Computed and virtual attributes

Quite often during the development of your application you will need to compute values or perform operations that dependend on the values contained in the rows you're selecting, inserting or updating in your database.

Emmett provides different apis that can help you in these cases: let's see them in details.

Computed fields

Changed in version 2.4

Sometimes you need some field values to be computed using other fields' values. Let's say, for example, that you have a table of items where you store the quantity and price for each of them. You often need the total value of the items you have in your store, and you don't want to compute this value every time in your application code.

A solution can be to compute the value when you change the price or the quantity of the item and store that value to the database too. In this case you can use the compute decorator:

from emmett.orm import Model, Field, compute

class Item(Model):
    price = Field.float()
    quantity =
    total = Field.float()

    @compute('total', watch=['price', 'quantity'])
    def compute_total(self, fields):
        return fields.price * fields.quantity

As you can see, the compute decorator accepts the name of the field where to store the result of the computation and and optional watch list of fields.

The function that performs the computation has to accept the operation fields as its first parameter, and it will be called both on inserts and updates.

Note: compute decorated methods receives only the fields' values involved in the operation. This means that fields will contain only the values passed by the insert/update operation and the relative default values.

Operation fields and watch parameter

Since computations will be triggered on every update operation that might happen on your model, and thus such operation might involve several records, you might end up in conditions where the operation doesn't include all the fields required for the computation. For example issuing this update instruction on the upper model:

Item.where(lambda i: i.quantity == 1).update(quantity=2)

would make impossible to re-compute the total value for the involved records.

The watch parameter is designed to avoid these conditions, since – under default behaviour – in an insert or update operation involving computations Emmett will:

  • execute all the computations without watch fields, ignoring the ones failing
  • execute all the computations with where the watch fields presence is completely satisfied
  • raise an exception, preventing the operation to continue, for those computations where the watch fields presence is not completely satisfied

Considering our upper example:

  • the total computation will be executed for all the operations including both price and quantity fields
  • an operation with only one of the price or quantity fields cannot be executed
  • operations not involving price or quantity fields won't trigger the computation

Note: to handle complex cases where you need to access the single record fields we suggest to use records' save method and relevant callbacks.

Virtual attributes

Changed in version 1.0

Virtual attributes are values returned by functions that will be injected to the involved rows every time you select them.

To clarify this concept, we will consider the same example we gave for the computed attributes and we will replace them with the rowattr decorator instead:

from emmett.orm import Model, Field, rowattr

class Item(Model):
    price = Field.float()
    quantity =

    def total(self, row):
        return row.price*row.quantity

As you can see, we don't have a real column in the table that will store the total value of the item, but we defined instead a method that evaluate it and add it to the selected rows.

Since virtual attributes are, by definition, virtuals, you can't use them in order to make queries.

You can access the values as the common fields:

>>> item = db(Item.price >= 2).select().first()

Virtual methods

Changed in version 1.0

Similarly to virtual attributes, these methods are helpers injected to the rows when you select them. Differently from virtual attributes, however, they will be methods indeed, and you should invoke them to access the value you're looking for.

Let's consider again the same example we saw above, and let's use the rowmethod decorator:

from emmett.orm import Model, Field, rowmethod

class Item(Model):
    price = Field.float()
    quantity =

    def total(self, row):
        return row.price*row.quantity

As we said, virtual methods are evaluated on demand, which means you have to invoke them when you want to access the values you need:

>>> item = db(db.Item.price > 2).select().first()

More on virtual methods

Field methods are a great instrument also to run database operations from the current selected object. In fact, since you're inside the model instance, you can access the model itself, the table and the database from within the method you're writing.

For example, let's say you have a table of messages referring to some topics and you want to easily get the next message from the current one. You can write down a method for that:

from emmett.orm import Model, belongs_to, rowmethod

class Message(Model):
    belongs_to('topic', 'author')
    body = Field.text()
    written_at = Field.datetime()

    def get_next_message(self, row):
        return self.db(
            (self.topic == row.topic) &
            (self.written_at > row.written_at)
            limitby=(0, 1)

Then, once we have a message, we can access the next quickly:

>>> message = db(db.Message.topic == 1).select().first()
>>> message
<Row {'id': 2L, 'topic': 1L, 'author': 1L, 'written_at': datetime.datetime(2015, 12, 22, 9, 18, 23, 118701), 'body': 'This is a test message'} >
>>> message.next_one()
<Row {'id': 3L, 'topic': 1L, 'author': 1L, 'written_at': datetime.datetime(2015, 12, 22, 9, 20, 21, 229511), 'body': 'This is another test message'} >

Virtual methods, as we saw for virtual fields, needs the row as first parameter, that will be injected by Emmett, but you can obviously add more parameters and pass values for them during invocation.