REST API application generator for Yii2, openapi 3.0 YAML -> Yii2.
Base on Gii, the Yii Framework Code Generator.
A code generator for OpenAPI and Yii Framework based PHP API application.
Input: OpenAPI 3.0 YAML or JSON (via cebe/php-openapi)
Output: Yii Framework Application with Controllers, Models, database schema
Currently available features:
- Generate Path mappings, Controllers and Actions for API Endpoints. CRUD Endpoints are ready-to-use, other Endpoints are generated as abstract functions that need to be implemented
- Generate Models and validation based on OpenAPI Schema
- Generate Database Schema from OpenAPI Schema
- Generates Database Migrations for schema changes
- Provide Dummy Data via Faker for development
- PHP 7.1 or higher (works fine with PHP 8)
composer require cebe/yii2-openapi:^2.0@beta
You can use this package in your existing application or start a new project using the yii2-app-api application template. For usage of the template, see instructions in the template repo readme.
In your existing Yii application config (works for console as well as web):
<?php
$config = [
// ... this is your application config ...
];
if (YII_ENV_DEV) {
// enable Gii module
$config['bootstrap'][] = 'gii';
$config['modules']['gii'] = [
'class' => yii\gii\Module::class,
'generators' => [
// add ApiGenerator to Gii module
'api' => \cebe\yii2openapi\generator\ApiGenerator::class,
],
];
}
return $config;
To use the web generator, open index.php?r=gii
and select the REST API Generator
.
On console you can run the generator with ./yii gii/api --openApiPath=@app/openapi.yaml
. Where @app/openapi.yaml
should be the absolute path to your OpenAPI spec file. This can be JSON as well as YAML (see also cebe/php-openapi for supported formats).
Run ./yii gii/api --help
for all options.
See Petstore example for example OpenAPI spec.
This library understands the following extensions to the OpenAPI spec:
You may specify custom PHP code for generating fake data for a property:
Post:
properties:
id:
type: integer
tags:
type: array
items:
type: string
example: ['one', 'two']
x-faker: "$faker->randomElements(['one', 'two', 'three', 'four'])"
Specify the table name for a Schema that defines a model which is stored in the database.
You can generate non-db model based on \yii\base\Model without migrations by setting x-table: false
Explicitly specify primary key name for table, if it is different from "id"
Post:
x-table: posts
x-pk: uid
properties:
uid:
type: integer
title:
type: string
Explicitly specify the database type for a column. (MUST contain only real DB type! (json
, jsonb
, uuid
, varchar
etc.)).
If x-db-type
is set to false
, property will be processed as virtual;
It will be added in model as public property, but skipped for migrations generation.
Example values of x-db-type
are:
false
(boolean false)- as string and its value can be like:
- text
- text[]
- INTEGER PRIMARY KEY AUTO_INCREMENT
- decimal(12,4)
- json
- varchar
- VARCHAR
- SMALLINT UNSIGNED ZEROFILL
- MEDIUMINT(10) UNSIGNED ZEROFILL COMMENT "comment" (note the double quotes here)
Such values are not allowed:
int null default null after low_price
(null and default will be handled bynullable
anddefault
keys respectively)- MEDIUMINT(10) UNSIGNED ZEROFILL NULL DEFAULT '7' COMMENT 'comment' AFTER
seti
, ADD INDEXt
(w
)
Specify table indexes
Post:
x-table: posts
x-indexes:
- 'visible,publish_date'
- 'unique:title' #for unique attributes also unique validation check will be added
- 'gist:metadata' #for postgres will generate index using GIST index type
properties:
id:
type: integer
x-db-type: INTEGER PRIMARY KEY AUTO_INCREMENT
title:
type: string
visible:
type: boolean
publish_date:
type: string
format: date
metadata:
type: object
x-db-type: JSON
default: '{}'
There are two ways for define many-to-many relations:
-
property name for many-to-many relation should be equal lower-cased, pluralized related schema name
-
referenced schema should contains mirrored reference to current schema
-
migration for junction table can be generated automatically - table name should be [pluralized, lower-cased schema_name1]2[pluralized, lower-cased schema name2], in alphabetical order; For example, for schemas Post and Tag - table should be posts2tags, for schemas Post and Attachement - table should be attachments2posts
Post:
properties:
...
tags:
type: array
items:
$ref: '#/components/schemas/Tag'
Tag:
properties:
...
posts:
type: array
items:
$ref: '#/components/schemas/Post'
This way allowed creating multiple many-to-many relations between to models
- define junction schema with all necessary attributes. There are only one important requirement - the junction schema name must be started with prefix 'junction_' (This prefix will be used internally only and will be trimmed before table and model generation)
# Model TeamMembers with table team_members will be generated with columns team_id, user_id and role
junction_TeamMembers:
team:
$ref: '#/components/schemas/Team'
user:
$ref: '#/components/schemas/User'
role:
type: string
- Both many-to-many related schemas must have properties with reference to "junction_*" schema. These properties will be used as relation names
Team:
properties:
...
team_members:
type: array
items:
$ref: '#/components/schemas/junction_TeamMembers'
User:
properties:
...
memberships: #You absolutely free with naming for relationship attributes
type: array
items:
$ref: '#/components/schemas/junction_TeamMembers'
- see both examples here tests/specs/many2many.yaml
When generating code from an OpenAPI description there are many possible ways to achive a fitting result. Thus there are some assumptions and limitations that are currently applied to make this work. Here is a (possibly incomplete) list:
- The current implementation works best with OpenAPI description that follows the JSON:API guidelines.
- The request and response format/schema is currently not extracted from OpenAPI schema and may need to be adjusted manually if it does not follow JSON:API
- column/field/property with name
id
is considered as Primary Key by this library and it is automatically handled by DB/Yii; so remove it from validationrules()
- other fields can currently be used as primary keys using the
x-pk
OpenAPI extension (see below) but it may not be work correctly in all cases, please report bugs if you find them.
- other fields can currently be used as primary keys using the
Other things to keep in mind:
When adding new fields in the API models, new migrations will be generated to add these fields to the table. For a project that is already in production, it should be considered to adjust the generated migration to add default values for existing data records.
One case where this is important is the addition of a new column with NOT NULL
contraint, which does not provide a default value.
Such a migration will fail when the table is not empty:
$this->addColumn('{{%company}}', 'name', $this->string(128)->notNull());
Fails on a PostgreSQL database with
add column name string(128) NOT NULL to table {{%company}} ...Exception: SQLSTATE[23502]: Not null violation: 7 ERROR: column "name" contains null values
The solution would be to create the column, allowing NULL, set the value to a default and add the null constraint later.
$this->addColumn('{{%company}}', 'name', $this->string(128)->null());
$this->update('{{%company}}', ['name' => 'No name']);
$this->alterColumn('{{%company}}', 'name', $this->string(128)->notNull());
NOT NULL
in DB migrations is determined by nullable
and required
properties of the OpenAPI schema.
e.g. attribute = 'my_property'.
- If you define attribute neither "required" nor via "nullable", then it is by default
NULL
:
ExampleSchema:
properties:
my_property:
type: string
- If you define attribute in "required", then it is
NOT NULL
ExampleSchema:
required:
- my_property
properties:
my_property:
type: string
- If you define attribute via "nullable", then it overrides "required", e.g. allow
NULL
in this case:
ExampleSchema:
required:
- my_property
properties:
my_property:
type: string
nullable: true
- If you define attribute via "nullable", then it overrides "required", e.g.
NOT NULL
in this case:
test_table:
required:
properties:
my_property:
type: string
nullable: false
It works on all 3 DB: MySQL, MariaDb and PgSQL.
test_table:
properties:
my_property:
enum:
- one
- two
- three
Note: Change in enum values are not very simple. For Mysql and Mariadb, migrations will be generated but in many cases custom modification in it are required. For Pgsql migrations for change in enum values will not be generated. It should be handled manually.
precision-default = 10 scale-default = 2
- You can define attribute like "numeric(precision,scale)":
test_table:
properties:
my_property:
x-db-type: decimal(12,4)
DB-Result = decimal(12,4)
- You can define attribute like "numeric(precision)" with default scale-default = 2:
test_table:
properties:
my_property:
x-db-type: decimal(12)
DB-Result = decimal(12,2)
- You can define attribute like "numeric" with precision-default = 10 and scale-default = 2:
test_table:
properties:
my_property:
x-db-type: decimal
DB-Result = decimal(10,2)
Gii Generator Form:
Generated files:
There commands are available to develop and check the tests. It is available inside the Docker container. To enter into bash shell of container, run make cli
.
cd tests
./yii migrate-mysql/up
./yii migrate-mysql/down 4
./yii migrate-maria/up
./yii migrate-maria/down 4
./yii migrate-pgsql/up
./yii migrate-pgsql/down 4
To apply multiple migration with one command:
./yii migrate-mysql/up --interactive=0 && \
./yii migrate-mysql/down --interactive=0 4 && \
./yii migrate-maria/up --interactive=0 && \
./yii migrate-maria/down --interactive=0 4 && \
./yii migrate-pgsql/up --interactive=0 && \
./yii migrate-pgsql/down --interactive=0 4
Need help with your API project?
Professional support, consulting as well as software development services are available:
https://www.cebe.cc/en/contact
Development of this library is sponsored by cebe.:cloud: "Your Professional Deployment Platform".