Skip to main content

Manage SQL queries as a Python API

Project description

aeSQLAPIus

Packaging status

CI codecov PyPI downloads Github commits (since latest release)

Summary

So you don't want to use ORM, and want to organize your SQL queries in a convenient way. Don't mix them with your python code, don't write execute and fetchrows by hand for each query. With aesqlapius:

  • Store your SQL queries separate from the code, in a dedicated file or directory hierarchy
  • Annotate each query with python-like function definition specifying input arguments and output types and patterns

aesqlapius builds a class out of this, where you can call your queries as plain methods. It handles arguments (pass positional or keyword arguments as you like, default values are also handled) and output types and patterns (you may specify whether a method returns iterator, list, dict of rows, or a single row, where row may be represented as a tuple, list, dict, single value or a custom type such as a dataclass).

Example

queries.sql:

-- def add_city(name, population = None) -> None: ...
INSERT INTO cities VALUES (%(name)s, %(population)s);

-- def list_cities() -> List[Single]: ...
SELECT name FROM cities ORDER BY name;

-- def get_population(city: str) -> Single[Single]: ...
SELECT population FROM cities WHERE name = %(city)s;

-- def get_city(city: str) -> Single[City]: ...
SELECT * FROM cities WHERE name = %(city)s;

script.py:

from aesqlapius import generate_api

db = psycopg2.connect('...')
api = generate_api('queries.sql', 'psycopg2', db)

api.add_city('Moscow', 12500000)
api.add_city('Paris')
api.add_city(population=380000, name='Berlin')

assert api.list_cities() == ['Berlin', 'Moscow', 'Paris']
assert api.get_population('Moscow') == 12500000

@dataclass
class City:
    name: str
    population: Optional[int]

assert api.get_city('Berlin') == City('Berlin', 3800000)

Reference

Python API

The module has a single entry point in form of a function:

def generate_api(path, driver, db=None, *, target=None, extension='.sql', namespace_mode='dirs', namespace_root='__init__')

This loads SQL queries from path (a file or directory) and returns an API class to use with specified database driver (psycopg2, sqlite3).

If db is specified, all generated methods are bound to the given database connection object:

db = psycopg2.connect('...')
api = generate_api('queries.sql', 'psycopg2', db)
api.my_method('arg1', 'arg2')

otherwise caller is expected to pass database connection object to each call:

db = psycopg2.connect('...')
api = generate_api('queries.sql', 'psycopg2')
api.my_method(db, 'arg1', 'arg2')

If target is specified, methods are injected into the given object (which is also returned from generate_api):

db = psycopg2.connect('...')
generate_api('queries.sql', 'psycopg2', db, target=db)
db.my_method('arg1', 'arg2')

extension (by default .sql) specifies which files are loaded from the queries directory.

namespace_mode controls how hierarchy of files is converted into hierarchy of objects when constructing the API class. There are 3 modes supported:

  • dirs (the default), which maps each subdirectory to a nested method namespace ignoring file names:
path under query dir function definition resulting API
root.sql -- def a(): ... api.a()
subdir/foo.sql -- def b(): ... api.subdir.b()
subdir/bar.sql -- def c(): ... api.subdir.c()
  • files which uses file names (after stripping the extension) as an extra nesting level:
path under query dir function resulting API
root.sql -- def a(): ... api.root.a()
subdir/foo.sql -- def b(): ... api.subdir.foo.b()
subdir/bar.sql -- def c(): ... api.subdir.bar.c()

In this mode, namespace_root allows to specify a special file name which circumvents this behavior, allowing to mimic how Python handles module namespaces. For example, when namespace_root = "__init__" (the default):

path under query dir function resulting API
__init__.sql -- def a(): ... api.a()
foo.sql -- def b(): ... api.foo.b()
subdir/__init__.sql -- def c(): ... api.subdir.c()
subdir/bar.sql -- def d(): ... api.subdir.bar.d()
  • flat mode which ignores hierarchy:
path under query dir function resulting API
root.sql -- def a(): ... api.a()
subdir/foo.sql -- def b(): ... api.b()
subdir/bar.sql -- def c(): ... api.c()

Query annotations

Each query managed by aesqlapius must be preceded with a -- (SQL comment) followed by a Python-style function definition:

-- def function_name(parameters, ...) -> return_type: ...
...some SQL code...

Parameters

Parameters allow optional literal default values and optional type annotations (which are currently ignored) and may be specified in both positional, keyword or mixed style in the resulting API:

-- def myfunc(foo, bar: str, baz=123) -> None: ...`
...some SQL code...
api.myfunc(1, bar="sometext")  # foo=1, bar="sometext", baz=123

Return value

Return value annotation is required and may either be None (when query does not return anything) or a nested type annotation with specific structure RowsFormat[RowFormat].

Outer RowsFormat specifies how multiple rows returned by the query are handled. Allowed values are:

  • Iterator[RowFormat] - return a row iterator.
  • List[RowFormat] - return a list of rows.
  • Single[RowFormat] - return a single row.
  • Dict[KeyColumn, RowFormat] - return a dictionary of rows. The column to be used as a dictionary key is specified in the first argument, e.g. Dict[0, ...] uses first returned column as key and `Dict['colname', ...] uses column named colname. Precede column index or name with unary minus to make it removed from the row contents.

Inner RowFormat specifies how data for each row is presented:

  • Tuple - return row as a tuple of values.
  • List - return row as a list of values.
  • Dict - return row as a dict, where keys are set to the column names returned by the query.
  • arbitrary type - TODO

Examples:

-- def example1() -> List[Tuple]: ...
SELECT 1, 'foo' UNION SELECT 2, 'bar';
-- def example2() -> Single[Tuple]: ...
SELECT 1, 'foo';
-- def example3() -> Iterator[Dict]: ...
SELECT 1 AS n, 'foo' AS s UNION SELECT 2 AS n, 'bar' AS s;
-- def example4() -> Dict['key', Dict]: ...
SELECT 'foo' AS key, 1 AS a, 2 AS b;
-- def example5() -> Dict[-'key', Dict]: ...
SELECT 'foo' AS key, 1 AS a, 2 AS b;
>>> api.example1()
[(1, 'foo'), (2, 'bar')]
>>> api.example2()
(1, 'foo')
>>> it = api.example3()
>>> next(it)
{'n': 1, 's': 'foo'}
>>> next(it)
{'n': 2, 's': 'bar'}
>>> api.example4()
{'foo': {'key': 'foo', 'a': 1, 'b': 2}}
>>> api.example5()
{'foo': {'a': 1, 'b': 2}}

Body

Function body is required to contain a single ellipsis.

License

MIT license, copyright (c) 2020 Dmitry Marakasov amdmi3@amdmi3.ru.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

aesqlapius-0.0.4.tar.gz (18.9 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page