Overview

The module pyTooling.Testing offers enhanced classes for writing unit tests with Python’s unittest framework, which is also what pytest runs.

Testcase

Testcase is the class to derive a testcase from. It is a unittest.TestCase - everything that class offers is available unchanged - and adds what pyTooling’s test suites otherwise write themselves.

from pyTooling.Testing import Testcase


class Slots(Testcase):
  def test_SlotsAreDerived(self) -> None:
    self.assertHasAttr(MyClass, "__slots__")

Deriving from it rather than from unittest.TestCase costs nothing and means a test suite picks up what is added later without changing every class again.

Assertions

assertHasAttr() and assertNotHasAttr() check whether an object has an attribute. They were added to unittest.TestCase in Python 3.14, so a test suite running on 3.11 to 3.13 cannot use them - Testcase provides them there.

On Python 3.14 and newer the class defines nothing of its own, so the standard library’s implementations and messages are used and the two behave identically on every supported interpreter.

Application Testing

A unit test imports the code it tests. An application test starts the installed program the way a user does, so the chain under test includes what importing cannot reach: the console_scripts entry point, the argument parsing and the exit code.

ApplicationTestcase derives from Testcase. It resolves the console script once per test class and offers two ways to start the program:

RunEntrypoint() runs the installed console script - the path a user takes, and therefore the one covering the entry-point wiring.
RunModule() runs python -m <module>. When this passes while the entry point fails, the packaging is at fault rather than the code.

Both capture stdout and stderr as text and take a timeout, so a hanging program fails the test instead of the test suite.

from pyTooling.Testing import ApplicationTestcase


class Commands(ApplicationTestcase):
  _consoleScript =  "myprogram"
  _runnableModule = "myPackage.CLI"

  def test_Version(self) -> None:
    result = self.RunEntrypoint("--version")

    self.assertExitCode(result)
    self.assertIn("myprogram", result.stdout)

Both class variables are mandatory: a test class naming neither cannot run anything, so setUpClass() raises a TestingError instead of letting every testcase in the class fail with a less obvious error. A console script that is not installed is reported the same way.

Assertions

assertExitCode() compares the exit code and, when they differ, reports the command line together with what the program printed. That output is what explains the failure, and it is gone once the test has finished, so it belongs in the assertion message rather than in the console.

Helpers

stripANSIColorCodes() removes ANSI escape sequences from a text. A program writing to a terminal colors its output while the same program in a pipe usually does not - comparing the stripped text is more robust than encoding a rule about when the codes appear.

Marker-based Collection

A test runner has to decide what a test is, and by default it decides from a name: pytest collects classes matching python_classes (Test*) and functions matching python_functions (test_*), and unittest’s loader collects methods starting with test. The identifier therefore does two jobs at once - it names the entity and it enables collection.

@testsuite and @testcase separate them.

from pyTooling.Testing import Testcase, testsuite, testcase

@testsuite("Version comparison")
class VersionComparison(Testcase):
  """
  This is a testsuite summary.

  Here follows a multiline
  testsuite description.
  """

  @testcase("A newer version compares greater")
  def NewerIsGreater(self) -> None:
    """
    This is a testcase summary.

    This can describe a testcase with more details
    using multiple lines.
    """
    self.assertGreater(Version("2.0"), Version("1.9"))

The class is collected because it is marked, not because of how it is spelled, and the title travels into the report as a property - see The four names of a test item.

Important

classname and name keep the identifiers and are not replaced by the titles, for two reasons.

They are the testcase’s node ID, which is what selects a test: on the command line, from an IDE’s run this test, and from the cache --last-failed reads. And a post-processing tool may reasonably expect them to be identifiers - free of spaces and punctuation - so a title in that position could break it.

A title is additional information, so it is reported as additional information.

The four names of a test item

A test item - a test suite or a testcase - has four names, and only the first is what Python calls it:

Name	Where it comes from
ID	the module, class or method name. It is the item’s `classname`/`name`, and what selects the test.
title	what the marker was given. Defaults to the ID.
summary	the first paragraph of the doc-string.
description	the doc-string.

They are four values, not a fallback chain: a title never replaces a summary, and a summary never becomes a title. A testcase can therefore carry a short label and a sentence and the full prose, and a report can show whichever of them it has room for.

@testsuite("Version comparison.")
class VersionComparison(Testcase):
  """
  Compare two release versions.

  Everything about comparing them.
  """

  @testcase("A newer version compares greater.")
  def NewerIsGreater(self) -> None:
    """
    A newer version compares greater than an older one.

    Only the minor number differs here.
    """

All of them except the ID reach the report as properties:

<testsuites name="pytest tests">
  <testsuite name="pytest" errors="0" failures="0" skipped="0"
             tests="1" time="0.016" timestamp="2026-08-24T23:55:41+00:00" hostname="build-01">
    <testcase classname="tests.unit.Versioning.Comparison.VersionComparison"
              name="test_NewerIsGreater" time="0.001">
      <properties>
        <property name="title" value="A newer version compares greater." />
        <property name="summary" value="A newer version compares greater than an older one." />
        <property name="description"
                  value="A newer version compares greater than an older one.&#10;&#10;Only the minor ..." />
        <property name="testsuiteTitle" value="Version comparison." />
        <property name="testsuiteSummary" value="Compare two release versions." />
        <property name="testsuiteDescription"
                  value="Compare two release versions.&#10;&#10;Everything about comparing them." />
      </properties>
    </testcase>
  </testsuite>
</testsuites>

classname is the testcase’s package path - the directories below the root, then the module, then the class - so a testcase in tests/unit/Versioning/Comparison.py is reported as tests.unit.Versioning.Comparison.VersionComparison. Its ID, in the table above, is the last part of that.

Hint

Because the title defaults to the ID, a marker can be added to an existing testcase without changing anything a report says about it - which is what makes a suite migratable one class at a time.

The names of a test suite level

A JUnit document has exactly one <testsuite> element for the whole session, and squeezes the hierarchy into a dotted classname - so a level between the root and the class has no element that could carry a title or a description.

The names of every level are therefore written as keys in the session’s <properties>, where the key is the level’s dotted path:

<?xml version="1.0" encoding="utf-8"?>
<testsuites name="pytest tests">
  <testsuite name="pytest" errors="0" failures="0" skipped="0" tests="1" time="0.011"
             timestamp="2026-08-25T07:01:50.230489+00:00" hostname="build-01">
    <properties>
      <property name="versioning.summary" value="The version handling test suite." />
      <property name="versioning.description"
                value="The version handling test suite.&#10;&#10;Everything about parsing ..." />
      <property name="versioning.test_comparison.summary" value="Version comparison tests." />
      <property name="versioning.test_comparison.description" value="Version comparison tests." />
      <property name="versioning.test_comparison.VersionComparison.title" value="Version comparison." />
    </properties>
    <testcase classname="versioning.test_comparison.VersionComparison" name="test_NewerIsGreater" time="0.004">
      <properties>
        <property name="title" value="A newer version compares greater." />
      </properties>
    </testcase>
  </testsuite>
</testsuites>

The innermost key is the testcase’s classname, and every outer one is a prefix of it - which is what lets a reader join the two.

Where a level’s names come from depends on what the level is:

a package or a module has only its doc-string, so its summary and full text become the summary and the description - the doc-string in __init__.py names the package;
a class carries a @testsuite marker, so it has a title as well.

A level contributes only the names it has, and a level with none is skipped.

Important

They are written once per session, not once per testcase. A <property> inside <testcase> would repeat for every testcase in that level, which is why the testcase carries only its own names.

Note

The PyTest-JUnit schema does allow <properties> on a <testsuite> element - it is pytest emitting a single one that leaves nowhere to put a per-class name. A Report Format of One’s Own is the format where the hierarchy is expressed directly instead of being encoded in a key.

Enabling the plugin

The collection itself is a pytest plugin, pyTooling.Testing.PyTest. pyTooling declares it as a pytest11 entry point, so an installed pyTooling registers it automatically and a test suite only has to mark something.

The plugin is inert until something is marked, so its presence changes nothing for a test suite that collects by name. Both styles work in one session and even in one file, which is what makes a gradual migration possible.

Two cases still name it explicitly:

pytest -p pyTooling.Testing.PyTest tests/unit      # a checkout that is not installed
pytest -p no:pyTooling.Testing.PyTest tests/unit   # switch the plugin off

The entry point’s name is the module’s name, so both spellings address the same plugin, and passing -p for an already registered plugin does nothing rather than registering it twice.

What the plugin does

pytest_pycollect_makeitem() turns a marked class into a collector and a marked method into a test item, so neither has to match python_classes or python_functions.
pytest_collection_modifyitems() attaches the titles to the item as user_properties - the channel the record_property() fixture uses. They are part of the test report, so they survive a pytest-xdist worker and reach the JUnit report as <property> elements.
Node IDs are never touched, so selection, pytest-xdist, --last-failed and IDE integration work exactly as they do without the plugin.
An unmarked method in a marked class is not collected. Marking is the whole statement of intent, so a helper method needs no naming convention to stay out of the report.
A marked unittest.TestCase is a special case. Such a class is collected by pytest’s unittest support, which asks unittest.TestLoader.getTestCaseNames() for the test methods - and that loader matches testMethodPrefix, which is "test". It is not the python_functions setting: with python_functions = check_*, a plain class collects check_* methods while a TestCase still collects test_* ones. The plugin therefore aliases each marked method under a name that loader accepts and lets pytest collect the class as usual.

A Report Format of One’s Own

JUnit XML cannot express two things a marked test suite has.

Test suites do not nest. A JUnit document holds one flat list of <testcase> elements, and the hierarchy is squeezed into a dotted classname - tests.unit.Versioning.VersionComparison. Every level between the root and the class is a substring, so nothing can be said about a level: it has no element to carry a title or a description.

An item has one name. The four names of a test item gives it four, and JUnit’s only place for the other three is a flat <property name= value=> pair, whose value is an attribute and therefore a single line.

pyTooling.Testing.ReportWriter writes a format that has both. It is a pytest11 entry point as well, so it needs no registration either - only the switch that turns it on:

pytest --pytooling-xml=report/unit/TestReport.xml --junit-xml=report/unit/unittest.xml

Both files are written in one session from the same reports, so a pipeline keeps the format its dashboard understands while the richer file is produced beside it.

<?xml version='1.0' encoding='utf-8'?>
<TestReport xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
            xsi:noNamespaceSchemaLocation="TestReport-v0.1.xsd"
            timestamp="2026-08-25T18:15:26.819107+00:00"
            duration="0.000671" tests="2" failures="0" errors="0" skipped="0">
  <Testsuite name="test_versioning">
    <Testsuite name="VersionComparison">
      <Title>Version comparison.</Title>
      <Summary>Compare two release versions.</Summary>
      <Description>Compare two release versions.

Everything about comparing them.</Description>
      <Testcase name="test_NewerIsGreater" status="passed" duration="0.000428"
                nodeID="test_versioning.py::VersionComparison::test_NewerIsGreater">
        <Title>A newer version compares greater.</Title>
        <Summary>A newer version compares greater than an older one.</Summary>
        <Description>A newer version compares greater than an older one.

Only the minor number differs here.</Description>
      </Testcase>
    </Testsuite>
  </Testsuite>
</TestReport>

The schema

The schema lives in the resource package pyTooling.Resources and is shipped with the distribution. Every generated file points at it with xsi:noNamespaceSchemaLocation, so a reader can validate without being told where it lives. getResourceFile() returns its path, whether pyTooling is installed, inside a wheel, or a checkout:

from pathlib                        import Path
from xmlschema                      import XMLSchema
from pyTooling                      import Resources
from pyTooling.Common               import getResourceFile
from pyTooling.Testing.ReportWriter import SCHEMA_FILES, SCHEMA_VERSION_LATEST

schemaPath: Path = getResourceFile(Resources, SCHEMA_FILES[SCHEMA_VERSION_LATEST])
XMLSchema(schemaPath).validate("report/unit/TestReport.xml")

Validating needs an XML schema library such as xmlschema. pyTooling does not depend on one: writing a report uses xml.etree.ElementTree from the standard library, so the schema is there for whoever reads the file.

The file name carries the format’s version. SCHEMA_FILES maps a version to its schema file, so a later version of the format is added beside the current one rather than replacing it, and a reader learns from a report’s xsi:noNamespaceSchemaLocation which version it is holding. The format states its own version this way; the report does not name the tool that wrote it, nor the machine it ran on.

name is an attribute on every item, because it is an identifier. Title, Summary and Description are elements, because they are prose - Description is typed preservingstring, so its line breaks survive.
<Testsuite> is recursive, so the hierarchy is as deep as the test suite is.
<Testcase> carries status from a fixed list, duration, and a nodeID - the test runner’s own identifier, so a reader of the report can re-run exactly that testcase.
An item writes only the names it has, so an unmarked testcase produces a <Testcase> element with no children.

Where the nesting comes from

The levels are the node ID’s own parts: the module path, then each class between it and the testcase. So tests/unit/Versioning.py::VersionComparison::test_NewerIsGreater becomes tests → unit → Versioning → VersionComparison, and a title or description attaches to whichever level declared one.