Database objects and operations

Once you defined your models and the structure of your database entities, you need to make operations with them. In the next paragraphs we will inspect all the ways to create, modify and fetch your data from the database.

Creating records

The first operation you may need, is to add new data to your database.

Given a very essential model:

class Dog(Model):
    name = Field()


the simplest way to create a new record is to use its create method:

>>> Dog.create(name="Pongo")
<Row {'errors': {}, 'id': 1}>

As you can see, the create method return a Row object, which contains the id of the created record and a dictionary named errors. This is because the create method will validate the input before trying to insert the new record.
In fact, if we add a validation rule to the Dog model:

class Dog(Model):
    name = Field()
    validation = {'name': {'presence': True}}

and we try to insert a dog without specifying a name:

>>> Dog.create()
<Row {'errors': {'name': 'Cannot be empty'}, 'id': None}>

we have the name field in the errors and the id set as None, meaning that no record has been created at all.

Emmett has also a more low level method to create records, that will skip the validation and insert the record directly into the database:

>>> db.Dog.insert(name="Peggy")

As you can see, the insert method of the table defined by the model will return directly the id of the inserted record, since no validation was performed.

Remember that if you're not in the request flow with the Database pipe, you have to commit your changes to effectively have them written into the database.

Accessing the created record

As we just seen from the above methods, when you create a new record, Emmett returns just the integer corresponding to the id of the database row. If you look deeply, you will find that actually the return value is not just an integer:

>>> rv = Dog.create("Penny")
>>> type(
<class 'pydal.helpers.classes.Reference'>

In fact, you can access the attributes of the record you just created:

{'id': 3, 'name': 'Penny'}

We will see more about the as_dict method in the next paragraphs.

Spatial fields helpers

Emmett provides some helpers on GIS fields (geography and geometry types) in order to simplify the workflow regarding these values.

Whenever you have a GIS column:

class City(Model):
    name = Field.string()
    location = Field.geography("POINT")

you can use the provided helpers from emmett.orm.geo module to produce fields' values:

from emmett.orm import geo

rv = City.create(
    name="Hill Valley",
    location=geo.Point(44, 12)

Also, GIS fields values are sub-class of str, but they provides some additional attributes:

'POINT(44 12)'
(44.0, 12.0)

On geometries representing collections, you also have the groups attribute:

>>> mp = geo.MultiPoint((1, 1), (2, 2))
>>> mp.groups
('POINT(1.000000 1.000000)', 'POINT(2.000000 2.000000)')

Making queries

As soon as you have rows in your tables, you need to query them to fetch the data you want. Emmett provides a very efficient way to write queries using python language, since you will use your model fields and their methods.

But, before we proceed learning the syntax to make queries, we have to understand the main principle behind Emmett querying: the sets. Every time you work with the database to filter data, you're actually using a Set of rows corresponding to your query. The Set class is fundamental in Emmett and allows you to make all the operations on the records corresponding to your query, as we will see in the next paragraphs .

So, how you make queries on your database? Let's say, for example, that you have a table containing events, defined by the model:

class Event(Model):
    name = Field()
    location = Field()
    participants =
    happens_at = Field.datetime()

and you want to query all the events for a certain location. You can use your Database instance and its where method for that:

>>> db.where(Event.location == "New York")
<Set (events.location = 'New York')>

or the more compact

>>> db(Event.location == "New York")
<Set (events.location = 'New York')>

that produce the same result.
As you can see, you can build queries using your model fields and the available operators:

operator description
== value is equal to
!= value differs from
< value is lower than
> value is greater than
<= value is lower than or equal to
>= value is greater than or equal to

Returning back to our Event model, we can, for example, get all the events that are not in New York:

db(Event.location != "New York")

or all the events with 200 or more participants:

db(Event.participants >= 200)

– Ok dude, what if I want to combine multiple where conditions?
just use the operators for the and, or and not conditions

Emmett provides the &, | and ~ operators for the and, or and not conditions, in order to combine multiple conditions on the same query.

For example, you may want all the events in New York that have less than 200 participants:

>>> db((Event.location == "New York") & (Event.participants < 200))
<Set ((events.location = 'New York') AND (events.participants < 200))>

or the events happening on a specific day:

    (Event.happens_at >= datetime(1955, 10, 5)) & 
    (Event.happens_at < datetime(1955, 10, 6))

or the future events that won't be in New York or in Chicago:

        (Event.location == "New York") |
        (Event.location == "Chicago")
    ) & (Event.happens_at >=

Model where method

In all the examples we've seen above, we applied multiple where conditions on the same table. Emmett offers also a more compact way to write these queries using directly the Model.where method and a lambda notation:

Event.where(lambda e: 
        (e.location == "New York") | (e.location == "Chicago")
    ) & (e.happens_at >=

The resulting Set will obviously be the same.

Query using tables

As we seen in the models section, adding a model to your Database instance will add a Table object accessible both with the model name and the table name.

Since the tables share the fields with models, you can use them for querying too. In fact you can write the same query in all these ways:

Event.where(lambda e: e.location == "New York")
db(Event.location == "New York")
db(db.Event.location == "New York")
db( == "New York")

and all of them will produce the same result. Just use the one you prefer or that results more convenient for your code.

Query all records in the table

When you want to work with all the records of a table, you have two options, one using the Model class and one with the db() syntax we have seen above:

# from the model
# using Database instance

Both the methods will return the Set corresponding to all the records of the table.

Additional query operators

Emmett also provides additional query operators that might be useful when you need particular conditions or for specific field types. Let's see them in detail.

belongs for the IN condition

When you need to perform sql IN conditions, you can use the belongs method:

locations = ["New York", "Chicago"]

In this example we're asking all the events not happening in New York or Chicago.

String and text operators

An operator you may be familiar with is the like one, that produces a LIKE operation on the database. It works pretty similar to writing a raw sql query with a LIKE condition:


where the % character is a wild-card meaning any sequence of characters, so the query will find any event starting with "party".

But Emmett provides also some shortcuts for the like operator with wild-card:


that will be the same of writing


Note that the like operator will usually be case-sensitive on most of the DBMS, so if you want to make case-insensitive queries, you should specify the option on like and the other helpers:

db("party%", case_sensitive=False))

You can also use the upper and lower helpers:


Emmett provides also a regexp method on fields that works in the same way of the like one but allows regular expressions syntax for the look-up expression. Just remember that only some DBMS support it (PostgreSQL, MySQL, Oracle and SQLite).

Date and time operators

Emmett provides some additional operators for date, time and datetime fields, in particular:

  • date and datetime fields have the day, month and year methods
  • time and datetime fields have the hour, minutes and seconds methods

So, for example, you can query the events of a specific year quite easily:

db(Event.happens_at.year() == 1985)

Engine specific operators

GIS operators

Emmett provides additional query operators specific to spatial extensions. Engines providing this kind of APIs can be Spatialite or PostGIS. The following table describes Emmett's ORM methods:

operator description
st_asgeojson returns a geometry as a GeoJSON element
st_astext returns WKT representation of the geometry/geography
st_x returns the X coordinate of a Point
st_y returns the Y coordinate of a Point
st_distance returns the distance between two geometry/geography values
st_simplify returns a simplified version of a geometry (Douglas-Peucker)
st_simplifypreservetopology returns a simplified and valid version of a geometry (Douglas-Peucker)
st_contains returns true if no points of B lie in the exterior of A
st_equals returns true if two geometries include the same set of points
st_intersects returns true if two geometries intersect
st_overlaps returns true if two geometries intersect and have the same dimension
st_touches returns true true if two geometries have at least one point in common, but their interiors do not intersect
st_within returns true if no points of A lie in the exterior of B
st_dwithin returns true if two geometries are within a given distance

PostgreSQL json operators

Emmett provides additional query operators specific to PostgreSQL engine. The following table describes the mapping between Emmett's ORM methods and the relevant PostgreSQL json/jsonb operators:

operator expression
jcontains jsonb @> jsonb
jhas jsonb ?(|&) text
jin jsonb <@ jsonb
jget json(b) -> integer|text
jgetv json(b) ->> integer|text
jpath json(b) #> text
jpathv json(b) #>> text

While jcontains, jhas and jin are boolean operators, the other ones expect a right operand:

db(Model.jsonb_field.jget("key") == "val")

Moreover, the jhas operator implements an all parameter for list arguments, so you can produce the different queries:

Model.jsonb_field.jhas("key")  # "models"."jsonb_field" ? 'key'
Model.jsonb_field.jhas(["key1", "key2"], all=False)  # "models"."jsonb_field" ?| array['key1','key2']
Model.jsonb_field.jhas(["key1", "key2"], all=True)  # "models"."jsonb_field" ?& array['key1','key2']

Selecting records

Once you have made a query to your database and have a Set, you can fetch the records with the select method:

>>> db(Event.location == "New York").select()
<Rows (2)>

The returning object of a select operation will always be a Rows object, which is an iterable of Row objects. A Row objects behaves quite like a dictionary, but allows you to access its elements as attributes, and implements some useful methods.

>>> rows = db(Event.location == "New York").select()
>>> for row in rows:
...     print(
Awesome party
Secret party
>>> rows[0]
<Row {'happens_at': datetime.datetime(2016, 1, 7, 23, 0, 0), 'name': 'Awesome party', 'participants': 300, 'location': 'New York', 'id': 1}>

The Rows and Row objects have also some helper methods you might find useful. For example, the Rows object has a first and a last methods:

>>> rows = db(Event.location == "New York").select()
>>> rows.first()
<Row {'happens_at': datetime.datetime(2016, 1, 7, 23, 0, 0), 'name': 'Awesome party', 'participants': 300, 'location': 'New York', 'id': 1}>
>>> rows.last()
<Row {'happens_at': datetime.datetime(2016, 1, 8, 23, 0, 0), 'name': 'Secret party', 'participants': 200, 'location': 'New York', 'id': 2}>

They work pretty the same like calling rows[0] and rows[-1] but while using integer position will raise an exception if the Rows object is empty, first() and last() will return None.

The first method can be useful also when you're looking for a single record:

event = db( == "Secret Party").select().first()
if event:
        "Event %s starts at %s" % (
  , str(event.happens_at)
    print("Event not found")

The Row object has an as_dict method that you might find useful for serialization, since it will produce a dictionary from the original object without any callable object. For example, if you're working with json apis, you can render the dictionary directly as the json response.

>>> rows = db(Event.location == "New York").select()
>>> rows.first().as_dict()
{'happens_at': datetime.datetime(2016, 1, 7, 23, 0, 0), 'name': 'Awesome party', 'participants': 300, 'location': 'New York', 'id': 1}

Similarly, the Rows object has both an as_dict and an as_list methods. While the as_list returns a list of rows serialized with as_dict, so you can avoid to call the as_dict of the rows recursively, the as_dict returns a dictionary that will have the ids of the rows as keys and the rows serialized with the as_dict method as values:

>>> rows.as_list()
[{'happens_at': datetime.datetime(2016, 1, ...}, {...}]
>>> rows.as_dict()
{1: {'happens_at': datetime.datetime(2016, 1, ...}, 2: {...}}

Now, let's proceed with the options of the select method. It accepts unnamed arguments: these are interpreted as the names of the fields that you want to fetch. For example, you can be explicit on fetching just the id and name and fields:

>>> rows = db(Event.location == "New York").select(,
>>> rows[0]
<Row {'id': 1, 'name': 'Awesome party'}>

If you don't specify arguments, Emmett will select all the fields for all the tables involved in the query. In fact, the explicit argument for the first example is:

db(Event.location == "New York").select(db.Event.ALL)

The ALL attribute of Table is, indeed, a special attribute that will select all the columns of the table.

Warning: the ALL attribute is available on Table objects only, not on Model obejcts


Changed in version 0.6

Emmett provides some shortcuts that might be useful when you want to select single records. For example, you can select a single record using the Model.get method with the query:

event = Event.get(name="Secret party")

or calling the table:

event = db.Event(name="Secret party")

both the methods will produce the same result of writing:

event = db( == "Secret party").select().first()

And if you want to select a record using the id, you can pass it as an unnamed parameter in both methods, or accessing it as a table item:

event = Event.get(1)
event = db.Event(1)
event = db.Event[1]

The Model class has also a first and a last methods, that will select the first and the last record of the table, with ascending ordering of the id field:

first_inserted = Event.first()
last_inserted = Event.last()


When you want to specify a ordering for selecting record, you can use the orderby option of the select method, that will produce an ORDER BY instruction in the sql query.

db(Event.location == "New York").select(

will return all the events in New York in ascending order by their dates (so the oldest one will be the first).

To have the rows in descending order (in this case the oldest one will be the last), just use the ~ operator:

db(Event.location == "New York").select(

You can also concatenate multiple fields for ordering using the | operator:

db(Event.location == "New York").select(


When you select records, you often want to limit the result to a specific number of records, and use pagination to get the consequent results. Emmett provides the paginate option in the select method, so for example


will return the first page of results, with 10 events per page. You can specify the size of the page using a tuple, so that

Event.all().select(paginate=(2, 25))

will return the second page, with 25 events per page.

Note: remember that paginate will always consider the first page number as 1, not 0

Emmett provides also a more sql-like option for limiting the results, the limitby one, that has the same syntax of the sql LIMIT BY instruction:

Event.all().select(limitby=(25, 50))

with the starting offset and the ending one. This line of code will produce the same result of using paginate=(2, 25).


When you need to aggregate the rows with the same values for specific columns, you can use the groupby option of the select method. For example, you can select all the locations for events in 2015:

db(Event.happens_at.year() == 2015).select(

You can also specify a grouping condition, for example aggregate only records that have 300 or more participants:

db(Event.happens_at.year() == 2015).select(
    having=(Event.participants >= 300))

The argument of having should be a query with the same syntax you used for db.where().

The select method also provides a distinct option, that has the same effect as grouping using all specified fields:

db(Event.happens_at.year() == 1955).select(

Counting and expressions

Among with the select method, sets come with a count method:

>>> Event.all().count()

But also fields have a count method. This is useful when you do aggregation as we seen in the above paragraph; for example you may want to count the number of events happened in 2015 grouped by their locations:

count =
db(Event.happens_at.year() == 2015).select(

The resulting rows will be something like this:

<Row {'events': {'location': 'New York'}, '_extra': {'COUNT(': 2}}>

And you can, for example, print the values using:

>>> for row in rows:
...     print(, row[count])
Chicago 1
New York 2

As you can see, you can access the count value using the variable as item of the row. Also notice that Emmett moved the location field into the events dictionary. This is done because you added elements that don't belongs to the events table itself, and Emmett wants to make this very explicit, grouping all the elements belonging to the table into a separated key of the rows.

Beside the count method, fields also have other methods useful to compute values from the records: the sum, avg, min, and max methods. They work all the same, like the count one. Let's say for example that you want to have the sum of all the participants to events in 1955 grouped by their locations:

summed = Event.participants.sum()
db(Event.happens_at.year() == 1955).select(

You will have the same result structure we've seen for count.

Updating records

When you need to update existing data inside your database, you can use two different methods, the first one is the update method of the Set object:

>>> db(Event.happens_at.year() == 1955).update(location="Hill Valley")

The update method accepts the column names and the values to change as named arguments, and it will update all the records corresponding to the set you have queried. The return value of the update method is, indeed, the number of records updated.

Since the update record is atomic, it also accepts expression built with model fields as arguments. As an example, you can increment a value:

db(Event.location == "Hill Valley").update(

As we've just seen, the update method is built on top of the Set object, so when you want to update a specific record, you should query for its id (or a combination of other values that makes the record unique):

db( == 1).update(participants=3)

But this is not the only option, in fact the Row object has an update_record method, which is the second method in Emmett to update an existing record. In order to use this method, you should have a selected row with the id included in the selected fields.

This will produce the same result of the last example:

>>> row = Event.get(1)
>>> row.update_record(participants=3)
<Row {'id': 1, 'location': 'Hill Valley' ...}>

where the main difference is that you made a SELECT sql operation and then an UPDATE one, while in the other example you did just the second one. Also, the update_record return the Row object updated to reflect the changed database record, instead of an integer.

Note: Row.update_record should not be confused with Row.update, that will change the Row object but not the database record.

Mind that, writing lines like this:

row = Event.get(1)

won't produce an atomic update on the record, but will just write to the database the last selected value plus one. If you're intended to increment a value, you should use the update method of the Set with the expression as parameter, as we've seen before.

Validation on updates

Now, since update and update_record won't trigger validations before effectively update the records in the database, Emmett also provides a validate_and_update method on the Set object, which works pretty the same of the update one:

>>> db( == 1).validate_and_update(location="New York")
<Row {'updated': 1, 'errors': {}}>

except that it will trigger the validation on the values and the effective update of the records only on its success.
As you can see the return value of the validate_and_update method will be a Row object containing the number of updated records under the updated attribute and the validation errors (if any) under the errors one.

Deleting records

Like for the update of records, Emmett provides two different methods to delete records:

  • the delete method on the Set object
  • the delete_record method of the Row object

Here are two examples:

>>> db(Event.location == "New York").delete()
>>> row = Event.get(3)
>>> row.delete_record()

As you can see both of these methods return the number of record removed.

Note: just like the update_record, the delete_record method requires you to select the id field in the rows.

Using model objects

Emmett also provides the ability to work directly with records, in addition to models' operations.

We will use the same Event model we presented in the above sections for the examples:

class Event(Model):
    name = Field(notnull=True)
    location = Field(notnull=True)
    participants =
    happens_at = Field.datetime()

Let's see all the available methods and steps in details.

Model new method

Every Model in Emmett provides a new method, which produces a clean record:

event =
    location="Hill Valley"
event.happens_at = datetime(
    1955, 11, 12,
    22, 4, 0

Records produced from the new method won't have primary key(s), and will have valued fields with defaults and passed parameters.

Record save method

New in version 2.4

Records produced with or selected with all the model fields will have a save method.

The save methods performs an insert or an update accordingly to the record contents:

# save() will produce an insert
event =
# save() will produce an update
event = Event.first()
event.location = "New York"

Note: differently from update or update_record methods, where you specify which fields should be updated, the save method will overwrite all the fields with the current record contents

The save method will return a boolean representing the operation fulfillment, unless you call save(raise_on_error=True) which will produce an exception.

Note: save will trigger both save callbacks and insert/update ones

Record destroy method

New in version 2.4

Records selected with all the model fields will have a destroy method.

The destroy methods performs a delete operation using the records' primary key(s):

event = Event.first()

The destroy method will return a boolean representing the operation fulfillment, unless you call destroy(raise_on_error=True) which will produce an exception.

Note: destroy will trigger both destroy callbacks and delete ones

Record refresh method

New in version 2.4

Records selected with all the model fields will have a refresh method.

The refresh methods performs a new selection of the record from the database and update the current object accordingly.

event = Event.first()

The refresh method will return a boolean representing the operation fulfillment.

Record changes

New in version 2.4

Records produced with or selected with all the model fields will track changes to their attributes between saves.

Here we list attributes and methods provided by Emmett to deal with row changes:

name type description
has_changed attribute boolean which states if record has changed
has_changed_value method returns a boolean wich states if the specified attribute has changed
get_value_change method returns a tuple containing the original and new values for the specified attribute or None
changes attribute returns a sdict with all the changed attributes and their values

And here is an example:

>>> event = Event.first()
>>> event.location = "New York"
>>> event.has_changed
>>> event.has_changed_value("location")
>>> event.get_value_change("location")
('Hill Valley', 'New York')
>>> event.changes
<sdict {'location': ('Hill Valley', 'New York')}>

Record validations

New in version 2.4

Records produced with or selected with all the model fields will provide helpers for validation.

Here we list attributes and methods provided by Emmett:

name type description
is_valid attribute boolean which states if record passes validations
validation_errors attribute returns a sdict with all the validation errors

And here is an example:

>>> event ="Lightning", location="Hill Valley")
>>> event.is_valid
>>> event.validation_errors
<sdict {'happens_at': 'Cannot be empty'}>
>>> event.happens_at = datetime(1955, 11, 12, 22, 4, 0)
>>> event.is_valid