A __getattr__ for class methods?

D

Dylan Moreland

I'm trying to implement a bunch of class methods in an ORM object in
order to provide functionality similar to Rails' ActiveRecord. This
means that if I have an SQL table mapped to the class "Person" with
columns name, city, and email, I can have class methods such as:

Person.find_by_name
Person.find_by_city_and_name
Person.find_by_name_and_city_and_email

I have a metaclass generating basic properties such as .name and .city,
but I don't want to generate a class method for every permutation of
the attributes. I'd like to have something much like __getattr__ for
instance attributes, so that if a method like
Person.find_by_city_and_email cannot be found, I can construct a call
to the basic find method that hides the SQL. Is there any way of doing
this, or am I trying to mirror a functionality that Python simply does
not have?
 
M

Michael Spencer

Dylan said:
I'm trying to implement a bunch of class methods in an ORM object in
order to provide functionality similar to Rails' ActiveRecord. This
means that if I have an SQL table mapped to the class "Person" with
columns name, city, and email, I can have class methods such as:

Person.find_by_name
Person.find_by_city_and_name
Person.find_by_name_and_city_and_email

I have a metaclass generating basic properties such as .name and .city,
but I don't want to generate a class method for every permutation of
the attributes. I'd like to have something much like __getattr__ for
instance attributes, so that if a method like
Person.find_by_city_and_email cannot be found, I can construct a call
to the basic find method that hides the SQL. Is there any way of doing
this, ...

Sure, define __getattr__ on the type of the class i.e., the metaclass, just as
you define it on a class to provide default-attribute-lookup to its instances:
... class __metaclass__(type):
... def __getattr__(cls, attr):
... return "%s.%s" % (cls.__name__, attr)
...
HTH

Michael
 
R

Roland Heiber

Dylan said:
I have a metaclass generating basic properties such as .name and .city,
but I don't want to generate a class method for every permutation of
the attributes. I'd like to have something much like __getattr__ for
instance attributes, so that if a method like
Person.find_by_city_and_email cannot be found, I can construct a call
to the basic find method that hides the SQL. Is there any way of doing
this, or am I trying to mirror a functionality that Python simply does
not have?

I don't get it? __getattr__ is exactly what you are looking for:

http://docs.python.org/ref/attribute-access.html#l2h-206

__getattr__( self, name)

Called when an attribute lookup has not found the attribute in the usual
places (i.e. it is not an instance attribute nor is it found in the
class tree for self). name is the attribute name. This method should
return the (computed) attribute value or raise an AttributeError exception.

HtH, Roland
 
D

Dylan Moreland

Michael said:
Sure, define __getattr__ on the type of the class i.e., the metaclass, just as
you define it on a class to provide default-attribute-lookup to its instances:

... class __metaclass__(type):
... def __getattr__(cls, attr):
... return "%s.%s" % (cls.__name__, attr)
...

HTH

Michael

Thanks! I only recently realized that I would have to learn metaclasses
in order to make this work, and I'm still a bit unclear on their
properties.
 
X

Xavier Morel

Dylan said:
I'm trying to implement a bunch of class methods in an ORM object in
order to provide functionality similar to Rails' ActiveRecord. This
means that if I have an SQL table mapped to the class "Person" with
columns name, city, and email, I can have class methods such as:

Person.find_by_name
Person.find_by_city_and_name
Person.find_by_name_and_city_and_email

I have a metaclass generating basic properties such as .name and .city,
but I don't want to generate a class method for every permutation of
the attributes. I'd like to have something much like __getattr__ for
instance attributes, so that if a method like
Person.find_by_city_and_email cannot be found, I can construct a call
to the basic find method that hides the SQL. Is there any way of doing
this, or am I trying to mirror a functionality that Python simply does
not have?
I'm not sure that the way you tackled this is the good approach: while
it's quite flexible as far as the search criteria go, it'll require less
than obvious code to match keywords (field names) and values, will lead
to somewhat verbose syntax (especially with many criteria), and the
syntax itself is unforgiving (every new search field requires at least 5
additional characters on top of the field name itself), brittle (you'll
have to do an extensive validation of your method name and fields unless
you want everything to break, and Python doesn't support your syntax)
and not really pythonic.

Since you seem to know ActiveRecord, you're probably familiar with the
way AR does this task: with a hash of column_name, value pairs. The
syntax is quite straightforward especially due to the fact that any "key
=> value" comma-separated pairs sequence generates a hash, without
requiring explicit hash boundaries ( { and } ). The syntax is extremely
straightforward since it relies on the language's syntax itself, and the
correctness of the grammar is checked by the compiler/interpreter itself
since no custom syntax is built. This construct is therefore quite
solid, on top of being obvious to a Rubyist.

Now, you're in luck, because Python has even better than that: **kwargs,
the optional keyword arguments.

As you probably know, Python has a good support for keyword args,
allowing you to fill your arguments out of order (and leave the 3th
argument at it's default value while specifying the value of the 5th
argument). But **kwargs goes beyond the regular explicit keyword
arguments: when specified, **kwargs is a dict populated with the
implicit keyword arguments (keyword:value pairs), the ones you haven't
specified in the argument tuple of your method.

This means that if I define a function as

def foo(bar=0, **kwargs):
print kwargs

Then calling foo() will print an empty dict
As will calling foo(3) or foo(bar=3), because the explicit "bar" keyword
argument is used.

Now if I call foo(something=5, somethingelse="woohoo"), kwargs will
evaluate to

{"somethingelse":"woohoo","something":5}

This means that all you have to do is replace your method definition with

and in the method itself iterate over the key:value pairs of kwargs to
automagically get both the field names and the values upon which your
search shall be performed.

Calling the method would then look something like:
the syntax is fairly obvious and pythonic, has a low verbosity, and
Python itself will do the parsing and the grammatical validation of your
method calls for you.
 
X

Xavier Morel

Oh, and I wondered too: is your goal to build an ORM, or do you just
need an ORM?

Cause if it's the latter then Python does already have some fairly good
ORMs such as SQLAlchemy or PyDO2, you don't *need* to create yours.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
474,283
Messages
2,571,409
Members
48,103
Latest member
MadieDeitz

Latest Threads

Top