Log Assertion

What?

A simple log assertion mechanism for Python unittests.

Why?

As is vox populi, you must also test the logging calls in your programs.

With logassert this is now very easy.

Why is it easy?

Because it provides a simple and expressive way to use it in the unit tests (see next section) but also because when the assertion fails, it presents a useful report that helps you to find out why it is failing.

For example, this may be a case when the logged line is slightly different:

Actual code:

    value = "123"
    logger.debug("Received value is %r", value)

The unit test assertion:

    assert "Received value is 123" in logs.debug

The report you get as it failed:

    AssertionError: assert for Regex("Received value is 123") in DEBUG failed; logged lines:
           DEBUG     "Received value is '123'"
    )

Or a case where the logged line is ok, but the level is incorrect:

Actual code:

    value = "123"
    logger.debug("Received value is %s", value)

The unit test assertion:

    assert "Received value is \d+" in logs.info

The report you get as it failed:

    AssertionError: assert for Regex("Received value is \d+") in DEBUG failed; logged lines:
           INFO      "Received value is 123"
    )

Awesome! How do I use logassert?

After installing, the same functionality is exposed in two very different ways, one that fits better the pytest semantics, the other one more suitable for classic unit tests.

For pytest

All you need to do is to declare logs in your test arguments, it works just like any other fixture.

Then you just check (using assert, as usual with pytest) if a specific line is in the logs for a specific level.

Example:

def test_bleh(logs)
    (...)
    assert "The meaning of life is 42!" in logs.debug

Actually, the line you write is a regular expression, so you can totally do (in case you're not exactly sure which the meaning of life is):

    assert "The meaning of life is \d+!" in logs.debug

The indicated string is searched to be inside the log lines, it doesn't need to be exact whole line. If you want that, just indicate it as with any regular expression:

    assert "^The meaning of life is \d+!$" in logs.debug

In a similar way you can also express the desire to check if it's at the beginning or at the end of the log lines.

NOTE: the message checked is the formatted one, after the logging system replaced all the parameters in the template and built the final string. In other words, if the code is logger.debug("My %s", "life"), the verification will be done in the final My life string.

If you want to verify that a text was logged, no matter at which level, just do:

    assert "The meaning of life is 42" in logs.any_level

To verify that some text was NOT logged, just juse the Python's syntax! For example:

    assert "A problem happened" not in logs.error

But I don't like regexes, I want the exact string

Then you just import Exact from logassert and wrap the string with that.

For example, in this case the .. means exactly two dots, no regex semantics at all:

    assert Exact("The meaning of life is ..") in logs.any_level

Anyway, I liked old behaviour of searching multiple strings

Then you may want to import Multiple from logassert and wrap the different strings you had in each call for the classic behaviour.

For example:

    assert Multiple("life", "meaning", "42") in logs.any_level

What if I want to check that nothing was logged?

The simplest way to do it is to use the NOTHING verifier that you can import from logassert:

    assert NOTHING in logs.debug

Note that it doesn't make sense to use it by the negative (...NOTHING not in logs...): is no really useful at testing level to know that "something was logged", you should improve the test to specifically verify what was logged.

Breaking the "per line barrier"

Sometimes it's useful to verify that several lines were logged, and that those lines are logged one after the other, as they build a "composite message".

To achieve that control on the logged lines you can use the Sequence helper, that receives all the lines to verify (regexes by default, but you can use the other helpers there):

    assert Sequence(
        "Got 2 errors and \d+ warnings:",
        Exact("  error 1: foo"),
        Exact("  error 2: bar"),
    ) in logs.debug

Examples

After logging...

    person = "madam"
    item = "wallet"
    logger.debug("Excuse me %s, you dropped your %s", person, item)

...the following test will just pass:

    assert "Excuse me .*?, you dropped your wallet" in logs.debug

However, the following will fail (different text!)...

    assert "Excuse me .*?, you lost your wallet" in logs.debug

...producing this message in your tests:

assert for regex 'Excuse me .*?, you lost your wallet' check in DEBUG, failed; logged lines:
      DEBUG     'Excuse me madam, you dropped your wallet'

This one will also fail (different level!)...

    assert "Excuse me .*?, you dropped your wallet" in logs.info

...producing this message in your tests:

assert for regex 'Excuse me .*?, you dropped your wallet' check in INFO, failed; logged lines:
       DEBUG     'Excuse me madam, you dropped your wallet'

A more complex example, with several log lines, and a specific assertion:

    logger.info("Starting system")
    places = ['/tmp/', '~/temp']
    logger.debug("Checking for config XYZ in all these places %s", places)
    logger.warning("bad config XYZ")

    assert "bad config XYZ" in logs.debug

See how the test failure message is super helpful:

assert for regex 'bad config XYZ' check in DEBUG, failed; logged lines:
       INFO      'Starting system'
       DEBUG     "Checking for config XYZ in all these places ['/tmp/', '~/temp']"
       WARNING   'bad config XYZ'

What about repeated verifications?

Sometimes it's needed to verify that something if logged only once (e.g. welcoming messages). In this cases it's super useful to use the reset method.

See the following test sequence:

def test_welcoming message(logs):
    custom_logger.info("foo")  # first log! it should trigger the welcoming message
    assert "Welcome" in logs.info

    logs.reset()
    custom_logger.info("foo")  # second log! it should NOT trigger the welcoming message
    assert "Welcome" not in logs.info

For classic TestCases

All you need to do is to call this module's setup() passing the test case instance, and the logger you want to supervise.

Like

class MyTestCase(unittest.TestCase):
    """Example."""

    def setUp(self):
        logassert.setup(self, 'mylogger')

In the example, mylogger is the name of the logging to supervise. If different subsystems of your code log in other loggers, this tester won't notice.

Then, to use it, just call the assertLogged method and it's family, passing all the strings you want to find. This is the default behaviour for backwards compatibility.

Example:

    def test_blah(self):
        (...)
        self.assertLoggedDebug('secret', 'life', '42')

That line will check that "secret", "life" and "42" are all logged in the same logging call, in DEBUG level.

So, if you logged this, the test will pass:

logger.debug("The secret of life, the universe and everything is %d", 42)

Note that the message checked is the one with all parameters replaced.

But if you logged any of the following, the test will fail (the first because it misses one of the string, the second because it has the wrong log level)::

logger.debug("The secret of life, the universe and everything is lost")
logger.info("The secret of life, the universe and everything is 42")

What can I test?

You'll have at disposition several assertion methods:

• self.assertLogged: will check that the strings were logged, no matter at which level

• self.assertLoggedLEVEL (being LEVEL one of Error, Warning, Info, or Debug): will check that the strings were logged at that specific level.

• self.assertNotLogged: will check that the strings were NOT logged, no matter at which level

• self.assertNotLoggedLEVEL (being LEVEL one of Error, Warning, Info, or Debug): will check that the strings were NOT logged at that specific level.

Support for structured logs

The structlog library is very commonly used by developers. It provides a simple way of logging using messages and dictionaries with structured data that later are processed in powerful ways.

For example you can do:

    ...
    result = "success"
    code = 37
    logger.debug("Process finished correctly", result=result, code=code)

How do you test that? Don't panic! logassert supports structlog :)

It is very similar to the regular logging checks, but formalizing that there is a structure with a message and other fields:

    assert Struct("Process finished", result="success") in logs.debug

When a string is used in the main message or any of the field values, the regular logassert rules apply (by default it is a regular expression and is searched in the logged text) but you can use all the power of the helpers, like checking for the exact string...

    assert Struct(Exact("Process finished correctly"), result="success") in logs.debug

...or using multiple strings...

    assert Struct(Multiple("correctly", "finished"), result="success") in logs.debug

... etc.

If the field value is not a string, it's matches just for equality:

    assert Struct("finished", code=37) in logs.debug
    assert Struct("finished", code=3) not in logs.debug

Complete structures

The previous examples just verified that the indicated fields exist in the logged lines, but they do NOT assert that those are ALL the logged fields.

If you want to check that the given message and fields match but also verify that the those are all the logged fields, you need to use CompleteStruct. E.g.:

    assert CompleteStruct("finished", code=37, result="success") in logs.debug

How to install

logassert is a very small pure Python library, easiest way to install is from PyPI:

    pip install logassert

Nice! But...

If you need help, or have any question, or found any issue, please open a ticket here.

Thanks in advance for your time.