Python __repr__ and None

Question

I'm quite new to Python and currently I need to have a __repr__ for a SqlAlchemy class. I have an integer column that can accept Null value and SqlAlchemy converts it to None. For example:

class Stats(Base):
   __tablename__ = "stats"
   description = Column(String(2000))
   mystat = Column(Integer, nullable=True)

What is the correct way to represent the "mystat" field in the __repr__ function when SqlAlchemy returns None?

Answer 1

The __repr__ should return a string that describes the object. If possible, it should be a valid Python expression that evaluates to an equal object. This is true for built-in types like int or str:

>>> x = 'foo'
>>> eval(repr(x)) == x
True

If that's not possible, it should be a '<...>' string uniquely describing the object. The default __repr__ is an example of this:

>>> class Foo(object):
        pass
>>>
>>> repr(Foo())
'<__main__.Foo object at 0x02A74E50>'

It uses the object's address in memory to uniquely identify it. Of course address doesn't tell us much about the object so it's useful to override __repr__ and return a string describing the object's state.

The object's state is defined by other objects it contains so it makes sense to include their repr in yours. This is exactly what list or dict do:

>>> repr(['bar', Foo()])
"['bar', <__main__.Foo object at 0x02A74710>]"

In your case, the state is in your Column properties so you want to use their repr. You can use the %r formatting for this, it inserts a repr() of the argument:

def __repr__(self):
    return '<Stats: description=%r, mystat=%r>' % (self.description, self.mystat)

An equivalent using the new formatting:

def __repr__(self):
    return '<Stats: description={0.description!r}, mystat={0.mystat!r}>'.format(self)

Answer 2

maybe repr(mystat) is what you want?

Answer 3

'Null' if mystat is None else mystat

Answer 4

Isn't the generic_repr decorator of sqlalchemy-utils providing a community based solution that suit your needs?

It leaves it as None.

Answer 5

The previous answer showing how to override Base.__repr__ was exactly what I needed. Thank you! Here it is re-written with f-strings for Python 3.7+ and overriding the flask-sqlalchemy db.Model.

def override_default_repr(self):
    class_name = self.__class__.__name__
    attribs = ", ".join(
        [
            f"{k}={self.__dict__[k]!r}"
            for k in sorted(self.__dict__.keys())
            if k[0] != "_"
        ]
    )
    return f"<{class_name}({attribs})>"

db.Model.__repr__ = override_default_repr

Answer 6

I was trying to find a generic __repr__ method that could be used on any SQLAlchemy object and only found this page. So, I decided to write my own and here's what I've done:

from sqlalchemy.ext.declarative import declarative_base
Base = declarative_base()

if __debug__:
    # monkey-patch in useful repr() for all objects, but only in dev
    def tablerepr(self):
        return "<{}({})>".format(
            self.__class__.__name__,
            ', '.join(
                ["{}={}".format(k, repr(self.__dict__[k]))
                    for k in sorted(self.__dict__.keys())
                    if k[0] != '_']
            )
        )
    Base.__repr__ = tablerepr

This extends the Base class to print out the contents of the particular instance. So, now every object I create that extends the Base will have a __repr__ method that prints more than just the class name and the instance hash.

EDIT: I added an if __debug__ as this change may cause a leak of information you may not want leaked in a production environment. I also added a sorted so the display will be consistent.