Testing

Tests enable you to quickly validate data across your project.

Note

Tests are run with the visivo test command (docs).

Tests can run any arbitrary python statement so long as it evaluates to True or False. Additionally tests can access trace data from across the project by using the $(project) context variable. By default trace data arrays will be represented as numpy arrays.

The combination of python based boolean statements and access to all trace data makes it super fast all while enabling complex assertions about your data.

Example

Revenue Sums MatchStandard DeviationAssert Value

tests:
  - name: revenue-date-grains-match
    logic: ">{ numpy.sum( ${project.traces['revenue-per-week'].props.y} ) = numpy.sum( ${project.traces['revenue-per-month'].props.y} ) }"

tests:
  - name: recent-std-less-double-normal
    logic: ">{ numpy.std( ${project.traces['revenue-per-week'].props.y} ) * 2 > numpy.std( ${project.traces['revenue-per-month'].props.y[:-10]} ) }"

tests:
  - name: recent-std-less-double-normal
    logic: ">{ round( ${project.traces['revenue-per-week'].props.y[10]} ) = 2901384 }"

You can also define tests within the trace it's self which gives you access to the information of the trace data that you're defining the test on through the ${trace} context variable. You can still access the ${project} context variable from within the trace definition.

Example

traces:
    - name: tested-trace
    model: ref(model)
    columns:
        account_name: account_name
    props:
        type: scatter
        x: column(project_created_at)
        y: column(project_name)
    tests:
        - logic: ">{ assert_that(numpy.sum( ${trace.props.x} ).is_equal_to(7) }"
        - logic: ">{ 'key account' in ${trace.columns.account_name} }"
        - logic: ">{ numpy.unique( ${trace.columns.account_name} ) = numpy.unique( ${project.traces[another-trace].columns.account_name} ) }

The assert_that() function from the assertpy library and the numpy library are available to reference. The combination of these two libraries enable a wide array of calculations and logical assertions.