An assertion is an expression that must return a Boolean (true if the
assertion succeeds, false otherwise).
It always relies on metric values, operators, units,
functions, and variables and must be written in the tests
section of the .blackfire.yaml file:
'All pages are fast':
- 'main.wall_time < 50ms'
- 'main.peak_memory < 10mb'
- 'main.network_out < 10kb'
'Not too many SQL queries on the homepage':
- 'metrics.sql.queries.count < 5'
In the above example, main.network_out, metrics.sql.queries.count, and
main.peak_memory are metric values. Unit like in 10mb can be explicit
in order to override default units.
Whenever possible, we recommend you write assertions that do not
depend on time. The main reason is that time is always a consequence, a symptom of a deeper issue.
Assertions support profile comparison as well
to assert the performance evolution of your code:
"Pages should not become slower":
- "percent(main.wall_time) < 10%" # time does not increase by more than 10%
- "diff(metrics.sql.queries.count) < 2" # less than 2 additional SQL statements
Comparison assertions are only evaluated when running builds.
Values are compared between the current build and a reference build.
The reference can be either the last successful periodic build, or a build that has been referenced in
the webhook command with
Blackfire exposes metrics that are associated with the current
profile. The value of one dimension of a metric can be used in assertions;
for instance, the count value of the sql.queries metric is stored in
Learn more about all built-in metrics exposed
in assertions by default and how to create your own.
The available dimensions for metrics are the following ones:
For each metric (count, time or memory), a default unit is defined:
When using 10 in an assertion for a time value, that evaluates to 10
milliseconds. It is possible to add an explicit unit to a metric, which will
override the default unit.
The following time units are supported: ms, s.
The following memory units are supported: kb, kib, mb, mib,
The following generic units are supported: k, ki, m,
mi, g, gi.
The following operators are supported in assertions:
Variables are useful when you configure several Blackfire environments to run performance tests on various machines
hosting the same application. Variables make it possible to use the same
.blackfire.yaml configuration on several servers that
have different purposes (e.g. development, staging, production).
Variables can be defined in the environment configuration and used in an
expression by passing the metric key to the var() function.
Variable values may also use units.
Example: You have configured two different environments in Blackfire:
The same application runs on all servers, but the debug mode increases memory
usage a lot. As such, you cannot set the same maximum value for your
assertions on memory.
Your .blackfire.yaml would look like this when using variables:
"Pages shouldn't use too much memory":
- "main.peak_memory < 10mb * var('memory_coeff')"
Now in both your Integration and Production Blackfire environments,
you have to create the memory_coeff variable:
memory_coeff == 2
memory_coeff == 1
The var() function may also receive a default value as a second
This default value is used whenever the variable is not defined in one of your
"Pages shouldn't use too much memory":
- "main.peak_memory < 10mb * var('memory_coeff', 1)"
The vars.xxx notation, supported in older versions of Blackfire, is now
deprecated. We strongly advise you to migrate your expression to use
the var() function instead.
When an assertion is run from an environment, the is_dev() function returns
false when the environment is configured for production usage.
When using builds, it is possible
to compare one build to another. This is useful when you want to validate a
code merge (e.g. a pull-request) by triggering a build webhook or a Blackfire Player collection of scenarios. It can also ensure that your code doesn’t
have performance regression over time using periodic builds.
Comparison can be made within assertion expressions thanks to the percent()
and diff() functions:
If you want to assert that a metric value does not increase by more than a
given percentage between two builds, use the percent() function:
"percent(main.wall_time) < 10%"
You can also test the evolution of a metric value in absolute terms with the
"diff(metrics.sql.queries.count) < 2"
In this example, the assertion checks that the profile from the newer build has
less than 2 additional SQL statements compared to the previous one.