $new_person as an inhabitant to it.
If we sort our results in the SQL statement by locations.id, we can be assured that whenever we create a new Person object, even if we didn't create a location during that particular loop of the while statement, the last element on the $locations array will always be the location that needs to have an inhabitant added it it (as opposed to some other location).
$results = pg_query("SELECT
locations.*,
people.id AS person_id,
people.name,
people.age
FROM locations
LEFT JOIN people
ON locations.id = people.home_id
ORDER BY locations.id ASC");