A fast, user friendly ORM and query builder which supports asyncio.
Lots of internal improvements, mostly to support new functionality in Piccolo Admin.
array_columns to Table._meta by @dantownsend in https://github.com/piccolo-orm/piccolo/pull/954
Array column for getting the number of dimensions of the array by @dantownsend in https://github.com/piccolo-orm/piccolo/pull/956
where clauses work in the docs by @dantownsend in https://github.com/piccolo-orm/piccolo/pull/963
Full Changelog: https://github.com/piccolo-orm/piccolo/compare/1.4.2...1.5.0
Improved how ModelBuilder handles recursive foreign keys.
Fixed an edge case with auto migrations.
If starting from a table like this, with a custom primary key column:
class MyTable(Table):
id = UUID(primary_key=True)
When a foreign key is added to the table which references itself:
class MyTable(Table):
id = UUID(primary_key=True)
fk = ForeignKey("self")
The auto migrations could fail in some situations.
Improved how create_pydantic_model handles Array columns:
Array(Array(Integer))) have more accurate types.Array(Email()) now validates that each item in the list is an email address.Array(Varchar(length=10)) now validates that each item is the correct length (i.e. 10 in this example).Some Pylance errors were fixed in the codebase.
Fixed a bug with nested array columns containing BigInt. For example:
class MyTable(Table):
my_column = Array(Array(BigInt))
Thanks to @AmazingAkai for reporting this issue.
Fixed a bug with foreign keys which reference BigSerial primary keys. Thanks to @Abdelhadi92 for reporting this issue.
Added the piccolo user list command - a quick and convenient way of listing Piccolo Admin users from the command line.
ModelBuilder now creates timezone aware datetime objects for Timestamptz columns.
Updated the ASGI templates.
SQLite auto migrations are now allowed. We used to raise an exception, but now we output a warning instead. While SQLite auto migrations aren't as feature rich as Postgres, they work fine for simple use cases.
There's now an alternative syntax for joins, which works really well with static type checkers like Mypy and Pylance.
The traditional syntax (which continues to work as before):
# Get the band name, and the manager's name from a related table
await Band.select(Band.name, Band.manager.name)
The alternative syntax is as follows:
await Band.select(Band.name, Band.manager._.name)
Note how we use ._. instead of . after a ForeignKey.
This offers a considerably better static typing experience. In the above example, type checkers know that Band.manager._.name refers to the name column on the Manager table. This means typos can be detected, and code navigation is easier.
get_related.esmerald ASGI framework.Piccolo allows the user to specify savepoint names which are used in transactions. For example:
async with DB.transaction() as transaction:
await Band.insert(Band(name='Pythonistas'))
# Passing in a savepoint name is optional:
savepoint_1 = await transaction.savepoint('savepoint_1')
await Band.insert(Band(name='Terrible band'))
# Oops, I made a mistake!
await savepoint_1.rollback_to()
Postgres doesn't allow us to parameterise savepoint names, which means there's a small chance of SQL injection, if for some reason the savepoint names were generated from end-user input. Even though the likelihood is very low, it's best to be safe. We now validate the savepoint name, to make sure it can only contain certain safe characters. Thanks to @Skelmis for making this change.
Added support for Python 3.12.
Modified create_pydantic_model, so additional information is returned in the JSON schema to distinguish between Timestamp and Timestamptz columns. This will be used for future Piccolo Admin enhancements.