Advanced usage

This final part of the databases documentation will explain some advanced features and behaviors you might need to understand when you're designing more complex applications or you need advanced and particular flows for your data.

Model inheritance and subclassing

The Model class can be subclassed and extended by design in many different ways. You can also define models without creating a table for them in the database and use them as meta classes for common fields, validations and other needs.

In fact, even if some properties regarding your model are bound dictionaries of the class itself, Emmett will hack these attributes in order to make them inherit and extend the ones from the upper models.

For instance, if you have a lot of models where you want to store the creation and update timestamps, you can write down a simple model for that:

class TimeStampModel(Model):
    created_at = Field.datetime()
    updated_at = Field.datetime()

    default_values = {
        'created_at': lambda: request.now,
        'updated_at': lambda: request.now
    }

    update_values = {
        'updated_at': lambda: request.now
    }

and then extend it with the other models:

class Post(TimeStampModel):
    title = Field()
    body = Field.text()
    status = Field()

    default_values = {
        'status': 'published'
    }

class Comment(TimeStampModel):
    text = Field.text()

Then if you pass just these models to your Database instance:

db.define_models(Post, Comment)

you will find the created_at and updated_at fields on these tables and the default and update values set for them.

As you can see, the Post model has its own default_values dictionary that defines the default value for the status field, but the resulting dictionary will contains also the values of the TimeStampModel dictionary. This behavior is injected by Emmett and is intended to avoid you the pain of rewriting the whole dictionaries when extending models.

You can obviosly override a value from the super model, in fact if you write something like this in the Post model:

default_values = {
    'updated_at': lambda: request.now+timedelta(minutes=5),
    'status': 'published'
}

it will change the default value defined from TimeStampModel for the updated_at value.

This is true for all the reserved properties related to the fields we seen in the models part of the documentation, and all the other attributes of the Model class will follow the normal inheritance behavior in the python language.

Models in Emmett also inherits all the properties decorated with the helper functions we seen in the previus chapters, and all the relations too.

In fact, if you defined scopes in TimeStampModel:

@scope('created_between')
def filter_creation_period(self, start, end):
    return (self.created_at >= start) & (self.created_at < end)

@scope('updated_between')
def filter_update_period(self, start, end):
    return (self.updated_at >= start) & (self.updated_at < end)

you will be able to use them on the Post and Comment classes.

You can also use multiple models as base classes for inheritance, and Emmett will merge the relevant properties with the order of the base classes. For example if we define another base model:

class Hidden(Model):
    is_hidden = Field.bool()

    default_values = {
        'is_hidden': True
    }

    @scope('not_hidden')
    def filter_not_hidden(self):
        return self.is_hidden == False

    @scope('only_hidden')
    def filter_only_hidden(self):
        return self.is_hidden == True

and update the inheritance for the Post class:

class Post(TimeStampModel, Hidden):
    # ...

you will have the fields, default values and scopes from this model too.

Warning: every time you using inheritance with model ensure every class is a subclass of the Model one.

Note: when you need to override a decorated method of the super model, ensure to decorate it also in the subclassed one and to use the same method name of the super one.

Customize has_one and has_many sets

Sometimes you will need to customize the way Emmett generates relations sets for has_one and has_many relations.

For example, you may want to change the behavior of a relation depending on some conditions:

from emmett import session

class User(Model):
    @has_many()
    def posts(self):
        if session.user.is_admin:
            return Post.where(lambda m: m.is_trashed == True)
        return Post.where(lambda m: m.is_trashed == False)

As you can see we used the has_many helper as a decorator over custom posts method we defined inside our User model. This way, the posts attribute of a user will have different sets depending on the condition we defined.

Another common scenario would be a polimorphic relation, where you want to create multiple association to a single model. Let's say, for example, that you want to use a single table to store your photos, but you want to refer this table to several ones, for example to a table of cars and also one of trucks:

class Photo(Model):
    url = Field()
    entity_type = Field()
    entity_id = Field.int()

class Car(Model):
    name = Field()
    price = Field.float()

    @has_many(field='entity_id')
    def photos(self):
        return Photo.where(lambda m: m.entity_type == 'Car')

class Truck(Model):
    name = Field()
    hp = Field.int()
    price = Field.float()

    @has_many(field='entity_id')
    def photos(self):
        return Photo.where(lambda m: m.entity_type == 'Truck')

As you can see, in this case we also specified the field Emmett should use as the foreign key for the relation – if not specified this field is the name of the downcase name of the model.

You can use polimorphic relations also for many-to-many relations, for example for a tagging system like this:

class Tag(Model):
    name = Field()

class Tagging(Model):
    belongs_to('tag')
    tagged_type = Field()
    tagged_id = Field()

class Post(Model):
    @has_many(field='tagged_id')
    def taggings(self):
        return Tagging.where(lambda m: m.tagged_type == 'Post')

    has_many({'tags': {'via': 'taggings'}})

class Video(Model):
    @has_many(field='tagged_id')
    def taggings(self):
        return Tagging.where(lambda m: m.tagged_type == 'Video')

    has_many({'tags': {'via': 'taggings'}})

Note that inheritance we explained in the section above is great to avoid repeating code:

class Taggable(Model):
    @has_many(field='tagged_id')
    def taggings(self):
        return Tagging.where(
            lambda m: m.tagged_type == self.tablename)

    has_many({'tags': {'via': 'taggings'}})

    @fieldmethod('taglist')
    def get_taglist(self, row):
        return row.tags().column('name')

class Post(Taggable):
    title = Field()

class Video(Taggable):
    title = Field()

Advanced indexes

New in version 0.7

Using expressions

Note: support for expressions in indexes depends on the DBMS in use.

Emmett supports expressions on indexes definition, which can be useful when you want to coalesce values from your rows:

class Post(Model):
    author = Field()
    is_private = Field.bool()
    title = Field()

    indexes = {
        'author_and_priv': {
            'fields': ['author'], 
            'expressions': lambda m: m.is_private.coalesce(False)
        }
    }

As you can see the expressions option accepts one or a list of lambda functions, that should accept just one parameter: the model. The returning value should by any Emmett valid sql expression on a field of the model itself.

Conditional indexes

Note: support for conditional indexes depends on the DBMS in use.

Emmett supports a where option on indexes definition, which can be used to define conditional indexes:

class Post(Model):
    author = Field()
    is_private = Field.bool()

    indexes = {
        'author_and_priv': {
            'fields': ['author'], 
            'where': lambda m: (m.is_private == False)
        }
    }

The where option accepts a lambda function that should accept the model as first parameter and should return any valid query using the Emmett query language.