negotiator

Negotiator

Proper Content Negotiation for Python

Introduction

Negotiator offers a framework for making content negotiation decisions based on the HTTP accept headers.

NOTE it currently only formally supports Accept and Accept-Language, but it is a short haul to support for Accept-Charset and Accept-Encoding (TODO)

Basic Usage

Import all the objects from the negotiator module

>>> from negotiator import ContentNegotiator, AcceptParameters, ContentType, Language

Specify the default parameters. These are the parameters which will be used in place of any HTTP Accept headers which are not present in the negotiation request. For example, if the Accept-Language header is not passed to the negotiator it will assume that the client request is for "en"

>>> default_params = AcceptParameters(ContentType("text/html"), Language("en"))

Specify the list of acceptable formats that the server supports

>>> acceptable = [AcceptParameters(ContentType("text/html"), Language("en"))]
>>> acceptable.append(AcceptParameters(ContentType("text/json"), Language("en")))

Create an instance of the negotiator, ready to accept negotiation requests

>>> cn = ContentNegotiator(default_params, acceptable)

A simple negotiate on the HTTP Accept header "text/json;q=1.0, text/html;q=0.9", asking for json, and if not json then html

>>> acceptable = cn.negotiate(accept="text/json;q=1.0, text/html;q=0.9")

The negotiator indicates that the best match the server can give to the client's request is text/json in english

>>> acceptable
AcceptParameters:: Content Type: text/json;Language: en;

Advanced Usage

Import all the objects from the negotiator module

>>> from negotiator import ContentNegotiator, AcceptParameters, ContentType, Language

Specify the default parameters. These are the parameters which will be used in place of any HTTP Accept headers which are not present in the negotiation request. For example, if the Accept-Language header is not passed to the negotiator it will assume that the client request is for "en"

>>> default_params = AcceptParameters(ContentType("text/html"), Language("en"))

Specify the list of acceptable formats that the server supports. For this advanced example we specify html, json and pdf in a variety of languages

>>> acceptable = [AcceptParameters(ContentType("text/html"), Language("en"))]
>>> acceptable.append(AcceptParameters(ContentType("text/html"), Language("fr")))
>>> acceptable.append(AcceptParameters(ContentType("text/html"), Language("de")))
>>> acceptable.append(AcceptParameters(ContentType("text/json"), Language("en")))
>>> acceptable.append(AcceptParameters(ContentType("text/json"), Language("cz")))
>>> acceptable.append(AcceptParameters(ContentType("application/pdf"), Language("de")))

specify the weighting that the negotiator should apply to the different Accept headers. A higher weighting towards content type will prefer content type variations over language variations (e.g. if there are two formats which are equally acceptable to the client, in different languages, a content_type weight higher than a language weight will return the parameters according to the server's preferred content type.

>>> weights = {"content_type" : 1.0, "language" : 0.5}

Create an instance of the negotiator, ready to accept negotiation requests

>>> cn = ContentNegotiator(default_params, acceptable, weights)

set up some more complex accept headers (you can try modifying the order of the elements without q values, and the q values themselves, to see different results).

>>> accept = "text/html, text/json;q=1.0, application/pdf;q=0.5"
>>> accept_language = "en;q=0.5, de, cz, fr"

negotiate over both headers, looking for an optimal solution to the client request

>>> acceptable = cn.negotiate(accept, accept_language)

The negotiator indicates the best fit to the client request is text/html in German

>>> acceptable
AcceptParameters:: Content Type: text/html;Language: de;

Preference Ordering Rules

The Negotiator organises the preferences in each accept header into a sequence, from highest q value to lowest, grouping together equal q values.

For example, the HTTP Accept header:

"text/html, text/json;q=1.0, application/pdf;q=0.5"

Would result in the following preference sequence (as a python dictionary):

{
    1.0 : ["text/json", "text/html"],
    0.5 : ["application/pdf"]
}

While the HTTP Accept-Language header:

"en;q=0.5, de, cz, fr"

Would result in the following preference sequence (as a python dictionary):

{
    1.0 : ["de"],
    0.8 : ["cz"],
    0.6 : ["fr"],
    0.5 : ["en"]
}

(In reality, the q values for de, cz and fr would be evenly spaced between 1.0 and 0.5, using floating point numbers as the keys)

Combined Preference Ordering Rules

The negotiator will compute all the possible allowed combinations and their weighted overall q values.

Given that the server supports the following combinations (from the code example above):

text/html, en
text/html, fr
text/html, de
text/json, en
text/json, cz
application/pdf, de

And given the weights:

w = {"content_type" : 1.0, "language" : 0.5}

We can calculate the combined q value of each allowed (by both server and client) option, using the equation:

overall_q = w["content_type"] * content_type_q + w["language"] * language_q

So, for the above options and q values from the previous section, we can generate the preference list (as a python dictionary):

{
    1.5  : ["text/html, de"],
    1.4  : ["text/json, cz"],
    1.3  : ["text/html, fr"],
    1.25 : ["text/html, en", "text/json, en"]
    1.0  : ["application/pdf, de"]
}

It is clear, then, why the negotiator in the Advanced Usage section selected "text/html, de" as its preferred format.