Single file PHP 7 script that adds a REST API to a MySQL 5.5 InnoDB database. PostgreSQL 9.1 and MS SQL Server 2012 are fully supported.
NB: Stable "v1" branch is: https://github.com/mevdschee/php-crud-api/tree/v1
Related projects:
- PHP-API-AUTH: Authentication add-on (for v1) supporting JWT or username/password.
- PHP-SP-API: Single file PHP script that adds a REST API to a SQL database.
- PHP-CRUD-UI: Single file PHP script that adds a UI to a PHP-CRUD-API (v1) project.
- VUE-CRUD-UI: Single file Vue.js script that adds a UI to a PHP-CRUD-API (v1) project.
There are also ports of this script in:
- Java JDBC by Ivan Kolchagov (v1)
- Java Spring Boot + jOOQ (v2: work in progress)
There are also proof-of-concept ports of this script that only support basic REST CRUD functionality in: PHP, Java, Go, C# .net core, Node.js and Python.
- PHP 7.0 or higher with PDO drivers for MySQL, PgSQL or SqlSrv enabled
- MySQL 5.6 / MariaDB 10.0 or higher for spatial features in MySQL
- PostGIS 2.0 or higher for spatial features in PostgreSQL 9.1 or higher
- SQL Server 2012 or higher (2017 for Linux support)
This is a single file application! Upload "api.php" somewhere and enjoy!
For local development you may run PHP's built-in web server:
php -S localhost:8080
Test the script by opening the following URL:
http://localhost:8080/api.php/records/posts/1
Dont forget to modify the configuration at the bottom of the file.
Edit the following lines in the bottom of the file "api.php":
$config = new Config([
'username' => 'xxx',
'password' => 'xxx',
'database' => 'xxx',
]);
These are all the configuration options and their default value between brackets:
- "driver": "mysql" (default), "pgsql" or "sqlsrv"
- "address": Hostname of the database server ("localhost")
- "port": TCP port of the database server (defaults to driver default)
- "username": Username of the user connecting to the database (no default)
- "password": Password of the user connecting to the database (no default)
- "database": Database the connecting is made to (no default)
- "middlewares": List of middlewares to load ("cors")
- "controllers": List of controllers to load ("records,columns,cache,openapi")
- "cacheType": "TempFile" (default), "Redis", "Memcache", "Memcached" or "NoCache"
- "cachePath": Path/address of the cache (defaults to system's "temp" directory)
- "cacheTime": Number of seconds the cache is valid (10)
- "debug": Show errors in the "X-Debug-Info" header (false)
The code resides in the "src" directory. You can access it at the URL:
http://localhost:8080/src/records/posts/1
You can compile all files into a single "api.php" file using:
php build.php
NB: The script appends the classes in alphabetical order (directories first).
These limitation were also present in v1:
- Primary keys should either be auto-increment (from 1 to 2^53) or UUID
- Composite primary or foreign keys are not supported
- Complex writes (transactions) are not supported
- Complex queries calling functions (like "concat" or "sum") are not supported
- MySQL storage engine must be either InnoDB or XtraDB
- Only MySQL, PostgreSQL and SQLServer support spatial/GIS functionality
These features match features in v1 (see branch "v1"):
- Single PHP file, easy to deploy.
- Very little code, easy to adapt and maintain
-
Streaming data, low memory footprint - Supports POST variables as input (x-www-form-urlencoded)
- Supports a JSON object as input
- Supports a JSON array as input (batch insert)
- Supports file upload from web forms (multipart/form-data)
-
Condensed JSON output: first row contains field names - Sanitize and validate input using callbacks
- Permission system for databases, tables, columns and records
- Multi-tenant database layouts are supported
- Multi-domain CORS support for cross-domain requests
- Support for reading joined results from multiple tables
- Search support on multiple criteria
- Pagination, seeking, sorting and column selection
- Relation detection with nested results (belongsTo, hasMany and HABTM)
-
Relation "transforms" (of condensed JSON) for PHP and JavaScript - Atomic increment support via PATCH (for counters)
- Binary fields supported with base64 encoding
- Spatial/GIS fields and filters supported with WKT
- Unstructured data support through JSON/JSONB
- Generate API documentation using OpenAPI tools
- Authentication via JWT token or username/password
-
SQLite support
NB: No checkmark means: not yet implemented. Striken means: will not be implemented.
These features are new and were not included in v1.
- Does not reflect on every request (better performance)
- Complex filters (with both "and" & "or") are supported
- Support for output of database structure in JSON
- Support for boolean and binary data in all database engines
- Support for relational data on read (not only on list operation)
- Support for middleware to modify all operations (also list)
- Error reporting in JSON with corresponding HTTP status
You can enable the following middleware using the "middlewares" config parameter:
- "cors": Support for CORS requests (enabled by default)
- "authorization": Restrict access to certain tables or columns
- "basicAuth": Support for "Basic Authentication"
- "firewall": Limit access to specific IP addresses
- "validation": Return input validation errors for custom rules
- "sanitation": Apply input sanitation on create and update
The "middlewares" config parameter is a comma separated list of enabled middlewares. You can tune the middleware behavior using middleware specific configuration parameters:
- "cors.allowedOrigins": The origins allowed in the CORS headers ("*")
- "cors.allowHeaders": The headers allowed in the CORS request ("Content-Type, X-XSRF-TOKEN")
- "cors.allowMethods": The methods allowed in the CORS request ("OPTIONS, GET, PUT, POST, DELETE, PATCH")
- "cors.allowCredentials": To allow credentials in the CORS request ("true")
- "cors.maxAge": The time that the CORS grant is valid in seconds ("1728000")
- "authorization.tableHandler": Handler to implement table authorization rules ("")
- "authorization.columnHandler": Handler to implement column authorization rules ("")
- "authorization.recordHandler": Handler to implement record authorization filter rules ("")
- "basicAuth.passwordFile": The file to read for username/password combinations (".htpasswd")
- "basicAuth.realm": Message shown when asking for credentials ("Username and password required")
- "firewall.reverseProxy": Set to "true" when a reverse proxy is used ("")
- "firewall.allowedIpAddresses": List of IP addresses that are allowed to connect ("")
- "validation.handler": Handler to implement validation rules for input values ("")
- "sanitation.handler": Handler to implement sanitation rules for input values ("")
If you don't specify these parameters in the configuration, then the default values (between brackets) are used.
TreeQL allow you to create a Tree of JSON objects based on your SQL database structure (relations).
It is loosely based on the REST standard and also inspired by json:api.
The example posts table has only a a few fields:
posts
=======
id
title
content
created
The CRUD + List operations below act on this table.
If you want to create a record the request can be written in URL format as:
POST /records/posts
You have to send a body containing:
{
"title": "Black is the new red",
"content": "This is the second post.",
"created": "2018-03-06T21:34:01Z"
}
And it will return the value of the primary key of the newly created record:
2
To read a record from this table the request can be written in URL format as:
GET /records/posts/1
Where "1" is the value of the primary key of the record that you want to read. It will return:
{
"id": 1
"title": "Hello world!",
"content": "Welcome to the first post.",
"created": "2018-03-05T20:12:56Z"
}
On read operations you may apply joins.
To update a record in this table the request can be written in URL format as:
PUT /records/posts/1
Where "1" is the value of the primary key of the record that you want to update. Send as a body:
{
"title": "Adjusted title!"
}
This adjusts the title of the post. And the return value is the number of rows that are set:
1
If you want to delete a record from this table the request can be written in URL format as:
DELETE /records/posts/1
And it will return the number of deleted rows:
1
To list records from this table the request can be written in URL format as:
GET /records/posts
It will return:
{
"records":[
{
"id": 1,
"title": "Hello world!",
"content": "Welcome to the first post.",
"created": "2018-03-05T20:12:56Z"
}
]
}
On list operations you may apply filters and joins.
Filters provide search functionality, on list calls, using the "filter" parameter. You need to specify the column name, a comma, the match type, another commma and the value you want to filter on. These are supported match types:
- "cs": contain string (string contains value)
- "sw": start with (string starts with value)
- "ew": end with (string end with value)
- "eq": equal (string or number matches exactly)
- "lt": lower than (number is lower than value)
- "le": lower or equal (number is lower than or equal to value)
- "ge": greater or equal (number is higher than or equal to value)
- "gt": greater than (number is higher than value)
- "bt": between (number is between two comma separated values)
- "in": in (number or string is in comma separated list of values)
- "is": is null (field contains "NULL" value)
You can negate all filters by prepending a "n" character, so that "eq" becomes "neq". Examples of filter usage are:
GET /records/categories?filter=name,eq,Internet
GET /records/categories?filter=name,sw,Inter
GET /records/categories?filter=id,le,1
GET /records/categories?filter=id,ngt,2
GET /records/categories?filter=id,bt,1,1
Output:
{
"records":[
{
"id": 1
"name": "Internet"
}
]
}
In the next section we dive deeper into how you can apply multiple filters on a single list call.
Filters can be a by applied by repeating the "filter" parameter in the URL. For example the following URL:
GET /records/categories?filter=id,gt,1&filter=id,lt,3
will request all categories "where id > 1 and id < 3". If you wanted "where id = 2 or id = 4" you should write:
GET /records/categories?filter1=id,eq,2&filter2=id,eq,4
As you see we added a number to the "filter" parameter to indicate that "OR" instead of "AND" should be applied. Note that you can also repeat "filter1" and create an "AND" within an "OR". Since you can also go one level deeper by adding a letter (a-f) you can create almost any reasonably complex condition tree.
NB: You can only filter on the requested table (not on it's included) and filters are only applied on list calls.
By default all columns are selected. With the "include" parameter you can select specific columns. You may use a dot to separate the table name from the column name. Multiple columns should be comma separated. An asterisk ("*") may be used as a wildcard to indicate "all columns". Similar to "include" you may use the "exclude" parameter to remove certain columns:
GET /records/categories/1?include=name
GET /records/categories/1?include=categories.name
GET /records/categories/1?exclude=categories.id
Output:
{
"name": "Internet"
}
NB: Columns that are used to include related entities are automatically added and cannot be left out of the output.
With the "order" parameter you can sort. By default the sort is in ascending order, but by specifying "desc" this can be reversed:
GET /records/categories?order=name,desc
GET /records/categories?order=id,desc&order=name
Output:
{
"records":[
{
"id": 3
"name": "Web development"
},
{
"id": 1
"name": "Internet"
}
]
}
NB: You may sort on multiple fields by using multiple "order" parameters. You can not order on "joined" columns.
The "page" parameter holds the requested page. The default page size is 20, but can be adjusted (e.g. to 50):
GET /records/categories?order=id&page=1
GET /records/categories?order=id&page=1,50
Output:
{
"records":[
{
"id": 1
"name": "Internet"
},
{
"id": 3
"name": "Web development"
}
],
"results": 2
}
NB: Pages that are not ordered cannot be paginated.
Let's say that you have a posts table that has comments (made by users) and the posts can have tags.
posts comments users post_tags tags
======= ======== ======= ========= =======
id id id id id
title post_id username post_id name
content user_id phone tag_id
created message
When you want to list posts with their comments users and tags you can ask for two "tree" paths:
posts -> comments -> users
posts -> post_tags -> tags
These paths have the same root and this request can be written in URL format as:
GET /records/posts?join=comments,users&join=tags
Here you are allowed to leave out the intermediate table that binds posts to tags. In this example you see all three table relation types (hasMany, belongsTo and hasAndBelongsToMany) in effect:
- "post" has many "comments"
- "comment" belongs to "user"
- "post" has and belongs to many "tags"
This may lead to the following JSON data:
{
"records":[
{
"id": 1,
"title": "Hello world!",
"content": "Welcome to the first post.",
"created": "2018-03-05T20:12:56Z",
"comments": [
{
id: 1,
post_id: 1,
user_id: {
id: 1,
username: "mevdschee",
phone: null,
},
message: "Hi!"
},
{
id: 2,
post_id: 1,
user_id: {
id: 1,
username: "mevdschee",
phone: null,
},
message: "Hi again!"
}
],
"tags": []
},
{
"id": 2,
"title": "Black is the new red",
"content": "This is the second post.",
"created": "2018-03-06T21:34:01Z",
"comments": [],
"tags": [
{
id: 1,
message: "Funny"
},
{
id: 2,
message: "Informational"
}
]
}
]
}
You see that the "belongsTo" relationships are detected and the foreign key value is replaced by the referenced object. In case of "hasMany" and "hasAndBelongsToMany" the table name is used a new property on the object.
When you want to create, read, update or delete you may specify multiple primary key values in the URL. You also need to send an array instead of an object in the request body for create and update.
To read a record from this table the request can be written in URL format as:
GET /records/posts/1,2
The result may be:
[
{
"id": 1,
"title": "Hello world!",
"content": "Welcome to the first post.",
"created": "2018-03-05T20:12:56Z"
},
{
"id": 2,
"title": "Black is the new red",
"content": "This is the second post.",
"created": "2018-03-06T21:34:01Z"
}
]
Similarly when you want to do a batch update the request in URL format is written as:
PUT /records/posts/1,2
Where "1" and "2" are the values of the primary keys of the records that you want to update. The body should contain the same number of objects as there are primary keys in the URL:
[
{
"title": "Adjusted title for ID 1"
},
{
"title": "Adjusted title for ID 2"
}
]
This adjusts the titles of the posts. And the return values are the number of rows that are set:
1,1
Which means that there were two update operations and each of them had set one row. Batch operations use database transactions, so they either all succeed or all fail (successful ones get roled back).
For spatial support there is an extra set of filters that can be applied on geometry columns and that starting with an "s":
- "sco": spatial contains (geometry contains another)
- "scr": spatial crosses (geometry crosses another)
- "sdi": spatial disjoint (geometry is disjoint from another)
- "seq": spatial equal (geometry is equal to another)
- "sin": spatial intersects (geometry intersects another)
- "sov": spatial overlaps (geometry overlaps another)
- "sto": spatial touches (geometry touches another)
- "swi": spatial within (geometry is within another)
- "sic": spatial is closed (geometry is closed and simple)
- "sis": spatial is simple (geometry is simple)
- "siv": spatial is valid (geometry is valid)
These filters are based on OGC standards and so is the WKT specification in which the geometry columns are represented.
By default all tables are reflected. If you want to restrict access to some tables you may add the 'authorization' middleware and define a 'authorization.tableHandler' function that returns 'false' for these tables.
'authorization.tableHandler' => function ($method, $path, $databaseName, $tableName) {
return $tableName != 'license_keys';
},
The above example will restrict access to the table 'license_keys' in all API calls.
'authorization.columnHandler' => function ($method, $path, $databaseName, $tableName, $columnName) {
return !($tableName == 'users' && $columnName == 'password');
},
The above example will restrict access to the 'password' field from the 'users' table in all API calls.
'authorization.recordHandler' => function ($method, $path, $databaseName, $tableName, $columnName) {
return ($tableName == 'users') ? 'filter=username,neq,admin' : '';
},
This will disallow viewing the user records where the username is 'admin'. It allows you to add a filter to every query.
By default all input is accepted and sent to the database. If you want to strip (certain) HTML tags before storing you may add the 'sanitation' middleware and define a 'sanitation.handler' function that returns the adjusted value.
'sanitation.handler' => function ($method, $tableName, $column, $value) {
return is_string($value) ? strip_tags($value) : $value;
},
The above example will strip all HTML tags from strings in the input.
By default all input is accepted. If you want to validate the input, you may add the 'validation' middleware and define a 'validation.handler' function that returns a boolean indicating whether or not the value is valid.
'validation.handler' => function ($method, $tableName, $column, $value, $context) {
return ($column['name'] == 'post_id' && !is_numeric($value)) ? 'must be numeric' : true;
},
When you edit a comment with id 4 using:
PUT /records/comments/4
And you send as a body:
{"post_id":"two"}
Then the server will return a '422' HTTP status code and nice error message:
{
"code": 1013,
"message": "Input validation failed for 'comments'",
"details": {
"post_id":"must be numeric"
}
}
You can parse this output to make form fields show up with a red border and their appropriate error message.
There are 4 cache engines that can be configured by the "cacheType" config parameter:
- TempFile (default)
- Redis
- Memcache
- Memcached
You can install the dependencies for the last three engines by running:
sudo apt install php-redis redis
sudo apt install php-memcache memcached
sudo apt install php-memcached memcached
The default engine has no dependencies and will use temporary files in the system "temp" path.
You may use the "cachePath" config parameter to specify the file system path for the temporary files or in case that you use a non-default "cacheType" the hostname (optionally with port) of the cache server.
These are the supported types with their default length/precision/scale:
character types
- varchar(255)
- clob
boolean types:
- boolean
integer types:
- integer
- bigint
floating point types:
- float
- double
decimal types:
- decimal(19,4)
date/time types:
- date
- time
- timestamp
binary types:
- varbinary(255)
- blob
other types:
- geometry /* non-jdbc type, extension with limited support */
JavaScript does not support 64 bit integers. All numbers are stored as 64 bit floating point values. The mantissa of a 64 bit floating point number is only 53 bit and that is why all integer numbers bigger than 53 bit may cause problems in JavaScript.
The following errors may be reported:
- 1000: Route not found (404 NOT FOUND)
- 1001: Table not found (404 NOT FOUND)
- 1002: Argument count mismatch (422 UNPROCESSABLE ENTITY)
- 1003: Record not found (404 NOT FOUND)
- 1004: Origin is forbidden (403 FORBIDDEN)
- 1005: Column not found (404 NOT FOUND)
- 1006: Table already exists (409 CONFLICT)
- 1007: Column already exists (409 CONFLICT)
- 1008: Cannot read HTTP message (422 UNPROCESSABLE ENTITY)
- 1009: Duplicate key exception (409 CONFLICT)
- 1010: Data integrity violation (409 CONFLICT)
- 1011: Authorization required (401 UNAUTHORIZED)
- 1012: Access denied (403 FORBIDDEN)
- 1013: Input validation failed (422 UNPROCESSABLE ENTITY)
- 9999: Unknown error (500: INTERNAL_SERVER_ERROR)
The following JSON structure is used:
{
"code":1002,
"message":"Argument count mismatch in '1'"
}
NB: Any non-error response will have status: 200 OK
I am testing mainly on Ubuntu and I have the following test setups:
- (Docker) Debian 9 with PHP 7.0, MariaDB 10.1, PostgreSQL 9.6 (PostGIS 2.3)
- (Docker) Ubuntu 16.04 with PHP 7.0, MariaDB 10.0, PostgreSQL 9.5 (PostGIS 2.2) and SQL Server 2017
- (Docker) Ubuntu 18.04 with PHP 7.2, MySQL 5.7, PostgreSQL 10.4 (PostGIS 2.4)
This covers not all environments (yet), so please notify me of failing tests and report your environment. I will try to cover most relevant setups in the "docker" folder of the project.
To run the functional tests locally you may run the following command:
php test.php
This runs the functional tests from the "tests" directory. It uses the database dumps (fixtures) and database configuration (config) from the corresponding subdirectories.
Install docker using the following commands and then logout and login for the changes to take effect:
sudo apt install docker.io
sudo usermod -aG docker ${USER}
To run the docker tests run "build_all.sh" and "run_all.sh" from the docker directory. The output should be:
================================================
Debian 9
================================================
[1/4] Starting MariaDB 10.1 ..... done
[2/4] Starting PostgreSQL 9.6 ... done
[3/4] Starting SQLServer 2017 ... skipped
[4/4] Cloning PHP-CRUD-API v2 ... done
------------------------------------------------
mysql: 80 tests ran in 2470 ms, 0 failed
pgsql: 80 tests ran in 680 ms, 0 failed
sqlsrv: skipped, driver not loaded
================================================
Ubuntu 16.04
================================================
[1/4] Starting MariaDB 10.0 ..... done
[2/4] Starting PostgreSQL 9.5 ... done
[3/4] Starting SQLServer 2017 ... done
[4/4] Cloning PHP-CRUD-API v2 ... done
------------------------------------------------
mysql: 80 tests ran in 2490 ms, 0 failed
pgsql: 80 tests ran in 710 ms, 0 failed
sqlsrv: 80 tests ran in 4204 ms, 0 failed
================================================
Ubuntu 18.04
================================================
[1/4] Starting MySQL 5.7 ........ done
[2/4] Starting PostgreSQL 10.4 .. done
[3/4] Starting SQLServer 2017 ... skipped
[4/4] Cloning PHP-CRUD-API v2 ... done
------------------------------------------------
mysql: 80 tests ran in 2882 ms, 0 failed
pgsql: 80 tests ran in 668 ms, 0 failed
sqlsrv: skipped, driver not loaded
The above test run (including starting up the databases) takes less than one minute on my machine.
$ ./run.sh
1) debian9
2) ubuntu16
3) ubuntu18
> 3
================================================
Ubuntu 18.04
================================================
[1/4] Starting MySQL 5.7 ........ done
[2/4] Starting PostgreSQL 10.4 .. done
[3/4] Starting SQLServer 2017 ... skipped
[4/4] Cloning PHP-CRUD-API v2 ... skipped
------------------------------------------------
mysql: 84 tests ran in 2899 ms, 0 failed
pgsql: 84 tests ran in 380 ms, 0 failed
sqlsrv: skipped, driver not loaded
root@b7ab9472e08f:/php-crud-api#
As you can see the "run.sh" script gives you access to a prompt in a chosen the docker environment. In this environment the local files are mounted. This allows for easy debugging on different environments. You may type "exit" when you are done.