PHP "php://input" vs $_POST
First, a basic truth about PHP.
PHP was not designed to explicitly give you a pure REST (GET, POST, PUT, PATCH, DELETE) like interface for handling HTTP requests.
However, the $_SERVER
, $_COOKIE
, $_POST
, $_GET
, and $_FILES
superglobals, and the function filter_input_array()
are very useful for the average person's / layman's needs.
The number one hidden advantage of $_POST
(and $_GET
) is that your input data is url-decoded automatically by PHP. You never even think about having to do it, especially for query string parameters within a standard GET
request, or HTTP body data submitted with a POST
request.
Other HTTP Request Methods
Those studying the underlying HTTP protocol and its various request methods come to understand that there are many HTTP request methods, including the often referenced PUT
, PATCH
(not used in Google's Apigee), and DELETE
.
In PHP, there are no superglobals or input filter functions for getting HTTP request body data when POST
is not used. What are disciples of Roy Fielding to do? ;-)
However, then you learn more ...
That being said, as you advance in your PHP programming knowledge and want to use JavaScript's XmlHttpRequest
object (jQuery for some), you come to see the limitation of this scheme.
$_POST
limits you to the use of two media types in the HTTP Content-Type
header:
application/x-www-form-urlencoded
, andmultipart/form-data
Thus, if you want to send data values to PHP on the server, and have it show up in the $_POST
superglobal, then you must urlencode it on the client-side and send said data as key/value pairs--an inconvenient step for novices (especially when trying to figure out if different parts of the URL require different forms of urlencoding: normal, raw, etc..).
For all you jQuery users, the $.ajax()
method is converting your JSON to URL encoded key/value pairs before transmitting them to the server. You can override this behavior by setting processData: false
. Just read the $.ajax() documentation, and don't forget to send the correct media type in the Content-Type header.
php://input, but ...
Even if you use php://input
instead of $_POST
for your HTTP POST
request body data, it will not work with an HTTP Content-Type
of multipart/form-data
This is the content type that you use on an HTML form when you want to allow file uploads!
<form enctype="multipart/form-data" accept-charset="utf-8" action="post">
<input type="file" name="resume">
</form>
Therefore, in traditional PHP, to deal with a diversity of content types from an HTTP POST
request, you will learn to use $_POST
or filter_input_array(POST)
, $_FILES
, and php://input
. There is no way to just use one, universal input source for HTTP POST
requests in PHP.
You cannot get files through $_POST
, filter_input_array(POST)
, or php://input
, and you cannot get JSON/XML/YAML in either filter_input_array(POST)
or $_POST
.
PHP Manual: php://input
php://input is a read-only stream that allows you to read raw data from the request body...php://input is not available with enctype="multipart/form-data".
PHP Frameworks to the rescue?
PHP frameworks like Codeigniter 4 and Laravel use a facade to provide a cleaner interface (IncomingRequest
or Request
objects) to the above. This is why professional PHP developers use frameworks instead of raw PHP.
Of course, if you like to program, you can devise your own facade object to provide what frameworks do. It is because I have taken time to investigate this issue that I am able to write this answer.
URL encoding? What the heck!!!???
Typically, if you are doing a normal, synchronous (when the entire page redraws) HTTP requests with an HTML form, the user-agent (web browser) will urlencode your form data for you. If you want to do an asynchronous HTTP requests using the XmlHttpRequest
object, then you must fashion a urlencoded string and send it, if you want that data to show up in the $_POST
superglobal.
How in touch are you with JavaScript? :-)
Converting from a JavaScript array or object to a urlencoded string bothers many developers (even with new APIs like Form Data). They would much rather just be able to send JSON, and it would be more efficient for the client code to do so.
Remember (wink, wink), the average web developer does not learn to use the XmlHttpRequest
object directly, global functions, string functions, array functions, and regular expressions like you and I ;-). Urlencoding for them is a nightmare. ;-)
PHP, what gives?
PHP's lack of intuitive XML and JSON handling turns many people off. You would think it would be part of PHP by now (sigh).
So many media types (MIME types in the past)
XML, JSON, and YAML all have media types that can be put into an HTTP Content-Type
header.
- application/xml
- applicaiton/json
- application/yaml (although IANA has no official designation listed)
Look how many media-types (formerly, MIME types) are defined by IANA.
Look how many HTTP headers there are.
php://input or bust
Using the php://input
stream allows you to circumvent the baby-sitting / hand holding level of abstraction that PHP has forced on the world. :-) With great power comes great responsibility!
Now, before you deal with data values streamed through php://input
, you should / must do a few things.
- Determine if the correct HTTP method has been indicated (GET, POST, PUT, PATCH, DELETE, ...)
- Determine if the HTTP Content-Type header has been transmitted.
- Determine if the value for the Content-Type is the desired media type.
- Determine if the data sent is well formed XML / JSON / YAML / etc.
- If necessary, convert the data to a PHP datatype: array or object.
- If any of these basic checks or conversions fails, throw an exception!
What about the character encoding?
AH, HA! Yes, you might want the data stream being sent into your application to be UTF-8 encoded, but how can you know if it is or not?
Two critical problems.
- You do not know how much data is coming through
php://input
. - You do not know for certain the current encoding of the data stream.
Are you going to attempt to handle stream data without knowing how much is there first? That is a terrible idea. You cannot rely exclusively on the HTTP Content-Length
header for guidance on the size of streamed input because it can be spoofed.
You are going to need a:
- Stream size detection algorithm.
- Application defined stream size limits (Apache / Nginx / PHP limits may be too broad).
Are you going to attempt to convert stream data to UTF-8 without knowing the current encoding of the stream? How? The iconv stream filter (iconv stream filter example) seems to want a starting and ending encoding, like this.
'convert.iconv.ISO-8859-1/UTF-8'
Thus, if you are conscientious, you will need:
- Stream encoding detection algorithm.
- Dynamic / runtime stream filter definition algorithm (because you cannot know the starting encoding a priori).
(Update: 'convert.iconv.UTF-8/UTF-8'
will force everything to UTF-8, but you still have to account for characters that the iconv library might not know how to translate. In other words, you have to some how define what action to take when a character cannot be translated: 1) Insert a dummy character, 2) Fail / throw and exception).
You cannot rely exclusively on the HTTP Content-Encoding
header, as this might indicate something like compression as in the following. This is not what you want to make a decision off of in regards to iconv.
Content-Encoding: gzip
Therefore, the general steps might be ...
Part I: HTTP Request Related
- Determine if the correct HTTP method has been indicated (GET, POST, PUT, PATCH, DELETE, ...)
- Determine if the HTTP Content-Type header has been transmitted.
- Determine if the value for the Content-Type is the desired media type.
Part II: Stream Data Related
- Determine the size of the input stream (optional, but recommended).
- Determine the encoding of the input stream.
- If necessary, convert the input stream to the desired character encoding (UTF-8).
- If necessary, reverse any application level compression or encryption, and then repeat steps 4, 5, and 6.
Part III: Data Type Related
- Determine if the data sent is well formed XML / JSON / YMAL / etc.
(Remember, the data can still be a URL encoded string which you must then parse and URL decode).
- If necessary, convert the data to a PHP datatype: array or object.
Part IV: Data Value Related
Filter input data.
Validate input data.
Now do you see?
The $_POST
superglobal, along with php.ini settings for limits on input, are simpler for the layman. However, dealing with character encoding is much more intuitive and efficient when using streams because there is no need to loop through superglobals (or arrays, generally) to check input values for the proper encoding.
The reason is that php://input
returns all the raw data after the HTTP-headers of the request, regardless of the content type.
The PHP superglobal $_POST
, only is supposed to wrap data that is either
application/x-www-form-urlencoded
(standard content type for simple form-posts) ormultipart/form-data
(mostly used for file uploads)
This is because these are the only content types that must be supported by user agents. So the server and PHP traditionally don't expect to receive any other content type (which doesn't mean they couldn't).
So, if you simply POST a good old HTML form
, the request looks something like this:
POST /page.php HTTP/1.1
key1=value1&key2=value2&key3=value3
But if you are working with Ajax a lot, this probaby also includes exchanging more complex data with types (string, int, bool) and structures (arrays, objects), so in most cases JSON is the best choice. But a request with a JSON-payload would look something like this:
POST /page.php HTTP/1.1
{"key1":"value1","key2":"value2","key3":"value3"}
The content would now be application/json
(or at least none of the above mentioned), so PHP's $_POST
-wrapper doesn't know how to handle that (yet).
The data is still there, you just can't access it through the wrapper. So you need to fetch it yourself in raw format with file_get_contents('php://input')
(as long as it's not multipart/form-data
-encoded).
This is also how you would access XML-data or any other non-standard content type.