thadafinser / useragentparser Goto Github PK

Some providers use more than UA only:

• https://github.com/serbanghita/Mobile-Detect/blob/master/Mobile_Detect.php#L136
• https://github.com/WhichBrowser/WhichBrowser/blob/master/libraries/whichbrowser.php#L97

https://udger.com/products/local_parser

https://github.com/Ikimea/Browser
https://github.com/lennerd/vipx-bot-detect
https://github.com/JayBizzle/Crawler-Detect

If the running script is placed in a a subdirectory, it will break the loading of the installed providers.

For ex., including autoload like this
require __DIR__.'../vendor/autoload.php';

produce a fatal error:

PHP Fatal error: Uncaught exception 'UserAgentParser\\Exception\\PackageNotLoadedException' with message 'You need to install the package piwik/device-detector to use this provider' in ..

As the title says. If the package is still supported is composer v2 support planned?

To solve the generics problem from ThaDafinser/UserAgentParserComparison#19

we should create something abstract to filter default/generic values.

Currently we have those variations:
https://github.com/ThaDafinser/UserAgentParser/blob/master/src/Provider/AbstractProvider.php
https://github.com/ThaDafinser/UserAgentParser/blob/master/src/Provider/Wurfl.php#L64-L66

https://github.com/ThaDafinser/UserAgentParser/blob/master/src/Provider/BrowscapPhp.php#L103-L115
https://github.com/ThaDafinser/UserAgentParser/blob/master/src/Provider/Wurfl.php#L142-L166

Possible solution

We change the possible values of $defaultValues to

• multi dimensional array
  • general for all fields
  • specific filter for a single field (e.g browser -> name)
• do it with regex and not only direct string comparison

Example current

namespace UserAgentParser\Provider;

class PiwikDeviceDetector extends AbstractProvider
{
    protected $defaultValues = [
        DeviceDetector::UNKNOWN,

        // bot names
        'Bot',
        'Generic Bot',
    ];
}

Example new

namespace UserAgentParser\Provider;

class PiwikDeviceDetector extends AbstractProvider
{
    protected $defaultValues = [

        'general' => [
            '/^' . DeviceDetector::UNKNOWN . '$/'
        ],

        'bot' => [
            'name' => [
                '/^Bot$/',
                '/^Generic Bot$/'
            ]
        ]
    ];
}

http://www.handsetdetection.com/pricing

This package requires thadafinser/package-info but does not appear to use it. Am I missing something?

see https://github.com/nfriedly/whatsmyua.com

https://51degrees.com/

The used properties win32 and win64 are true only for Windows. If tyou want to detect also other platforms, use the platform_bits property.
The platform version can be detected with the platform_version property.
The device brand can be detected with the device_brand_name property.
The device model can be detected with the device_name property.

To detect the device properties you need the full_php_browscap.ini file.

Repo-Url: https://github.com/mimmi20/BrowserDetector

We have true 100% code coverage in case of unittests, but the providers are all mocked.

So we need to add some tests, which check with the lowest and highest supported version of the provider, if they still work.

Any good ideas, how to implement this right?

https://github.com/sinergi/php-browser-detector

https://github.com/sam-media/wurfl-api

or

https://github.com/mimmi20/Wurfl

BrowscapPhp:

• model: "Windows Desktop"
• model: "Linux Desktop"

SinergiBrowserDetector:

• model: "Windows Phone"

Wurfl:

• model: "SmartTV"
• browser: "Java Applet"

Device model is mixed with tablet/phone and brand information (so remove for now!)

https://github.com/zsxsoft/php-useragent

It would be great if information about the bit-width of the operating system could be included. Several of the providers give this information, so it shouldn't be too hard to do :)

add integration tests with special characters which cover the escaping works.

Add the type + bot detection

field: user_agent_type

browser
mobile
ebook_reader
crawler
analyser
tool
tv
mediaplayer
gamesystem
library
unknown

I've created a separate Composer package the tests I run for WhichBrowser. The package is called whichbrowser/testrunner and includes the runner.php script and all the yaml files. It is automatically updated whenever there is a new release of WhichBrowser.

If you want I can probably create a pull request for the actual extracting of the UA strings.

But it is also possible to extract it by using the runner.php script from the command line:

php vendor/whichbrowser/testrunner/runner.php list

or

php vendor/whichbrowser/testrunner/runner.php list television

or

php vendor/whichbrowser/testrunner/runner.php list mobile tablet television gaming

The target of this project is wherever one or more of the following:

• create an overview/matrix:
  • which Providers have a good quality (generation time, valid data, ...)
  • use different UA databases to generate those matrixes
• maybe release this as a counter part for UA parsing, like https://github.com/geocoder-php/Geocoder

...for now it's just a prototype:

• install it with composer
• execute php tests/test.php for the matrix

//cc @sgiehl could i use the Piwik-DeviceDetector fixtures (https://github.com/piwik/device-detector/tree/master/Tests/fixtures) to generate matrix with all provider results?

WhichBrowser adds the nickname of the OS release before the version number. So the version is "Yosemite 10.10" and not just "10.10". UserAgentParser interprets this as "0.10".

See for example:
http://useragent.mkf.solutions/?userAgent=Kodi%2F14.0+%28Macintosh%3B+Intel+Mac+OS+X+10_10_3%29+App_Bitness%2F64+Version%2F14.0-Git%3A2014-12-23-ad747d9-dirty

This also happens with many other libraries with versions like "XP" for Windows XP. It is interpreted by UserAgentParser as "0".

See for example:
http://useragent.mkf.solutions/?userAgent=Microsoft+Office%2F14.0+%28Windows+NT+5.1%3B+Microsoft+Outlook+14.0.4536%3B+Pro%3B+MSOffice+14%29

The function getParts() should probably skip over any non digit characters in the version string before splitting it into parts.

Also when setComplete() is called, it first stores the original string as the $complete property, then hydrates all the parts. But the hydrateVersionParts() function calls setMajor(), setMinor() and setPatch() which in turn call calculateComplete() which overwrites the original $complete property that was initially set. This means the original string calculated by the parser is lost.

https://github.com/endorphin-studio/browser-detector

http://www.detectright.com/

crossjoin/Browscap is a UA parser, who uses the browscap.ini too, but is faster and requires lower amout of RAM. Actually only a file cache is supported. The browscap-php library has adopted a lot of features from it.

Hello @mimmi20 (sorry to ping again)

i start now to finalize the BrowscapPhp provider.

Could you maybe check if i missed here something, or done something wrong in the implementation?
https://github.com/ThaDafinser/UserAgentParser/blob/master/src/Provider/BrowscapPhp.php

For testing i create a small browscap.ini file and used it in the tests.
https://github.com/ThaDafinser/UserAgentParser/blob/master/tests/fixtures/browscap.ini

Is this fine or do i need to mock i further?
https://github.com/ThaDafinser/UserAgentParser/blob/master/tests/unit/Provider/BrowscapPhpTest.php#L41

Use instead reflection like here: https://github.com/ThaDafinser/UserAgentParser/pull/20/files

Hello,

Thank you for this great package.

I notice one of your dependencies, piwik/device-detector, is abandoned: https://packagist.org/packages/piwik/device-detector
Their maintainer advices to use https://packagist.org/packages/matomo/device-detector instead.

Do you consider update your package with it?

Thank you
Jourdan

Currently requires PHP ~5.6|~7.0. It'd be nice if we could use this with the latest PHP 8 release. There ain't many breaking changes, shouldn't be complicated (if this project is still alive).

Instead of doing this

if ($this->isRealResult($deviceRaw->model, 'device', 'model') === true) {
    $device->setModel($deviceRaw->model);
}

Do this and return null if it's not a valid value

$device->setModel($this->getRealResult($deviceRaw->model, 'device', 'model'));

There are some providers where the value is not always available - there we still need a isset() if check before that

@NielsLeenheer i added now unitTests for all providers, but WhichBrowser fails on PHP7.

I excluded now PHP7 temporary, that the tests can pass here.

Do you have any plans for PHP7 compability?
https://travis-ci.org/ThaDafinser/UserAgentParser/jobs/90565282#L284

Each provider has it's own device types..

BrowscapPhp

https://github.com/browscap/browscap/wiki/Resource:-Device-Database#device_type

Console
TV Device
Tablet
Mobile Phone
Mobile Device
Desktop
Ebook Reader
Car Entertainment System
Digital Camera

PiwikDeviceDetector

https://github.com/piwik/device-detector/blob/master/Parser/Device/DeviceParserAbstract.php#L42-L54

    protected static $deviceTypes = array(
        'desktop'               => self::DEVICE_TYPE_DESKTOP,
        'smartphone'            => self::DEVICE_TYPE_SMARTPHONE,
        'tablet'                => self::DEVICE_TYPE_TABLET,
        'feature phone'         => self::DEVICE_TYPE_FEATURE_PHONE,
        'console'               => self::DEVICE_TYPE_CONSOLE,
        'tv'                    => self::DEVICE_TYPE_TV,
        'car browser'           => self::DEVICE_TYPE_CAR_BROWSER,
        'smart display'         => self::DEVICE_TYPE_SMART_DISPLAY,
        'camera'                => self::DEVICE_TYPE_CAMERA,
        'portable media player' => self::DEVICE_TYPE_PORTABLE_MEDIA_PAYER,
        'phablet'               => self::DEVICE_TYPE_PHABLET
    );

UAParser

Can be some of those: https://raw.githubusercontent.com/ua-parser/uap-core/master/regexes.yaml

Generic Tablet
Generic Smartphone
Generic Feature Phone
Spider

WhichBrowser

https://github.com/WhichBrowser/WhichBrowser/blob/master/libraries/whichbrowser.php#L30-L48

define ('TYPE_DESKTOP', 'desktop');
define ('TYPE_MOBILE', 'mobile');
define ('TYPE_DECT', 'dect');
define ('TYPE_TABLET', 'tablet');
define ('TYPE_GAMING', 'gaming');
define ('TYPE_EREADER', 'ereader');
define ('TYPE_MEDIA', 'media');
define ('TYPE_HEADSET', 'headset');
define ('TYPE_WATCH', 'watch');
define ('TYPE_EMULATOR', 'emulator');
define ('TYPE_TELEVISION', 'television');
define ('TYPE_MONITOR', 'monitor');
define ('TYPE_CAMERA', 'camera');
define ('TYPE_SIGNAGE', 'signage');
define ('TYPE_WHITEBOARD', 'whiteboard');
define ('TYPE_GPS', 'gps');
define ('TYPE_CAR', 'car');
define ('TYPE_POS', 'pos');
define ('TYPE_BOT', 'bot');

Woothee

class DataSet
{
const DATASET_CATEGORY_PC          = 'pc';
const DATASET_CATEGORY_SMARTPHONE  = 'smartphone';
const DATASET_CATEGORY_MOBILEPHONE = 'mobilephone';
const DATASET_CATEGORY_CRAWLER     = 'crawler';
const DATASET_CATEGORY_APPLIANCE   = 'appliance';
const DATASET_CATEGORY_MISC        = 'misc';
}

Wurfl

http://web.wurfl.io/

form_factor
Desktop
App
Tablet
Smartphone
Feature Phone
Smart-TV
Robot
Other non-Mobile
Other Mobile

Yzalis

https://github.com/yzalis/UAParser/blob/master/src/UAParser/Result/DeviceResult.php#L111-L113

mobile
tablet
desktop

The required vendor section should be moved to optional https://github.com/ThaDafinser/UserAgentParser/blob/master/composer.json#L51

• move them to suggested
• throw a new exception when the package is not included

https://github.com/Ikimea/Browser

https://github.com/serbanghita/Mobile-Detect

They are working on a complete detection here...wait for this?
https://github.com/serbanghita/Mobile-Detect/tree/devel-3#getting-started

mimmi20/Wurfl#15 (comment)

Repo-Url: https://github.com/Wolfcast/BrowserDetection

When a UA is tested on the web, the UA is cutted if the UA conains a ".

Sample UA: Mozilla/5.0 (Linux; Android 5.1; Alba 7" Build/LMY47I) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.93 Safari/537.36

Hello together,

i create this prototype project in order to try out the different UserAgent parsers available for PHP.

First i wanted to try if it's currently possible to create an abstract result set, like it's done for GeoIP already https://github.com/geocoder-php/Geocoder
Second i wanted to have a Chain Adapter on top it to get results from different vendors to get better results https://github.com/ThaDafinser/UserAgentParser/blob/master/bin/generateMatrix.php#L45-L53

Last but not least i wanted to compare the different implementations - how complete and good are they?

At this stage the results of the comparison are not completed, since some vendor adapter will be not complete or have errors.
But i still wanted to share those results to you all, to get (hopefully) some feedback about your Adapter implementation and get improvements https://github.com/ThaDafinser/UserAgentParser/tree/master/src/Provider

See the (current not correct or complete) results

Download the HTML files from here and open it just in a browser https://github.com/ThaDafinser/UserAgentParser/tree/master/data/results

It should look like this screenshot.
Click on "toggle" to see the original data and the result detail

Note about used datasets to generate the matrix

I currently did not used datasets from each vendor, so the comparison is not fair currently. (and of course missing convertions or errors)

• browscap/browscap-php the resources have placeholders...need to check how they can we replaced accordingly example file
• donatj/phpuseragentparser
• piwik/device-detector
• ua-parser/uap-php waiting for a release ua-parser/uap-core#114
• whichbrowser/whichbrowser found no test data list.
• woothee/woothee waiting for a release woothee/woothee#19
• yzalis/ua-parser

Better would be an additional independent list...need to find a good source first

Next steps

• get some feedback, about the provider adapter
• complete and correct the provider adapters to get better results and a true comparison
• refactor all code...it's ugly written...just a prototype
• seperate UserAgentParser and the Matrix generation
• release initial UserAgentParser
• add more test datasets for the matrix
• create a better comparison markup...the current on is not so "nice"
• release a Matrix generator website

I hope it's ok to ping you all ...if not just tell me and i wont do it again 👍
@mimmi20
@donatj
@sgiehl
@lstrojny
@NielsLeenheer
@yuya-takeyama
@blaugueux

You can try to add nice dumper (https://packagist.org/packages/raulfraile/ladybug) for your live demonstration: http://useragent.mkf.solutions/

https://github.com/bcit-ci/CodeIgniter

They have this class and config:
https://github.com/bcit-ci/CodeIgniter/blob/develop/application/config/user_agents.php
https://github.com/bcit-ci/CodeIgniter/blob/develop/system/libraries/User_agent.php

There are a lot of free and paid cloud services around.

Implement at least 1 or 2 of them to get started

In order to be able to deploy the library in various environments, it would be better to check if the primitive functions/classes of the different providers are present instead of checking if the package has been deployed in a specific vendor directory.

Even if it's a best practice to put external libraries in a vendor directory, all projects don't respect it and can't deploy your great library.

I created a pull request for the DonatjUAParser Provider : #60

Let me know if it makes sense to you

We remove currently some generics in various providers, to prevent the user from "false positive" detection results.
This plays also really well with the chain, since only then the next provider has the chance to give "good results"

But maybe some need to disabled this and want the original results at all time?

Is this project still alive? On the test page I see tat the underlying data was not update since over two years... what looks like a "project dead" indicator to me...

Hi,
I a using the latest version of this library. Now theres an issue. Not sure if its a bug.
I already have this library installed PiwikDeviceDetector but i keep getting the following exception error. You need to install the package piwik/device-detector to use this provider. I have done all i could to debug but i can seem to find a solution. The only one is to comment out the following:
public function __construct(DeviceDetector $parser = null)

    {
        if ($parser === null && ! file_exists('vendor/' . $this->getPackageName() . '/composer.json')) {
         Icommented this line - // throw new PackageNotLoadedException('You need to install the package ' . $this->getPackageName() . ' to use this provider');
        }

        $this->parser = $parser;
    }

I moved the requested library into the vendor/ directory as its supposed to, but it didn't solve either.

I am on windows and it doesn't allow this kind of file naming structure piwik/device-detector, because i believe this is the directory that is being looked, and since it can not be located, an error is thrown.

Any suggestions?