web2py zen

April 27, 2010
  • Web2py is a WSGI compliant enterprise web framework .
  • Enterprise” here means mainly one thing: we DO NOT CHANGE THE API because professional users who works in teams and on long term projects must count on stable API add documentation. “Enterprise” do not mean we focus on very large enterprises, we probably focus more on small and medium size ones, nor it means we have every possible “enterprise” feature. We are very straightforward about the web2py features. “enterprise” here sends a message: web2py is not developed by a bunch of kids who try to follow the latest trends and never deliver a stable product. web2py is built by experienced developers who know how to find a balance between usability, stability and features.
  • WSGI server parameters can be passed as command line options
  • Always promises backward compatibility
  • Web2py is not planned to move to python 3 any time soon because
    it will break backward compatibility of all applications and because
    database drivers do not work on 3. Moreover web2py apps would no
    longer work on GAE.
    The main and probably the only issue is with strings becoming unicode
    strings by default. It would not be difficult to change web2py to
    work with 3.0 and it was designed following all the other 3.0
    guidelines but there is no way to make the apps backward compatible.
  • you can make your web2py interact with your cgi scripts :
    You can run web2py on a port (say 8000) and your CGI script on a
    different port (say 8001) using a different web server.
    In this case the web2py program can call the CGI scripts pass parameters and receive response. Example

    import urllib

    You can read more in the urllib and urllib2 documentation).
    You cannot call the cgi scripts from web2py without going over HTTP.
    This is not a web2py limitation, this is a CGI one. In fact CGI
    assumes that every CGI script is executed in its own process and thus
    has its own folder, environment variables and stdout/stdin. Therefore
    a CGI script is not thread safe and cannot be executed inside a
    thread (threads shares folder, environment and IO streams). web2py
    uses threads.

  • web2py does support IF_MODIFIED_SINCE and PARTIAL REQUESTS out of the box.
    You can add any header you want including ETAGS.
  • There’s a routing system in web2py which unfortunately doesn’t work on application level basis (yet)
    Don’t use the routes module to enforce the use of SSL on a few URLs
    it would not be safe. If you are behind a proxy web2py does
    not know whether the original request was over https or not. The
    method of using X_FORWARDED_FOR has known vulnerabilities. You should use apache to force SSL.
  • Web2py runs on :
    • runs on python >= 2.4 but two things to note:
      • if you save data using CRYPT() validator used with password fields in 2.4 you will not able to read them in 2.5 or 2.6 so be careful python 2.y doesn’t have hashlib used by CRYPT() validator
      • There is not uuid module in Python2.4 , You could make one that will satisfy web2py. You can make it as a module ofcourse
        ## uuid.py 
        import random,time 
        def uuid4(): return '%s%s" % 
        ## end file 
    • web2py runs on jython.
    • No support for stackless python [yet] and using it actually causes errors
  • Web2py philosophy is not to use separate and un-integrated modules.In fact for some of the web2py components there is a better module out there, the point is that the web2py modules are designed to work together while the alternatives are not.
    You will not appreciate this until you try it. Here is an example: Pygments is a much better general purpose syntax highlighter than the web2py’s one, nevertheless web2py’s one can highlight web2py code, create clickable links from web2py keywords to the web2py online documentation, is faster and fits in 10k. If you need to syntax highlight Java or PHP code you can still easy_install and use Pygments. Although you can, the web2py modules are not designed to be taken out and used separately.
  • Web2py’s philosophy is : if you want scalability, it is best to store as much as possible on disk and use a centralized database when you need transactions and joins.
    This is why web2py by default stores sessions and files on disk.
  • Web2py vs others
  • From Django to web2py
  • From TurboGears to web2py
  • From Rails to web2py
  • Web2py for J2EE programmers
  • from php to web2py
  • web2py can run on a cluster environment like OpenVMS.In fact This is not a problem on web2py as long as all processes see the same
    filesystem. If they do not see the same filesystem you will not be able to use sessions, tickets and upload files (unless you do some
    Check this section in the manual book
  • web2py suits agile development, with an interactive shell used for testing, debugging and experimenting with your applications.
    you can run interactive shell that will run in ‘ipython’ 😀 by doing something like:

    cd web2py
    python web2py.py -S application_name -M
    # -M means to import also all the models in the application

    and congratulations, you can do something like:

    print db.tables

    to get list of all database tables used by this application.
    You can ofcourse do more .
    In fact one of the interesting stuff that you can make is:

    python web2py.py -S application_name -M -R file.py

    the interesting thing is that file.py can include code that will be executed as if it were a part of a controller.
    If the application specified by -S doesn’t exist, shell offers you to create it. isn’t that cool? 😀
    you can also import your custom modules and use it inside shell:

    exec('from applications.%s.modules.my_custom_mod import *' %(request.application))

    Your module will run in its own context, so if you want to do some operations like manipulating request, or some database stuff, you have to pass those objects to your module functions or class explicitly when calling them, this implies of course that you define your functions or classes with that in mind

    class DoThat(object):
        def __init__(self, request, db, cache, T):

    To execute a model file of another application:

    You can use the previous mechanism to execute other application controllers but there may be issues in importing controllers in other ways since they may contain authorization code outside functions. and it’s not a good idea to import controllers .

    you can even run the shell without ipython if you just don’t want it
    simply pass the -P (uppercase) argument when trying to run the shell.
    In the shell you have to explicitly db.commit() or db.rollback() your transactions.

  • Web2py actions support doctesting, but becareful …. doctets are not thread safe. This is not because of web2py. All
    testing libraries in python are not thread safe because they redirect
  • In fact The web based administrative interface can only test
    doctests in controllers. The shell has a -test option that will run
    any test you like and this does not require web2py running.
  • Web2py admin interface doesn’t allow remote access unless it is secure, other wise connection is denied.
    You can access it locally though without these restrictions, only admin password is required.
    Admin should be used for development and should never be exposed.
    If you want to expose admin:

    • make sure it goes over HTTPS
    • always logout when done
    • Edit admin/controllers/default.py and at the top write:

      The secure cookies are not enabled by default because admins usually access the admin application through their lcalhost, but it you’re tending to access it remotely you should add secure cookies.

    • Moreover all the */appadmin/* pages should go over https.
    • Another way , that may or may not be easier , is running two
      instances of web2py on different ports. One is exposes by without a
      password and so no admin. One is not exposed, runs only on localhost
      and has a password for admin. Then you connect to the second via a
      ssh tunnel. This is very secure, easy to setup, and you can do
      everything you do now.

    If somebody intercept your communications and steals the session
    cookies for admin, they can become admin.
    These rules apply to web2py as well as to every system that does
    authentication. even gmail has this problem.
    To make things even more secure admin session is default to expire after 10 minutes.

  • Web2py forms including SQLFORM, SQLFORM.factory, all inherit from class FORM and you access their components via form.components:


    or you can deal with forms as list of lists and edit its components according to your needs [just print a form and youi’ll know how to access it using this method.



    It is never a good idea to access form components since the internal
    implementation may change in the future (Although there ‘re no plans to do so).
    you can access them that way if there’s no other way to do so.

  • Web2py allows you to byte code compile applications and distribute them .
    you can even distribute your applications in a compiled binary files.
    This way, one of the advantages is that templates are only parsed
    once when the application is compiled and this makes web2py apps fast
    If you are using the binary distribution of web2py. It ignores
    your python installation because it uses its own.
    This may cause some errors when trying to import some modules.
  • Web2py is an MVC frame work , while all models are executed for every request.
    Technically a model file is not a module because they access symbols
    that are only defined in the context they are intended to be executed
    (file request and response). So importing a model explicitly can
    create problems.
    Put stuff in models if:

      Have something to do with accessing db data or helpers for visualizing db data
      They consist of a one single file with no dependences
      The code only makes sense inside a web2py app and I would not be reusing it outside web2py

    To create your tables, you may either do it manually or use a web based tool for this purpose.
    It’s able to generate web2py’s DAL compatible code.
    This great tool can’t be a part of web2py framework itself because its license conflicts with web2py’s license that have an exception to ship the binary web2py code with your web applications without the source code as long as you didn’t touch the source code .
    You can’t use controller functions in another controller because technically controller is not a module.
    you need to gather all your common functions,and put them inside a model or module, or even in gluon.contrib so that they can be accessible to all your applications without no problem

    When a request comes to a controller the requested controller file is loaded this is not achieved using reload() but exec() which executes the controller file and only takes about 0.03 seconds only.
    Using exec() also has the advantage that the framework does not
    need to monitor modules for changes before reloading (like Pylons)
    which may cause a slow down for apps with multiple files. web2py only
    execs those controllers that are requested.

    For framework like django, application is just the reuse unit, but not the execution unit. And in web2py app is execution unit, but
    not reuse unit. For example, I have user management app, and I want to reuse it, in Django, I can import the user model, user controller
    functions, and use template also. But in web2py I can’t directly
    import and reuse them. Only I can think out is copy the file which I
    want to use to target app. There are many ways to combine different apps together, but the usage of exec make it difficult. And how to orginaze the development unit and execution unit is a design issue. If we only care about the execution unit, so how to use others app functionalities in our app? Only through xml-rpc? Or copy the files into the application.

    In the contrary:

    There are also other problems with modules related to the search path
    and different applications may end up with conflicting modules.
    There may also be conflicting classes because all modules would be
    imported in the same context and if two modules have classes with the
    same name they would overwrite each other. This is not an issue if one just made an import like:

    from app1 import Class1
    from app2 import Class1

    Now you can use app1.Class1 or app2.Class1 wihtout any problems.

    In web2py exec() behavior is similar to:
    For example:

    class A: 
    print A.name 
    class A: 
    print A.name 
    exec('print A.name',{},c) 

    which is nice: the two classes called A ignore each other. If you use
    modules you end up with a lot of potential conflicts between multiple
    apps. In Django for example you would have to careful in designing
    apps that do not conflict. In web2py you do not need to think about
    it. You have to do a little of extra work to share stuff but at least
    you know for sure that no unwanted sharing takes place.

    the ‘exec’ have a good thing , that is the module can be GC by python afte request is over.

    sometimes you need to call some controllers functions and pass to them some variables[request variables] to make some functionality based on the values of these variables.In this case you may face one or more situations like those:

    • If you are doing a redirect to a controller function that is
      supposed to receive user input, you use :

      URL(r=request, c='controller_name', args=[...], vars=dict(..=.., ..=..))

      args are list why ? It allows you to do ‘/’.join(request.args) and rebuild the original string easily.

      vars is a dictionary why ? because this is the best fit for it key/value pairs .

    • If you want to call a controller function that is not intended to
      receive user input and is in the same controller, save the variables
      in session and retrieve them later.
    • If you want to call a controller function that exposes some
      functionality but has no state (so no use of sessions), expose it via
      xmlrpc .
      You can use urllib2.urlopen to call other functions, so you can do something similar to the following
    • example:
      Warning: This’s just an example made for test purposes to clarify things.
      Assume application name is ‘t2’

      def index():
          return dict(url=A('click here', _href=URL(r=request, f='another', args=['a'], vars=dict(b='b'))))
      def another():
          print response._caller(another)
          import urllib2, urllib
          if request.args(0) and request.vars.has_key('b'):        
              test_values = {'hidden_var' : 1}
              data = urllib.urlencode(test_values)        
                   req = urllib2.urlopen(''%data).read()             
                   return dict(x=XML(req))        
              except Exception, e:
                   print 'Error:', e
      def third():
          form = SQLFORM(db.my_table, request.vars.hidden_var)
          if form.accepts(request.vars, session):
          return form
  • Web2py includes a separate folder for your custom modules that you want to use, in this way you can use your own python libraries in your web application.
    things to note when using modules:
    Put stuff in modules if:

      The do not require access to request, response, cache, session and thus can be used with ot without web2py .
      They consist of multiple files
      I need them only in some (but not all) controller functions.
  • whenever you want to import something from a module don’t do something like:

    • from applications.yourapp.modules.Custom_Validators import IS_AAA

      making your import dependent on the name of the app is not a good idea, instead use something like:

      exec('from applications.%s.modules.validators import IS_AAA' %  
  • modules can’t see session, request, response objects of web2py, if you want to make operations using one or more web2py objects, you’ve to make a module function that takes those objects as its parameters.
  • The first time you need to run web2py with all its files. After that
    you can remove any installed application you want including admin,
    examples and welcome.

    Nothing breaks except that you can no longer visit pages that you do
    not have. You only get an internal error if you try to do that.

    welcome.tar is the scaffold application. That’s the one file you
    should not remove.

    Notice that if you remove admin you can no longer login as
    administrator and the database administration (appadmin) will be
    disabled for all you apps, unless you adit the appadmin.py files and
    change the authentication mechanism.
    You may need to rename your applicationa ‘init’, so that it can be the default application (the one that starts, once you start web2py) .

  • In web2py (and many web frameworks) each request is served by a different thread. This if done for speed. The different threads share
    the same memory space and the same environment variables. By default, you cannot execute in a thread any function or module that attempts to change environment variables or gain exclusive access to memory or other OS resource without mutex locking.
    For example you cannot do os.chdir(‘somedir’) else all threads would
    change folder and web2py would behave weird.
    There are ways many around this. Run processes, not thread; perform a
    mutex lock; create a background process that access exclusively the
    thread-unsafe module.
    So be careful :
    When ever making a web application using web2py, DON’T CHANGE THE CURRENT WORKING DIRECTORY , DON’T USE os.chdir() in your web applications.
    Changing working directory may cause your application to fail and moreover it’s not thread safe.
    To know more about this issue and more, check this interesting topic in the web2py user group:
    Very, Very interesting discussion

  • web2py can serve xmlrpc requests (even to itself, without deadlocks).
  • when uploading files using web2py, they’re renamed by adding to the name a random string .
    This is important for situations like :
    If the real name conflicts with another file or
    contains characters not supported by the file system
  • Web2py provides an easy way to access cookies:
    The value can be a string or a Morsel Object .
    Check that for more info
    Do Not Use cookies your self, and store everything always server side.
    Let web2py manage session cookies for you, it’s done automatically and you need not to worry about any thing.
  • web2py provides a good mechanism for internationalization through the T object that supports the lazy translation mode and immediate mode
    More information about T
    T returns an object that can be serialized as a string
    web2py does not allow you to do :


    but it does allow :

    T("bla %(name)s",dict(name='Tim')) 


    • You need to be careful concatenating strings. String concatenation is one of those “no-no”s of internationalization.
      The reason for this is that different languages have different sentence
      structures.For instance, some languages negation is before a word and other’s negationis in other parts of the word.Many slavic languages, like Czech, have the interesting feature that word order isn’t particularly important, so for Pete’s sake don’t start with one of those as your base language that everything’s translated too.The worst example is something like German where a negation can happen in two different parts of the sentence, and you just get screwed concatenating that string.If you’re gluing together strings, or you need to put dynamic content in an internationalized string, it’s really best if you put it in a string that’s parameterized: “hello %(name)s”, dict(“name”: “Tim”)
  • Web2py makes a good use of decorators to make your life easier
    for example:

    def do_it():
        # This won't be executed unless user is logged in and belongs to the 'admin' group

    Web2py philosophy is not to use decorators for validation since decorators are associated to a function (i.e. a page) while validation is associated to a form (and a page can have multiple forms).
    A form should be an object that knows how to validate itself.

  • Web2py mixes between the database tables and forms that should enter data into those tables; by using SQLFORM you can deal with a database table and enter data into it directly.

    #in a model:
    db.define_table('my_table', Field('my_field', 'string' ,requires=[....]))
    #in a controller
    form = SQLFORM(db.my_Table)
    • “requires” is assigned one or more validators [they work on form level not database level], i.e IS_NOT_EMPTY()
      To make a mix between validators you just add them to the list assigned to “requires”
    • A validator is a two way filter.
      user_input -> validator -> data_for_db OR error
      user_output <- validator <- data_from_db
    • using multiple validators at the same time makes sense and it’s just as if they’re ANDed together.
    • AND makes sense because it is like piping validators. When data
      goes in it goes through one validator after another and has to pass
      them all. If it misses one that one returns the error message. When
      data comes out it goes through the same validators in reversed order.
    • OR is problematic because if user input does not pass any validator,
      which error should be generated? The error is different that the
      errors associated to the individual validators.
      Moreoever when data comes out which validator should do the
      formatting? Consider this example:

      IS_DATE('%Y-%m-%d') OR IS_DATE('%d-%m%Y')

      This cannot be.
      To understand this well, consider this example:
      consider IS_DATE
      it has a constructor a __call__() method and a formatter method,
      when data is validated in a form it is filtered by IS_DATE.__call__
      and a string, say ’01/01/2007′ is converted to a datetime object.
      when data is presented into an edit SQLFORM or a SQLTABLE, formatter , it is called and the datetime object is converted back to ’01/01/2007′.
      The constructor takes an argument that specifies how the formatter
      can be done.
      If you have a list of validators and they all pass the the __call__
      method, they are called in the order of the list and the formatter methods are called in reversed order.
      Not so many validators use formatter method, but the problem is that : wil this continue for ever ?
      People can write validators that take formatter method and thus using OR will be problematic.

    • The bottom line is that in order to implement an OR
      some combination should be allowed and others should be forbidden. This will make individual validators more complex objects that they are now and their behavior less intuitive.
    • The only Validator to use OR is : IS_EMPTY_OR()

  • To build a great and rapid development environment, dictionaries returned by web2py’s actions [controller functions] are automatically rendered by a generic view , without the need to make a view file for this action .
    This’s done as you may know using a generic view files that use the
    response._vars which holds the returning values of the action [function]
    and is rendered using

  • Web2py can supports both POST and GET variables together sent with each other and you can extract each set.

     request.post_vars, request.get_vars

    Web2py takes care of the situations like when MS-Windows user input data which will be sent with ‘\r’ in the end of your input [‘\n\r’ is the line terminator in windows but not in linux .. In linux it’s just ‘\n’] .any way user input will have to be filtered using something like:

    string.replace('\r', '')
  • One another greatest thing is that you may not create a view file for every action you want to expose:
    • you may edit the generic view file [generic.html] which includes the following lines of code :

      {{extend 'layout.html'}}
      You should not modify this file. 
      It is used as default when a view is not provided for your controllers
      <button onclick="document.location='{{=URL("admin","default","design",
      <button onclick="jQuery('#request').slideToggle()">request</button>
      <div class="hidden" id="request"><h2>request</h2>{{=BEAUTIFY(request)}}</div>
      <button onclick="jQuery('#session').slideToggle()">session</button>
      <div class="hidden" id="session"><h2>session</h2>{{=BEAUTIFY(session)}}</div>
      <button onclick="jQuery('#response').slideToggle()">response</button>
      <div class="hidden" id="response"><h2>response</h2>{{=BEAUTIFY(response)}}</div>

      this stuff is useful in development time since it gives you buttons to check session variables, request and response but in production you may just make it look like:

      {{extend 'layout.html'}}

      and bingoo !!!!!! you get every thing almost automated and you don’t need to add one view file per action .
      Ofcourse this is not practical all the time , and in real-life you always have some thing to add to the view but sometimes you don’t have 🙂
      and you can rely on the generic.html then to render your view.

    • Another situation is that you can make use of one view file in different actions in different controllers … wooooow !!!!
      look at this example :

      def test():
           response.flash=T('Welcome to web2py')
           response.view='default/view.html'  ### this indicates the view file
           return dict(message=T('Hello World'))
      #Here's the same version of the function but more suitable for caching the view .
      @cache(request.env.path_info, time_expire=5, cache_model=cache.ram)
        def test():
           response.flash=T('Welcome to web2py')
           return response.render(dict(message=T('Hello World')))
    • Web2py views are different than django’s and Macko templates, in web2py : If a variable is accessed in the view, if has to be defined.
      in django, it just ignore it
      so in your view in web2py, if you’re returning a form that may sometimes have no value, you’ve to make additional check in the view:

      {{if reponse._vars.has_key('form'):}}
      # or
      {{if form:}}
    • One another interesting feature that suits the rapid development environment is that you can have multiple views per action(controller function), that is it , you can render output as html, rss, xml.
      You can do this by either depending on the generic views files “generic.xml, generic.xxx, …” that will render different actions based on the extension you provide in the url or you can make your own views, ending with different views for one action like [index.html, index.xml, ….]
      You can just open up one generic view file and follow the same behavior in your custom view files for other extensions other than html

      def index():
         form = FORM(INPUT(_name='Name'), INPUT(_type='submit'))
         if form.accepts(request.vars, session):
         return dict(form=form)

      Now trying to access page as index.html will return the form as expected, but trying to access index.xml will return something like:

      <form action="" enctype="multipart/form-data" method="post"><input name="Name" type="text" /><input type="submit" /><div class="hidden"><input name="_formkey" type="hidden" value="43ab2943-2f1c-4ae6-bdaf-73e3224967ee" /><input name="_formname" type="hidden" value="default" /></div></form>

      Can you see the beauty of this? You can use similar behaviors for debugging without trying to print the form itself in your code.
      Faster right ? easier right ? 😀

  • No daemons [except for cron]
  • Daemons tend to need starting/stopping and affect application portability.
  • If you’ve long running processes run by a thread , server would kill it any way. you need your daemon to run it
  • It is not a good idea to store the entire file in the db. better to store name only. That’s why web2py saves uploaded images to the directory called ‘uploads’ and the name in the db.
    In fact looking into gluon.sql.py, you’ll find something like:

    if fieldtype=='blob': 

    web2py is not currently using blobs by the book. The reason is that
    blob values need to be escaped differently by different database
    backends and in some cases this is not well documented. web2py avoids
    the problem by storing the base64 encoded data, thus using the blob
    as a text field.
    The current behavior has both advantages (simpler code, faster encoding, humanly readable SQL string all the time, works with databases that may not have a blob type) and disadvantages (the storage is increased by 30%, you cannot search with LIKE,, in a blob).

  • Field('blobf','blob')  # Not Recommended
    Field('uploadf','upload') # Reommended
  • There’re 3 reasons for using error tickets for displaying errors:
    • Not to separate between production and debugging modes.
    • It is conceptually possible to cause bad errors that Web2py
      may not be able to read the file that contains the error and even
      requires web2py restart.
    • No configuration files for the framework itself, but you can make your own configuration file for your web application.
  • Error messages are by default saved on file system not in database why?
    Simply because in development time most errors come from database itself, and thus errors won’t come if they’re stored there.
    And because web2py’s philosophy is than no separation between development and production stages, it’s the best choice to have errors stored on file system.
  • No more template languages, only python code.
    This doesn’t mean that you can’t use other template languages.
    You can use easily use your favorite template language.
    Django’s philosophy is to let view to designer and use template languages because in most cases designers are not programmers, but this’s not web2py’s philosophy.
    In fact, using web2py you can return to the view all elements that you want and just let designer control how things look like and when a designer needs to do some customization, he can still use usual HTML tags.

    • In your view you can use {{}} to write your python code into it
    • Multi lines are allowed between single {{}}.
    • no indentation is required, when using python in views
    • if statements, for, and while need to be closed using {{pass}} because in view , there’s no indentation
    • {{if x > 6:}}
    • web2py can have any implementation of any template language, its design allows so but why would one trade python for a template language ? !!!
  • unpickeled objects stored in session causes problems.ex: trying to store in session a database table object causes aplication to crash
  • session.table = db.my_table   # Don't do this

    SQLDB objects are not pickleble because of the open connection.
    In fact you cannot store objects that belong to a class starting with SQL*, i.e. connections, record sets, queries, individual records, etc. You also cannot store classes (only objects), functions and lambdas.
    [classes starting with SQL like SQLFIELD are having short names now like Field, ….. both names are working in fact ]

    user = db(db.users.username==username).select().first()
    You cannot store a user object in the session.The reason is that user has a method user.update_record which requires a db connection and it is not serializable.

    If cPickle.dumps(obj) does not raise an exception than you can do

  • web2py provides an ajax app that provides a python console.
    It uses pickle to store state therefore it is somewhat limited in what it can do and it has security issues but it is a fun toy to play with and show off python.
    Too bad in its present for it cannot be used to interact with SQLDB
    which is not pickleble because of the open connection.
  • It is standard in MVC frameworks including web2py that each page/form is a self submitting and it is uniquely associated to a controller function (Rails calls them actions). The reason is that a form knows how to validate its own input variables. Upon validation the page redirects to a different page(action). In web2py URL builds URLs within web2py.
    So generally it’s a bad behavior to have a form to validate inputs of another form

    For more Information click here

  • web2py prevents double submission of forms using a unique key passed to the form accepts() function via session.
    It makes sure if you click on a form twice, it is processed only once. This prevents user error and some forms of reply attacks. If you do not pass the session variable this mechanism is disabled.

    if form.accepts(request.vars, session)

    This mechanism fail though if you’ve multiple forms in the same page
    and in this case we should pass to the accepts function of every form explicit and different form names to distinguish between forms when they’re being accepted.

    if form1.accepts(request.vars, formname='custom_name'):
    if form2.accepts(request.vars, formname='custom_name2'):
        #do_another thing
  • Web2py had its own Database Abstraction Layer, the thing that bothers some people and push them to argue about the benefits of this.
    and why not using say for example SQLAlchemy the famous ORM.
    SQLAlchemy is better than the web2py ORM in dealing with legacy databases. web2py has restrictions in this case. But the web2py ORM is much better integrated with the rest of the framework than SQLAlchemy is integrated with, for example, Pylons or TurboGears. If you do not use the entire web2py framework you are better off with SQLAlchemy. If you do use web2py, you are better off with its own ORM.

    Any way the new Google appengine Bigtable datastore makes any SQL-based ORMs moot now anyway, so all we really need to worry about are the business logic and presentation, both of which web2py excels at.

  • web2py’s Database Abstraction Layer (DAL) supports many kinds of RDBMS like sqlite, mysql, oracle, …..
    In fact any new application created by web2py has a support of sqlite by default and this’s on purpose because this provides a quick solution for making an application that’s up and running on the fly without many configurations required or something.

    To change this [may be in a real production mode], you’ve to change one line in models/db.py from

    db = DAL('sqlite://storage.sqlite')

    into the appropriate dbms url in your system, for example to support mysql you simply have to do :

  • More interestingly, you can use DAL without database using :

    db = DAL(None)

    then you can define tables, and create SQLFORMs from those tables and test the validation mechanism
    and generate the real sql queries generated by DAL and that are hidden by the DAL syntax

    # Model
     db = DAL(None)
    db.define_table('members', Field('name', requires=IS_NOT_EMPTY()))
    def index():
        form = SQLFORM(db.members)
        if form.accepts(request.vars, session):
            response.flash = T('hey it worked')
        print db._lastsql
       return dict(form=form)

    Congratulations !!! without need to introduce extra/new syntax.
    Internally DAL(None) behaves like a sqlite db and you can use it test query generations as well but there is no sqlite file so nothing is stored and no overhead.

  • Things to take care of when using sqlite though are :

    • SQLite locks the database. Only one thread can safely access it.
    • Only the RENAME TABLE and ADD COLUMN variants of the ALTER TABLE command are supported. Other kinds of ALTER TABLE operations such as DROP COLUMN, ALTER COLUMN, ADD CONSTRAINT, and so forth are omitted.

      If you want to make more complex changes in the structure of a table, you will have to recreate the table. You can save existing data to a temporary table, drop the old table, create the new table, then copy the data back in from the temporary table.

      For example, suppose you have a table named “t1” with columns names “a”, “b”, and “c” and that you want to delete column “c” from this table. The following steps illustrate how this could be done:

          CREATE TEMPORARY TABLE t1_backup(a,b);
          INSERT INTO t1_backup SELECT a,b FROM t1;
          DROP TABLE t1;
          CREATE TABLE t1(a,b);
          INSERT INTO t1 SELECT a,b FROM t1_backup;
          DROP TABLE t1_backup;
    • To learn more about sqlite you can take a look at :
      Features not supported by sqlite
      Sqlite faq

  • Previous limitations belong to Sqlite only, other dbms runs smoothly
    In fact to ensure that web2py is helpful in an agile environment,
    web2py’s Database Abstraction Layer (DAL) supports migration,
    so you can easily make something like:

    db.define_table('my_table', Field('my_field', .....), migrate=True)
    # migrate=True

    setting migrate to True comes in rescue in the development time, you can just change the structure of your table [add more fields, remove ones] and your changes will take effect on the fly, upon the next request [all models are executed at every request and thus those changes will take effect then if migrate = True].
    This fact can lead to an interesting stuff, such as that you can make a const that holds the value of the migrate variable and set it to True while development and return it to False when you’re ready to publish your work.

  • web2py DAL does not perform any hidden Database IO. You have to call count, insert, delete, update or select to do any IO. This is a feature, not a limitation.
    At the current time no need for an ORM on top of it because the current DAL is more usable then existing ORMs.
  • Web2py DAL makes a good work for preventing uncompleted commitments to database to ruin your data
    All calls are wrapped into a transaction, and any uncaught exception
    causes the transaction to roll back. If the request succeeds, the trans-
    action is committed.
  • DAL supports distributed transactions
  • Web2py DAL provides us with some interesting stuff that makes your life easier :
    • db().select(db.table_name.field, orderby='<random>')
      You can get a random order every time the query is initiated.</li>
      db._lastsql -> returns the last sql code generated by DAL
    • most web2py expressions can easily be serialized in SQL. For example:

      Warning Warning Warning :
      Always use when ever doing queries the operators ‘&’, ‘|’ , Don’t use ‘and’, ‘or’ .

    • you can print str(myquery) and see the SQL that corresponds to the
    • methods :
       .delete(), .update(), .insert() and .select() 

      they all have a :

      ._delete(), ._update(), ._indert(), ._select()

      that instead of performing the operation return the SQL that does.

    • You can list all tables in database using : db.tables() which leads to some interesting stuff like:
      db.define_table('rating', Field('table_name'), Field('record_id', 'integer'), Field('rate', 'integer'))
      db.rating.table_name.requires = IS_IN_SET(db.tables())

      and whenever trying to insert some record you can do something like:

      db.rating.insert(table_name = str(db.game), record_id=..., rate=...)
    • DAL supports ‘NOT’ using the ‘~’, so you can do something like:

      db( ~db.table.field.belongs((1,2,3))).select() 
    • Web2py’s DAL supports limitby for pagination.Supporting limitby in oracle was a problem since oracle doesn’t support it.
      Both web2py and Django support this feature differently, While the code generated by Django is more readable, it is slower
      because it asks Oracle to first select all rows, then select the
      limited subset. The web2py way does it with three nested select that,
      as discussed in This article , allows Oracle to perform additional optimizations (and it works with joins too).
  • web2py’s Database Abstraction Layer (DAL) supports unicode by default
    you can check for that in gluon/sql.py

            charset = m.group('charset') or 'utf8'
             self._pool_connection(lambda : MySQLdb.Connection(
            # table should use this utf-8 when creating it
            if self._db._dbname == 'mysql':
                fields.append('PRIMARY KEY(%s)' % self.fields[0])
                other = ' ENGINE=InnoDB CHARACTER SET utf8;'
  • To deal with oracle database, web2py needs cx_Oracle
  • single quotes are escaped in Oracle SQL using repeated single quotes and web2py takes care of this
  • DAL makes a great job in simplifying selecting from database, so you can do :
    db((db.my_table.id>0) & (db.my_table.second_field == 'value') ).select()
    or even easier :
    db(db.my_table.id>0)(db.my_table.second_field == 'value').select()
    db(a)(b)(c) is the same as db((a)&(b)&(c))
    in fact the former is interpreted as taking the (c) subset of the (b)  
    subset of the (a) subset of db. 
  • dropping a database table manually causes web2py to fail and not working, with an error message that table doesn’t exist? while this doesn’t happen in case of dropping the table through web2py and the reason for this is :
    • when tables are created the first time web2py makes a .table file under applications/yourapp/databases
      that is how keeps track of table structure.
    • If you drop the tables manually, you need to delete those files as well or they will not be re-created.
  • You can make 2 web2py’s applications share the same database, you can copy the model files within database tables are defined to another application and start to use it.
    One proble though is that there should be only one application that is allowed to create tables and the other should have ‘migrate=False’ s the last argument in db.defin_table()

    Web2py keeps track of the migration files in a folder called database, web2py generates a random name for it and you should [if you tend set migration to True to give it a name]

    migrate='table_name.table'   => '%s.table'%table_name

    if the database is re-factored with a name change, we wouldn’t want to lose the data from the column, just have the column name changed
    So what to do?

    Find the .table file for that table and set migrate= to that filename.
    One should actually use migrate=”mytable.table” to give a name to the table files from the beginning create instead of sing the cryptic
    names created vy web2py.

  • You can’t reference other tables using other fields than their id, which is a good practice and in the same time efficient (strings need to be hashed while integers don’t. )

  • Dal has some limitations though:
    • Doesn’t support inheritance, and although you can do something like:
      db.define_Table('my_table', ....)
      db.new_table = db.my_table

      There’s no sync between them, changes in one of them are not reflected in the other.

    • Doesn’t support multiple insertion in the same query.
    • No support for special field types(Enum, Array e.t.c) and XML fields in POSTGRESQL.
    • You can’t rename a column, if you renamed it in your model, this will drop the old and create a new empty one [in case migration=True].
      To rename a column, just :
      1)create the new column.
      2)drop the old one.
      3)copy data
    • No support for self referencing foreign key [No parent/child relationship support], instead you can make an integer field and us it to store a self referencing key, with the parent field has the value of this field set to NUll or -1.
    • many-to-many relationship in web2py has to be explicit (as in rails, not as in Django). This means you have to create the intermediate link table.
      The intermediate table in a many-2-many relation, in the general case, can hold more than two keys.
      You can use sqldesigner, but you need to create the intermediate table to link the many- to the -many.
    • DAL doesn’t support something like :
      table A depends on table B
      table B depends on table C
      table C depends on table A
  • In web2py, menus can be created automatically :
    • check menu.py in your models in your web application folder
    • you can also check for the current used action in a menu [To make an active item]
      ['item name',(request.function=='action'), URL(r=request,f='action',vars={}, args=[])],

      as you can see you check with request.function==’action’ where ‘action’ is the name of the function called [change it to the name of the controller function that is referred to by this menu item]

  • Making a private function is easy in web2py :
    • Functions defined in a model are private.
    • Functions that are defined in controllers and that takes arguments are private.
    • Functions defined in controllers and start with ‘__’ [double underscores] are private.
      But functions starting with single ‘_’ are not private and they should not be because some people use them for an ajax callback.
    • Functions defined in controllers and having a space after the () and before the ‘:’ are private.

      def pri_action() :
      # do some thing
  • web2py provide us with many many HTML helpers that makes your life easier and instead of writing an HTML code all the time, you can use them as [python code] either in the view or in your controller functions (actions) .
    They do a grat job and makes you use python instead of HTML code in your actions.

    A('Download', _href=URL(....)) instead of <a href=''>'Download'</a>

    You can also have a server side highlighting using CODE helper that will display code in HTML with highlighting.

     {{=CODE(\"print 'hello world'\", language='python', link=None,
                counter=1, styles={})}}

    another example:

    BR -> <br/>
    P(..., cr2br=True)  Will replace ``\\n`` by ``<br />`` if the `cr2br` attribute is provided.
    P() -> <p>
  • You need to know that the philosophy of web2py HELPERS is:
    • Not to deliver all possible HTML helpers since it’s very easy to do a one , just look at gluon.html.py and you’ll see how easy it is
      you just make a class that extends the DIV class and into it you type
      tag = ‘whatever tag you want to implement’ .

      class FIELDSET(DIV):
          tag = 'fieldset'

      The fieldset tag should just wrap whatever is in it, since most designers never use tables to format forms.
      The fieldset tag helps create visible rubrics with very large forms

  • In web2py, you can return result to a user while making some computation that takes longer time.
    • For this don’t use forking because it doesn’t work in windows.
    • If the computation can be interrupted you can have an ajax call from
      the client initiate the computation in a separate controller function.
    • If the computation can not be interrupted I would run it a separate
      thread (not process). Mind that wsgiserver will kill threads that run
      too long.