Request Parameter Handling in Sling

Servlet API

The Servlet API specification provides the following methods to access the parameters of a request

The actual encoding of the parameters is all but safe because the encoding of URLs is not very well defined and browsers do not set the character encoding when sending post data. Fortunately, they use the same character encoding for sending back form content as was used by the server to send the form.

Sling API

To overcome the restrictions and to provide uniform access to request parameters the Sling API in addition to the Servlet API methods to access parameters provides an abstraction of parameters which is applicable to all parameters sent by clients, the RequestParameter interface. Through this interface, each parameter may be analyzed for these topics:

To accomodate this new interface as well as to provide easy access in the traditional way the SlingHttpServletRequest interface adds following methods to the standard Servlet API parameter access methods:

All parameters are handled the same, that is all methods give access to the same parameters regardless of whether the parameters were transmitted in the request query, as part of form encoded data or as part of a multipart/form-data request.

As of Sling Engine 2.1.0 the order or request parameters in the getRequestParameterMap(), getParameterMap(), and getParameterNams() is preserved as follows:

• The first entries are the parameters reported by the servlet container. The order of these parameters amongst each other is not defined. The SlingHttpServletRequest provides them in the same order as provided by the servlet container.
• After the servlet container provided parameters are parameters extracted from the request in case multipart/form-data POST requests. The order of these parameters is preserved as they are submitted in the request. This conforms to HTML 4.01 spec on forms submitted with multipart/form-data encoding: A "multipart/form-data" message contains a series of parts, each representing a successful control. The parts are sent to the processing agent in the same order the corresponding controls appear in the document stream. Part boundaries should not occur in any of the data; how this is done lies outside the scope of this specification (17.13.4 Form content types)

Be warned: Only rely on request parameter ordering multipart/form-data POST requests without a query part in the request URL.

Effects of Sling on Servlet API parameter methods

From within Sling servlets/scripts you can no longer rely on the original semantics of the Servlet API methods for dealing with parameters as

• ServletRequest.getParameter(String)
• ServletRequest.getParameterValues(String)
• ServletRequest.getParameterMap()
• ServletRequest.getParameterNames()
• ServletRequest.getParts() and
• ServletRequest.getPart(String)

internally use the Sling parameter support (and therefore have the same implications on e.g. encoding). You should preferably use the Sling methods getRequestParameter* instead.

Calling ServletRequest.getInputStream() is not supported, nor relying on some 3rd party libraries which are internally using that method like Apache Commons Fileupload. This is because the Sling parameter support needs exclusive access to the request's input stream.

Character Encoding

Traditionally, the encoding of parameters, especially in text area input forms, has been a big issue. To solve this issue Sling introduces the following convention:

• All forms should contain a hidden field of the name _charset_ containing the actual encoding used to send the form from the server to the client
• All forms should be sent with UTF-8 character encoding

The first rule is essential as it helps decoding the form input correctly. The second rule is not actually a very hard requirement but to enable support for all (or most) character sets used, using UTF-8 is one of the best choices anyway.

When Sling is now receiving a request and is asked for the parameters, the parameters are parsed in two phases: The first phase just parses the raw input data using an identity transformation of bytes to characters. This identity transformation happens to generate strings as the original data was generated with ISO-8859-1 encoding. The second phase locates the _charset_ parameter and fixes the character encodings of the parameters as follows:

• All names of the parameters are re-encoded
• The parameter values are re-encoded, unless the parameter value is an uploaded file. Actually the parameter (not the files of course) are internally as byte[] where the conversion to a string is done on the fly (and yes, the conversion using the _charset_ character encoding is of course cached for performance reasons)
• If the parameter is an uploaded file, the file name is re-encoded on the fly when accessed