ormar

Overview

Theormarpackage is an async mini ORM for Python, with support forPostgres, MySQL,andSQLite.

The main benefits of usingormarare:

• getting anasync ORM that can be used with async frameworks(fastapi, starlette etc.)
• getting justone model to maintain- you don't have to maintain pydantic and other orm models (sqlalchemy, peewee, gino etc.)

The goal was to create a simple ORM that can beused directly (as request and response models) withfastapithat bases it's data validation on pydantic.

Ormar - apart from the obvious "ORM" in name - gets its name fromormarin Swedish which meanssnakes,andormarin Croatian which meanscabinet.

And what's a better name for Python ORM than snakes cabinet:)

If you like ormar remember to star the repository ingithub!

The bigger community we build, the easier it will be to catch bugs and attract contributors;)

Documentation

Check out thedocumentationfor details.

Note that for brevity most of the documentation snippets omit the creation of the database and scheduling the execution of functions for asynchronous run.

If you want more real life examples than in the documentation you can see thetestsfolder, since they actually have to create and connect to a database in most of the tests.

Yet remember that those are - well - tests and not all solutions are suitable to be used in real life applications.

Part of thefastapiecosystem

As part of the fastapi ecosystemormaris supported in libraries that somehow work with databases.

As of nowormaris supported by:

• fastapi-crudrouter
• fastapi-pagination

If you maintain or use a different library and would like it to supportormarlet us know how we can help.

Dependencies

Ormar is built with:

• sqlalchemy corefor query building.
• databasesfor cross-database async support.
• pydanticfor data validation.
• typing_extensionsfor Python 3.6 - 3.7

License

ormaris built as open-sorce software and will remain completely free (MIT license).

As I write open-source code to solve everyday problems in my work or to promote and build strong Python community you can say thank you and buy me a coffee or sponsor me with a monthly amount to help ensure my work remains free and maintained.

Sponsor - Github Sponsors

Migrating fromsqlalchemyand existing databases

If you currently usesqlalchemyand would like to switch toormarcheck out the auto-translation tool that can help you with translating existing sqlalchemy orm models so you do not have to do it manually.

Betaversions available at github:sqlalchemy-to-ormar or simplypip install sqlalchemy-to-ormar

sqlalchemy-to-ormarcan be used in pair withsqlacodegento auto-map/ generateormarmodels from existing database, even if you don't usesqlalchemyfor your project.

Migrations & Database creation

Because ormar is built on SQLAlchemy core, you can usealembicto provide database migrations (and you really should for production code).

For tests and basic applications thesqlalchemyis more than enough:

# note this is just a partial snippet full working example below
# 1. Imports
importsqlalchemy
importdatabases

# 2. Initialization
DATABASE_URL="sqlite:///db.sqlite"
database=databases.Database(DATABASE_URL)
metadata=sqlalchemy.MetaData()

# Define models here

# 3. Database creation and tables creation
engine=sqlalchemy.create_engine(DATABASE_URL)
metadata.create_all(engine)

For a sample configuration of alembic and more information regarding migrations and database creation visitmigrationsdocumentation section.

Package versions

ormar is still under development: We recommend pinning any dependencies (with i.e.ormar~=0.9.1)

ormaralso follows the release numeration that breaking changes bump the major number, while other changes and fixes bump minor number, so with the latter you should be safe to update, yet always read thereleasesdocs before. example: (0.5.2 -> 0.6.0 - breaking, 0.5.2 -> 0.5.3 - non breaking).

Asynchronous Python

Note thatormaris an asynchronous ORM, which means that you have toawaitthe calls to the methods, that are scheduled for execution in an event loop. Python has a builtin module asynciothat allows you to do just that.

Note that most "normal" Python interpreters do not allow execution ofawait outside of a function (because you actually schedule this function for delayed execution and don't get the result immediately).

In a modern web framework (likefastapi), the framework will handle this for you, but if you plan to do this on your own you need to perform this manually like described in the quick start below.

Quick Start

Note that you can find the same script in examples folder on github.

fromtypingimportOptional

importdatabases
importpydantic

importormar
importsqlalchemy

DATABASE_URL="sqlite:///db.sqlite"
base_ormar_config=ormar.OrmarConfig(
database=databases.Database(DATABASE_URL),
metadata=sqlalchemy.MetaData(),
engine=sqlalchemy.create_engine(DATABASE_URL),
)


# Note that all type hints are optional
# below is a perfectly valid model declaration
# class Author(ormar.Model):
# ormar_config = base_ormar_config.copy(tablename= "authors" )
#
# id = ormar.Integer(primary_key=True) # <= notice no field types
# name = ormar.String(max_length=100)


classAuthor(ormar.Model):
ormar_config=base_ormar_config.copy(tablename="authors")

id:int=ormar.Integer(primary_key=True)
name:str=ormar.String(max_length=100)


classBook(ormar.Model):
ormar_config=base_ormar_config.copy(tablename="books")

id:int=ormar.Integer(primary_key=True)
author:Optional[Author]=ormar.ForeignKey(Author)
title:str=ormar.String(max_length=100)
year:int=ormar.Integer(nullable=True)


# create the database
# note that in production you should use migrations
# note that this is not required if you connect to existing database
# just to be sure we clear the db before
base_ormar_config.metadata.drop_all(base_ormar_config.engine)
base_ormar_config.metadata.create_all(base_ormar_config.engine)


# all functions below are divided into functionality categories
# note how all functions are defined with async - hence can use await AND needs to
# be awaited on their own
asyncdefcreate():
# Create some records to work with through QuerySet.create method.
# Note that queryset is exposed on each Model's class as objects
tolkien=awaitAuthor.objects.create(name="J.R.R. Tolkien")
awaitBook.objects.create(author=tolkien,title="The Hobbit",year=1937)
awaitBook.objects.create(author=tolkien,title="The Lord of the Rings",year=1955)
awaitBook.objects.create(author=tolkien,title="The Silmarillion",year=1977)

# alternative creation of object divided into 2 steps
sapkowski=Author(name="Andrzej Sapkowski")
# do some stuff
awaitsapkowski.save()

# or save() after initialization
awaitBook(author=sapkowski,title="The Witcher",year=1990).save()
awaitBook(author=sapkowski,title="The Tower of Fools",year=2002).save()

# to read more about inserting data into the database
# visit: https://collerek.github.io/ormar/queries/create/


asyncdefread():
# Fetch an instance, without loading a foreign key relationship on it.
# Django style
book=awaitBook.objects.get(title="The Hobbit")
# or Python style
book=awaitBook.objects.get(Book.title=="The Hobbit")
book2=awaitBook.objects.first()

# first() fetch the instance with lower primary key value
assertbook==book2

# you can access all fields on loaded model
assertbook.title=="The Hobbit"
assertbook.year==1937

# when no condition is passed to get()
# it behaves as last() based on primary key column
book3=awaitBook.objects.get()
assertbook3.title=="The Tower of Fools"

# When you have a relation, ormar always defines a related model for you
# even when all you loaded is a foreign key value like in this example
assertisinstance(book.author,Author)
# primary key is populated from foreign key stored in books table
assertbook.author.pk==1
# since the related model was not loaded all other fields are None
assertbook.author.nameisNone

# Load the relationship from the database when you already have the related model
# alternatively see joins section below
awaitbook.author.load()
assertbook.author.name=="J.R.R. Tolkien"

# get all rows for given model
authors=awaitAuthor.objects.all()
assertlen(authors)==2

# to read more about reading data from the database
# visit: https://collerek.github.io/ormar/queries/read/


asyncdefupdate():
# read existing row from db
tolkien=awaitAuthor.objects.get(name="J.R.R. Tolkien")
asserttolkien.name=="J.R.R. Tolkien"
tolkien_id=tolkien.id

# change the selected property
tolkien.name="John Ronald Reuel Tolkien"
# call update on a model instance
awaittolkien.update()

# confirm that object was updated
tolkien=awaitAuthor.objects.get(name="John Ronald Reuel Tolkien")
asserttolkien.name=="John Ronald Reuel Tolkien"
asserttolkien.id==tolkien_id

# alternatively update data without loading
awaitAuthor.objects.filter(name__contains="Tolkien").update(name="J.R.R. Tolkien")

# to read more about updating data in the database
# visit: https://collerek.github.io/ormar/queries/update/


asyncdefdelete():
silmarillion=awaitBook.objects.get(year=1977)
# call delete() on instance
awaitsilmarillion.delete()

# alternatively delete without loading
awaitBook.objects.delete(title="The Tower of Fools")

# note that when there is no record ormar raises NoMatch exception
try:
awaitBook.objects.get(year=1977)
exceptormar.NoMatch:
print("No book from 1977!")

# to read more about deleting data from the database
# visit: https://collerek.github.io/ormar/queries/delete/

# note that despite the fact that record no longer exists in database
# the object above is still accessible and you can use it (and i.e. save()) again.
tolkien=silmarillion.author
awaitBook.objects.create(author=tolkien,title="The Silmarillion",year=1977)


asyncdefjoins():
# Tho join two models use select_related

# Django style
book=awaitBook.objects.select_related("author").get(title="The Hobbit")
# Python style
book=awaitBook.objects.select_related(Book.author).get(
Book.title=="The Hobbit"
)

# now the author is already prefetched
assertbook.author.name=="J.R.R. Tolkien"

# By default you also get a second side of the relation
# constructed as lowercase source model name +'s' (books in this case)
# you can also provide custom name with parameter related_name

# Django style
author=awaitAuthor.objects.select_related("books").all(name="J.R.R. Tolkien")
# Python style
author=awaitAuthor.objects.select_related(Author.books).all(
Author.name=="J.R.R. Tolkien"
)
assertlen(author[0].books)==3

# for reverse and many to many relations you can also prefetch_related
# that executes a separate query for each of related models

# Django style
author=awaitAuthor.objects.prefetch_related("books").get(name="J.R.R. Tolkien")
# Python style
author=awaitAuthor.objects.prefetch_related(Author.books).get(
Author.name=="J.R.R. Tolkien"
)
assertlen(author.books)==3

# to read more about relations
# visit: https://collerek.github.io/ormar/relations/

# to read more about joins and subqueries
# visit: https://collerek.github.io/ormar/queries/joins-and-subqueries/


asyncdeffilter_and_sort():
# to filter the query you can use filter() or pass key-value pars to
# get(), all() etc.
# to use special methods or access related model fields use double
# underscore like to filter by the name of the author use author__name
# Django style
books=awaitBook.objects.all(author__name="J.R.R. Tolkien")
# Python style
books=awaitBook.objects.all(Book.author.name=="J.R.R. Tolkien")
assertlen(books)==3

# filter can accept special methods also separated with double underscore
# to issue sql query ` where authors.name like "%tolkien%" ` that is not
# case sensitive (hence small t in Tolkien)
# Django style
books=awaitBook.objects.filter(author__name__icontains="tolkien").all()
# Python style
books=awaitBook.objects.filter(Book.author.name.icontains("tolkien")).all()
assertlen(books)==3

# to sort use order_by() function of queryset
# to sort decreasing use hyphen before the field name
# same as with filter you can use double underscores to access related fields
# Django style
books=(
awaitBook.objects.filter(author__name__icontains="tolkien")
.order_by("-year")
.all()
)
# Python style
books=(
awaitBook.objects.filter(Book.author.name.icontains("tolkien"))
.order_by(Book.year.desc())
.all()
)
assertlen(books)==3
assertbooks[0].title=="The Silmarillion"
assertbooks[2].title=="The Hobbit"

# to read more about filtering and ordering
# visit: https://collerek.github.io/ormar/queries/filter-and-sort/


asyncdefsubset_of_columns():
# to exclude some columns from loading when querying the database
# you can use fields() method
hobbit=awaitBook.objects.fields(["title"]).get(title="The Hobbit")
# note that fields not included in fields are empty (set to None)
asserthobbit.yearisNone
asserthobbit.authorisNone

# selected field is there
asserthobbit.title=="The Hobbit"

# alternatively you can provide columns you want to exclude
hobbit=awaitBook.objects.exclude_fields(["year"]).get(title="The Hobbit")
# year is still not set
asserthobbit.yearisNone
# but author is back
asserthobbit.authorisnotNone

# also you cannot exclude primary key column - it's always there
# even if you EXPLICITLY exclude it it will be there

# note that each model have a shortcut for primary_key column which is pk
# and you can filter/access/set the values by this alias like below
asserthobbit.pkisnotNone

# note that you cannot exclude fields that are not nullable
# (required) in model definition
try:
awaitBook.objects.exclude_fields(["title"]).get(title="The Hobbit")
exceptpydantic.ValidationError:
print("Cannot exclude non nullable field title")

# to read more about selecting subset of columns
# visit: https://collerek.github.io/ormar/queries/select-columns/


asyncdefpagination():
# to limit number of returned rows use limit()
books=awaitBook.objects.limit(1).all()
assertlen(books)==1
assertbooks[0].title=="The Hobbit"

# to offset number of returned rows use offset()
books=awaitBook.objects.limit(1).offset(1).all()
assertlen(books)==1
assertbooks[0].title=="The Lord of the Rings"

# alternatively use paginate that combines both
books=awaitBook.objects.paginate(page=2,page_size=2).all()
assertlen(books)==2
# note that we removed one book of Sapkowski in delete()
# and recreated The Silmarillion - by default when no order_by is set
# ordering sorts by primary_key column
assertbooks[0].title=="The Witcher"
assertbooks[1].title=="The Silmarillion"

# to read more about pagination and number of rows
# visit: https://collerek.github.io/ormar/queries/pagination-and-rows-number/


asyncdefaggregations():
# count:
assert2==awaitAuthor.objects.count()

# exists
assertawaitBook.objects.filter(title="The Hobbit").exists()

# maximum
assert1990==awaitBook.objects.max(columns=["year"])

# minimum
assert1937==awaitBook.objects.min(columns=["year"])

# average
assert1964.75==awaitBook.objects.avg(columns=["year"])

# sum
assert7859==awaitBook.objects.sum(columns=["year"])

# to read more about aggregated functions
# visit: https://collerek.github.io/ormar/queries/aggregations/


asyncdefraw_data():
# extract raw data in a form of dicts or tuples
# note that this skips the validation(!) as models are
# not created from parsed data

# get list of objects as dicts
assertawaitBook.objects.values()==[
{"id":1,"author":1,"title":"The Hobbit","year":1937},
{"id":2,"author":1,"title":"The Lord of the Rings","year":1955},
{"id":4,"author":2,"title":"The Witcher","year":1990},
{"id":5,"author":1,"title":"The Silmarillion","year":1977},
]

# get list of objects as tuples
assertawaitBook.objects.values_list()==[
(1,1,"The Hobbit",1937),
(2,1,"The Lord of the Rings",1955),
(4,2,"The Witcher",1990),
(5,1,"The Silmarillion",1977),
]

# filter data - note how you always get a list
assertawaitBook.objects.filter(title="The Hobbit").values()==[
{"id":1,"author":1,"title":"The Hobbit","year":1937}
]

# select only wanted fields
assertawaitBook.objects.filter(title="The Hobbit").values(["id","title"])==[
{"id":1,"title":"The Hobbit"}
]

# if you select only one column you could flatten it with values_list
assertawaitBook.objects.values_list("title",flatten=True)==[
"The Hobbit",
"The Lord of the Rings",
"The Witcher",
"The Silmarillion",
]

# to read more about extracting raw values
# visit: https://collerek.github.io/ormar/queries/aggregations/


asyncdefwith_connect(function):
# note that for any other backend than sqlite you actually need to
# connect to the database to perform db operations
asyncwithbase_ormar_config.database:
awaitfunction()

# note that if you use framework like `fastapi` you shouldn't connect
# in your endpoints but have a global connection pool
# check https://collerek.github.io/ormar/fastapi/ and section with db connection


# gather and execute all functions
# note - normally import should be at the beginning of the file
importasyncio

# note that normally you use gather() function to run several functions
# concurrently but we actually modify the data and we rely on the order of functions
forfuncin[
create,
read,
update,
delete,
joins,
filter_and_sort,
subset_of_columns,
pagination,
aggregations,
raw_data,
]:
print(f "Executing:{func.__name__}")
asyncio.run(with_connect(func))

# drop the database tables
base_ormar_config.metadata.drop_all(base_ormar_config.engine)

Ormar Specification

QuerySet methods

• create(**kwargs): -> Model
• get(*args, **kwargs): -> Model
• get_or_none(*args, **kwargs): -> Optional[Model]
• get_or_create(_defaults: Optional[Dict[str, Any]] = None, *args, **kwargs) -> Tuple[Model, bool]
• first(*args, **kwargs): -> Model
• update(each: bool = False, **kwargs) -> int
• update_or_create(**kwargs) -> Model
• bulk_create(objects: List[Model]) -> None
• bulk_update(objects: List[Model], columns: List[str] = None) -> None
• delete(*args, each: bool = False, **kwargs) -> int
• all(*args, **kwargs) -> List[Optional[Model]]
• iterate(*args, **kwargs) -> AsyncGenerator[Model]
• filter(*args, **kwargs) -> QuerySet
• exclude(*args, **kwargs) -> QuerySet
• select_related(related: Union[List, str]) -> QuerySet
• prefetch_related(related: Union[List, str]) -> QuerySet
• limit(limit_count: int) -> QuerySet
• offset(offset: int) -> QuerySet
• count(distinct: bool = True) -> int
• exists() -> bool
• max(columns: List[str]) -> Any
• min(columns: List[str]) -> Any
• avg(columns: List[str]) -> Any
• sum(columns: List[str]) -> Any
• fields(columns: Union[List, str, set, dict]) -> QuerySet
• exclude_fields(columns: Union[List, str, set, dict]) -> QuerySet
• order_by(columns:Union[List, str]) -> QuerySet
• values(fields: Union[List, str, Set, Dict])
• values_list(fields: Union[List, str, Set, Dict])

Relation types

• One to many - withForeignKey(to: Model)
• Many to many - withManyToMany(to: Model, Optional[through]: Model)

Model fields types

Available Model Fields (with required args - optional ones in docs):

• String(max_length)
• Text()
• Boolean()
• Integer()
• Float()
• Date()
• Time()
• DateTime()
• JSON()
• BigInteger()
• SmallInteger()
• Decimal(scale, precision)
• UUID()
• LargeBinary(max_length)
• Enum(enum_class)
• Enumlike Field - by passingchoicesto any other Field type
• EncryptedString- by passingencrypt_secretandencrypt_backend
• ForeignKey(to)
• ManyToMany(to)

Available fields options

The following keyword arguments are supported on all field types.

• primary_key: bool
• nullable: bool
• default: Any
• server_default: Any
• index: bool
• unique: bool
• choices: typing.Sequence
• name: str

All fields are required unless one of the following is set:

• nullable- Creates a nullable column. Sets the default toFalse.Read the fields common parameters for details.
• sql_nullable- Used to set different setting for pydantic and the database. Sets the default tonullablevalue. Read the fields common parameters for details.
• default- Set a default value for the field.Not available for relation fields
• server_default- Set a default value for the field on server side (like sqlalchemy'sfunc.now()).Not available for relation fields
• primary keywithautoincrement- When a column is set to primary key and autoincrement is set on this column. Autoincrement is set by default on int primary keys.

Available signals

Signals allow to trigger your function for a given event on a given Model.

• pre_save
• post_save
• pre_update
• post_update
• pre_delete
• post_delete
• pre_relation_add
• post_relation_add
• pre_relation_remove
• post_relation_remove
• post_bulk_update