CuttlePool
CuttlePool is a general purpose, thread-safe resource pooling implementation for use with long lived resources and/or resources that are expensive to instantiate. It's key features are:
- Pool overflow
- Creates additional resources if the pool capacity has been reached and will remove the overflow when demand for resources decreases.
- Resource harvesting
- Any resources that haven't been returned to the pool and are no longer referenced by anything outside the pool are returned to the pool. This helps prevent pool depletion when resources aren't explicitly returned to the pool and the resource wrapper is garbage collected.
- Resource queuing
- If all else fails and no resource can be immediately found or made, the pool will wait a specified amount of time for a resource to be returned to the pool before raising an exception.
How-to Guide
Using CuttlePool requires subclassing a CuttlePool
object with a user
defined method normalize_resource()
and ping()
. The example below uses
sqlite3
connections as a resource, but CuttlePool is not limited to
connection drivers.
>>> import sqlite3 >>> from cuttlepool import CuttlePool >>> class SQLitePool(CuttlePool): ... def normalize_resource(self, resource): ... resource.row_factory = None ... def ping(self, resource): ... try: ... rv = resource.execute('SELECT 1').fetchall() ... return (1,) in rv ... except sqlite3.Error: ... return False >>> pool = SQLitePool(factory=sqlite3.connect, database='ricks_lab')
Let's break this down line by line.
First, the sqlite3
module is imported. sqlite3.connect
will be the
underlying resource factory.
CuttlePool
is imported and subclassed. The normalize_resource()
method takes a resource, in this case a sqlite3.Connection
instance created
by sqlite3.connect
, as a parameter and changes it's properties. This is
important because a resource can be modified while it's outside of the pool and
any modifications made during that time will persist; this can have unintended
consequences when the resource is later retrieved from the pool.
Next the ping()
method is implemented, which also takes a resource, the
same as normalize_resource()
, as a parameter. ping()
ensures the
resource is functional; in this case, it checks that the sqlite3.Connection
instance is open. If the resource is functional, ping()
returns True
else it returns False
. In the above example, a simple statement is executed
and if the expected result is returned, it means the resource is open and
True
is returned. The implementation of this method is really dependent on
the resource created by the pool and may not even be necessary.
Finally an instance of SQLitePool
is made. The sqlite3.connect
method is
passed to the instance along with the database name.
The CuttlePool
object and as a result the SQLitePool
object accepts any
parameters that the underlying resource factory accepts as keyword arguments.
There are three other parameters the pool object accepts that are unrelated to
the resource factory. capacity
sets the max number of resources the pool
will hold at any given time. overflow
sets the max number of additional
resources the pool will create when depleted. All overflow resources will be
removed from the pool if the pool is at capacity. timeout
sets the amount
of time in seconds the pool will wait for a resource to become free if the pool
is depleted when a request for a resource is made.
A resource from the pool can be treated the same way as an instance created by
the resource factory passed to the pool. In our example a resource can be used
just like a sqlite3.Connection
instance.
>>> con = pool.get_resource() >>> cur = con.cursor() >>> cur.execute(('INSERT INTO garage (invention_name, state) ' ... 'VALUES (%s, %s)'), ('Space Cruiser', 'damaged')) >>> con.commit() >>> cur.close() >>> con.close()
Calling close()
on the resource returns it to the pool instead of closing
it. It is not necessary to call close()
though. The pool tracks resources
so any unreferenced resources will be collected and returned to the pool. It is
still a good idea to call close()
though, since explicit is better than
implicit.
Note
Once close()
is called on the resource object, it renders the
object useless. The resource object received from the pool is a wrapper
around the actual resource object and calling close()
on it returns
the resource to the pool and removes it from the wrapper effectively
leaving it an empty shell to be garbage collected.
To automatically "close" resources, get_resource()
can be used in a
with
statement.
>>> with pool.get_resource() as con: ... cur = con.cursor() ... cur.execute(('INSERT INTO garage (invention_name, state) ' ... 'VALUES (%s, %s)'), ('Space Cruiser', 'damaged')) ... con.commit() ... cur.close()
API
The API can be found at read the docs.
FAQ
How do I install it?
pip install cuttlepool
Contributing
It's highly recommended to develop in a virtualenv.
Fork the repository.
Clone the repository:
git clone https://github.com/<your_username>/cuttlepool.git
Install the package in editable mode:
cd cuttlepool pip install -e .[dev]
Now you're set. See the next section for running tests.
Running the tests
Tests can be run with the command pytest
.
Where can I get help?
If you haven't read the How-to guide above, please do that first. Otherwise, check the issue tracker. Your issue may be addressed there and if it isn't please file an issue :)