Skip to main content

pywurfl - Python tools for processing and querying the Wireless Universal Resource File (WURFL)

Project description

WURFL and Python

#### by Armand Lynch (lyncha at users dot sourceforge dot net)

**Please note: As of pywurfl v7.0.0 all textual capabilities (including WURFL
IDs and user agents) are unicode strings and all textual parameters to the API
functions should be given as unicode strings.**

[pywurfl][1] is a [Python][3] language package that makes dealing with the
[WURFL][2] in Python a little easier. It contains tools that allow you to
retrieve objects that represent devices defined in the WURFL or manipulate the
WURFL device hierarchy by using a simple set of [API][9] functions or a
pywurfl specific [query language][4]. Also included within the package is a
[WURFL processor][5] class that provides an event based API that can be used
to alleviate some of the work when processing the WURFL sequentially.


pywurfl is Copyright 2004-2011, Armand Lynch (lyncha at users dot sourceforge
dot net) The code is free software; you can redistribute it and/or modify it
under the terms of the [LGPL][6] License (see the file LICENSE included with
the distribution).


Python >= 2.6

Required Modules

[Levenshtein Module][7] >= 0.10.1 is required for the user agent similarity

Optional Modules

[pyparsing][8] >= 1.5 is required if you want to use the pywurfl query language.


The pywurfl package contains a script that translates a WURFL
compatible XML file into a python class hierarchy that the pywurfl API can use
directly. The default name for the output file is Type the following
at the command line to produce it:

python wurfl.xml


In order to [patch][13] your wurfl.xml file, use the [WURFL XSLT Tools][14].
Then run wurfl2python on the patched file.

A quick usage example

After you have created the module, you can use the following code to
get a device object based on a user agent and print it to stdout.

from wurfl import devices
from pywurfl.algorithms import TwoStepAnalysis

user_agent = u"Nokia3350/1.0 (05.01)"
search_algorithm = TwoStepAnalysis(devices)

device = devices.select_ua(user_agent, search=search_algorithm)

# Print out the specialized capabilities for this device.
print device

That's it.

pywurfl API

To get an object that represents a device, you can use one of 2 methods of the
'devices' object imported from the module.

select_id(unicode, [actual_device_root=False], [instance=True])

This method returns a device object based on the WURFL ID provided.

If `actual_device_root` is `True` then the `select_id` method will return the
requested device or a device in its fallbacks if it is an actual device.

If instance is `False` then the `select_id` method will return a class object
instead of an instance.

select_ua(unicode, [actual_device_root=False], [search=False], [instance=True])

This method returns a device object based on the user agent provided.

If `actual_device_root` is `True` then the `select_ua` method will return the
requested device or a device in its fallbacks if it is an actual device.

The search argument takes an instance of [pywurfl.algorithms.Algorithm][12].
At this time, only four algorithms are provided: TwoStepAnalysis, Tokenizer,
Levenshtein distance and JaroWinkler. The TwoStepAnalysis algorithm is the
default algorithm used in Java and PHP APIs, you can find a description of it
[here][11]. The other three algorithms are older and probably shouldn't be
used in new programs.

If instance is `False` then the `select_ua` method will return a class object
instead of an instance.

### More API methods

There are a few more methods that you can use on the 'devices' object to
manipulate the device class hierarchy itself.

add(parent, devid, devua, [actual_device_root=False], [capabilities=None])

add_capability(group, capability, object)


insert_after(parent, devid, devua, [actual_device_root=False], [capabilities=None])

insert_before(child, devid, devua, [actual_device_root=False], [capabilities=None])





Here's an example

from wurfl import devices

# Add a new device
u'Mozilla/25.0 (X11; U; Linux i686; en-US; rv:25.0.0) Gecko/21000711 Firefox/28.1.5',

# Add a new capability group

# Add some capabilities to the teleporter group
devices.add_capability(u'teleporter', u'teleportation_device', False)
devices.add_capability(u'teleporter', u'distance', 20) # in km
devices.add_capability(u'teleporter', u'can_recover_from_errors', False)

# Add a new device overriding a default capability value
# Note that no devices had a 'teleportation_device' attribute until we added it
u'Mozilla/25.0 (X11; U; Linux i686; en-US; rv:26.0.0) Gecko/21000712 Firefox/28.1.6',
capabilities={u'teleportation_device': True})

# Add another group and capabilities
devices.add_capability(u'python', u'py_version', u'2.6')
devices.add_capability(u'python', u'py_heap_size', 0)

# Remove an unused group

# Not interested in tiff files

Check the [documentation][9] for more information.


It's also possible to serialize changes that you make to a WURFL compatible
XML file.

from wurfl import devices
from pywurfl.serialize import Serialize

# Remove some groups and their capabilities from the WURFL hierarchy


Search Algorithm Classes

The algorithms module contains four algorithm classes (TwoStepAnalysis,
Tokenizer, JaroWinkler and LevenshteinDistance). When instantiating any of
these classes, a callable object will be returned that can be used to search a
'devices' object with the provided user agent.

TwoStepAnalysis(devices, [use_normalized_ua=False])

In order to use the TwoStepAnalysis algorithm, you must initialize it with a
set of known devices. `use_normalized_ua` determines whether or not the search
algorithm requires that a normalized user agent be presented to it. The
TwoStepAnalysis algorithm does its own normalization internally, hence the
`False` specification here.

Tokenizer([devwindow], [use_normalized_ua=True])

The devwindow argument determines the upper limit of device matches before the
algorithm would return the generic device.

JaroWinkler([accuracy=1.0], [weight=0.05], [use_normalized_ua=True])

The accuracy argument determines the lower limit at which pywurfl will
determine if a user agent matches another. If no device can be found that
scores equal to or greater than accuracy, a generic device is returned. Valid
values are between 0.0 and 1.0


The LevenshteinDistance algorithm only has an optional `use_normalized_ua` argument.

from wurfl import devices
from pywurfl.algorithms import JaroWinkler, Tokenizer, LevenshteinDistance

tsa = TwoStepAnalysis(devices)
tokenizer = Tokenizer()
jarow = JaroWinkler()
levdis = LevenshteinDistance()

user_agent = u"Nokia3350/1.0 (05.01)"
device1 = jarow(user_agent, devices)
device2 = tokenizer(user_agent, devices)
device3 = levdis(user_agent, devices)

device4 = devices.select_ua(user_agent, search=tsa)
device5 = devices.select_ua(user_agent, search=jarow)
device6 = devices.select_ua(user_agent, search=tokenizer)
device7 = devices.select_ua(user_agent, search=levdis)

It's also very easy to define your own algorithm for use in pywurfl in case
the algorithms provided don't serve your needs. Just subclass the
pywurfl.algorithms.Algorithm class and follow the protocol.

Device Objects

The object returned by either `select_id` or `select_ua` is usually a Device

device = devices.select_id(unicode)

A device object has many attributes. The device id, user agent, `fall_back`
and `actual_device_root` attributes are exposed with the following attributes
of the device object:

device.fall_back # All devices have a fall_back attribute

Any capability that is defined in the WURFL becomes an attribute of the device
object. For example:


All attributes are converted into their respective Python types. For example:

device.ringtone # Attribute is boolean
device.preferred_markup # Attribute is a unicode string
device.rows # Attribute is an integer

You can iterate over a device object to select each capability and its
corresponding value. For example, to print out all capabilities of a device
object you can use the code below:

for group, capability, value in device:
print group, capability, value

Every device has a shared groups attribute which is a Python dictionary where
the keys are the capability group names as defined in the WURFL and the values
are lists of the capability names for that specific group.

for group in sorted(device.groups):
print group

### Classes and Instances

The API methods can also return a class object instead of an instance. What
wurfl2python does is produce a module that creates a single inheritance class
hierarchy of all WURFL devices. You can use this to your advantage if you want
to change the attributes of a device at run-time and have all of its
descendants represent that change.

# get an arbitrary device instance
device = devices.select_id(u'blackberry_generic_ver3_sub2')

# get the generic device *class*
gen = devices.select_id(u'generic', instance=False)

# modify the generic class
gen.teleportation_device = False

# since all devices inherit from the generic device, this will not raise
# an attribute error now
device.teleportation_device # == False

If you want to maintain the integrity of the class hierarchy, you should use
the add/remove/insert API methods on the 'devices' object mentioned above.

Query Language

The pywurfl package includes a query language that makes it easier to retrieve
a list of devices, WURFL IDs or user agents based on the capabilities of a
device. The best way to see what the query language looks like and what it can
do is with an example.

from wurfl import devices
from pywurfl.ql import QL # Import the query function generator

# Retrieve a function that will query the devices object
query = QL(devices)

# QL also adds a query method to devices (devices.query)

q1 = u"""select id where ringtone=true and rows < 5 and
columns > 5 and preferred_markup = 'wml_1_1'"""

for wurfl_id in query(q1):
print wurfl_id

# Let's look for some nice phones
q2 = u"""select device where all(ringtone_mp3, ringtone_aac, wallpaper_png,
streaming_mp4) = true"""

# Notice that we can also retrieve device classes
for device in devices.query(q2, instance=False):
print device.brand_name, device.model_name

# We can also use the methods on the capability types to refine our queries.
# Note that you should *always* quote the strings that are passed to functions
# and those that are used in comparisons.
q3 = u"""select ua where brand_name.lower()='nokia'"""
for ua in query(q3):
print ua

q4 = u"""select ua where brand_name.replace('No', 'Si').lower()='sikia'"""
for ua in query(q4):
print ua

q5 = u"""select ua where model_name.isdigit()=true and actual_device_root=true"""
for ua in query(q5):
print ua

# There are also a couple of regex methods (match and imatch) that were added
# to the string type to make those kind of queries possible. Use imatch to
# ignore case.
q6 = u"""select ua where brand_name.match('^No')=true"""
for ua in query(q6):
print ua

# and arbitrary nesting is supported
q7 = u"""select ua where brand_name.replace('Nokia', brand_name.lower())='nokia'"""
for ua in query(q7):
print ua

A full description of the query language is included in the documentation.

The WURFL Processor

The WURFL processor is a general class that walks a WURFL XML file and
executes hooks as specific events occur in a fashion similar to SAX. The best
way to understand the WURFL processor is to look at its [documentation][10].
For an example of how to use use it, look at the source for


To *Pau Aliagas*, *Gabriele Fantini* and *Michele Bariani*: Thank you for the
many patches, bug reports and improvements.

Comments and/or suggestions are appreciated.

[1]: "pywurfl"
[2]: "WURFL"
[3]: "Python"
[4]: "Query Language"
[5]: "WURFL processor"
[6]: "LGPL"
[7]: "Levenshtein Module"
[8]: "pyparsing"
[9]: "API documentation"
[10]: "WURFL processor documentation"
[11]: "TwoStepAnalysis description"
[12]: "Algorithm class API documentation"
[13]: "What is a patch file"
[14]: "WURFL XSLT Tools"

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pywurfl-7.2.1.tar.gz (237.3 kB view hashes)

Uploaded source

Built Distribution

pywurfl-7.2.1-py2.6.egg (76.2 kB view hashes)

Uploaded 2 6

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page