Blockbuster is a Python package designed to detect and prevent blocking calls within an asynchronous event loop. It is particularly useful when executing tests to ensure that your asynchronous code does not inadvertently call blocking operations, which can lead to performance bottlenecks and unpredictable behavior.
In Python, the asynchronous event loop allows for concurrent execution of tasks without the need for multiple threads or processes. This is achieved by running tasks cooperatively, where tasks yield control back to the event loop when they are waiting for I/O operations or other long-running tasks to complete.
However, blocking calls, such as file I/O operations or certain networking operations, can halt the entire event loop, preventing other tasks from running. This can lead to increased latency and reduced performance, defeating the purpose of using asynchronous programming.
The difficulty with blocking calls is that they are not always obvious, especially when working with third-party libraries or legacy code. This is where Blockbuster comes in: it helps you identify and eliminate blocking calls in your codebase during testing, ensuring that your asynchronous code runs smoothly and efficiently. It does this by wrapping common blocking functions and raising an exception when they are called within an asynchronous context.
Notes:
- Blockbuster currently only detects
asyncioevent loops. - Blockbuster is tested only with CPython. It may work with other Python implementations if it's possible to monkey-patch the functions with
setattr.
The package is named blockbuster.
For instance with pip:
pip install blockbusterIt is recommended to constrain the version of Blockbuster.
Blockbuster doesn't strictly follow semver. Breaking changes such as new rules added may be introduced between minor versions, but not between patch versions.
So it is recommended to constrain the Blockbuster version on the minor version.
For instance, with uv:
uv add "blockbuster>=1.5.5,<1.6"To activate BlockBuster manually, create an instance of the BlockBuster class and call the activate() method:
from blockbuster import BlockBuster
blockbuster = BlockBuster()
blockbuster.activate()Once activated, BlockBuster will raise a BlockingError exception whenever a blocking call is detected within an asyncio event loop.
To deactivate BlockBuster, call the deactivate() method:
blockbuster.deactivate()BlockBuster can also be activated using a context manager, which automatically activates and deactivates the checks within the with block:
from blockbuster import blockbuster_ctx
with blockbuster_ctx():
# Your test code hereBlockbuster is intended to be used with testing frameworks like pytest to catch blocking calls.
Here's how you can integrate Blockbuster into your pytest test suite:
import pytest
import time
from blockbuster import BlockBuster, BlockingError, blockbuster_ctx
from typing import Iterator
@pytest.fixture(autouse=True)
def blockbuster() -> Iterator[BlockBuster]:
with blockbuster_ctx() as bb:
yield bb
async def test_time_sleep() -> None:
with pytest.raises(BlockingError, match="sleep"):
time.sleep(1) # This should raise a BlockingErrorBy using the blockbuster_ctx context manager, Blockbuster is automatically activated for every test, and blocking calls will raise a BlockingError.
Blockbuster works by wrapping common blocking functions from various modules (e.g., os, socket, time) and replacing them with versions that check if they are being called from within an asyncio event loop.
If such a call is detected, Blockbuster raises a BlockingError to indicate that a blocking operation is being performed inappropriately.
Blockbuster supports by default the following functions and modules:
- Time Functions:
time.sleep
- OS Functions:
os.getcwdos.statvfsos.sendfileos.renameos.removeos.unlinkos.mkdiros.rmdiros.linkos.symlinkos.readlinkos.listdiros.scandiros.accessos.statos.replaceos.reados.write
- OS path Functions:
os.path.ismountos.path.samestatos.path.sameopenfileos.path.islinkos.path.abspath
- IO Functions:
io.BufferedReader.readio.BufferedWriter.writeio.BufferedRandom.readio.BufferedRandom.writeio.TextIOWrapper.readio.TextIOWrapper.write
- Socket Functions:
socket.socket.connectsocket.socket.acceptsocket.socket.sendsocket.socket.sendallsocket.socket.sendtosocket.socket.recvsocket.socket.recv_intosocket.socket.recvfromsocket.socket.recvfrom_intosocket.socket.recvmsgssl.SSLSocket.writessl.SSLSocket.sendssl.SSLSocket.readssl.SSLSocket.recv
- SQLite Functions:
sqlite3.Cursor.executesqlite3.Cursor.executemanysqlite3.Cursor.executescriptsqlite3.Cursor.fetchonesqlite3.Cursor.fetchmanysqlite3.Cursor.fetchallsqlite3.Connection.executesqlite3.Connection.executemanysqlite3.Connection.executescriptsqlite3.Connection.commitsqlite3.Connection.rollback
- Thread lock Functions:
threading.Lock.acquirethreading.Lock.acquire_lock
- Built-in Functions:
input
Some exceptions to the rules are already in place:
- Importing modules does blocking calls as it interacts with the file system. Since this operation is cached and very hard to avoid, it is excluded from the detection.
- Blocking calls done by the
pydevddebugger. - Blocking calls done by the
pytestframework.
Blockbuster is not a silver bullet and may not catch all blocking calls. In particular, it will not catch blocking calls that are done by third-party libraries that do blocking calls in C extensions. For these third-party libraries, you can declare your own custom rules to Blockbuster to catch these blocking calls.
Eg.:
from blockbuster import BlockBuster, BlockBusterFunction
import mymodule
blockbuster = BlockBuster()
blockbuster.functions["my_module.my_function"] = BlockBusterFunction(my_module, "my_function")
blockbuster.activate()Note: if blockbuster has already been activated, you will need to activate the custom rule yourself.
from blockbuster import blockbuster_ctx, BlockBusterFunction
import mymodule
with blockbuster_ctx() as blockbuster:
blockbuster.functions["my_module.my_function"] = BlockBusterFunction(my_module, "my_function")
blockbuster.functions["my_module.my_function"].activate()You can customize Blockbuster to allow blocking calls in specific functions by using the can_block_in method of the BlockBusterFunction class.
This method allows you to specify exceptions for particular files and functions where blocking calls are allowed.
from blockbuster import BlockBuster
blockbuster = BlockBuster()
blockbuster.activate()
blockbuster.functions["os.stat"].can_block_in("specific_file.py", {"allowed_function"})If you need to deactivate specific checks, you can directly call the deactivate method on the corresponding BlockBusterFunction instance:
from blockbuster import BlockBuster
blockbuster = BlockBuster()
blockbuster.activate()
blockbuster.functions["socket.socket.connect"].deactivate()Contributions are welcome! If you encounter any issues or have suggestions for improvements, please open an issue on the GitHub repository.
Blockbuster uses uv to manage its development environment.
See the uv documentation for more informationand how to install it.
To install the required dependencies, run the following command:
uv syncTests are written using pytest.
To run the tests, use the following command:
uv run pytestCode formatting is done using ruff. To format the code, run the following command:
uv run ruff formatCode linting is done using ruff.
To lint the code, fixing any issues that can be automatically fixed, run the following command:
uv run ruff check --fixThis project is licensed under the Apache License 2.0 - see the LICENSE file for details.
Blockbuster uses the forbiddenfruit library to monkey-patch CPython immutable builtin functions and methods.
Blockbuster was greatly inspired by the BlockHound library for Java, which serves a similar purpose of detecting blocking calls in JVM reactive applications.