Overview

PBBT -- Pluggable Black-Box Testing toolkit

Overview

PBBT is a regression test harness for black-box testing. It is suitable for testing complex software components with well-defined input and output interfaces.

  input    +----------+   output
o--------> | Software | --------->o
           +----------+

In black-box testing, a test case is a combination of input and expected output data. The test harness executes the software with the given input and verifies that the produced output coincides with the expected output.

Black-box testing could be implemented for many different types of software. For example,

  • a database system: the input is a SQL statement, the output is a set of records;
  • a web service: the input is an HTTP request, the output is an HTTP response;
  • a command-line utility: the input is a sequence of command-line parameters and stdin, the output is stdout;
  • a GUI application: different approaches are possible; for instance, the input could be a sequence of user actions, and the output could be the state of a particular widget.

PBBT is a Python library and an application which allows you to:

  • use built-in test types for testing command-line scripts and Python code;
  • register custom test types;
  • prepare test input in a succinct YAML format;
  • in the train mode, run the test cases and record expected output;
  • in the check mode, run the test cases and verify that the produced output coincides with the pre-recorded expected output.

PBBT is a free software released under MIT license. PBBT is created by Clark C. Evans and Kirill Simonov from Prometheus Research, LLC.

Using PBBT

To install PBBT, you can use pip package manager:

# pip install pbbt

This command downloads and installs the latest version of PBBT from Python Package Index. After successful installation, you should be able to import pbbt Python module and run pbbt command-line utility.

To start using PBBT, you need to create a file with input data. For example, create input.yaml with the following content:

py: |
  print "Hello, World!"

This file is in YAML format, which is a data serialization language similar to JSON, and, in fact, a superset of JSON. The file above could be represented in JSON as:

{ "py": "print \"Hello, World!\"\n" }

For description of YAML syntax and semantics, see http://yaml.org/.

Next, execute PBBT in training mode to generate expected output data. Run:

$ pbbt input.yaml output.yaml --train

and accept new output when asked. PBBT will write output data to output.yaml:

py: print-hello-world
stdout: |
  Hello, World!

Now you can start PBBT in checking mode, in which it executes test cases and verifies that expected and actual output data coincide:

$ pbbt input.yaml output.yaml

To add more test cases to input.yaml, you need to convert it to a test suite:

title: My Tests
tests:
- py: |
    print "Hello, World!"
- sh: echo Hello, World!

The file now contains a test suite My Tests with two test cases: one as in the previous example, and another that executes a shell command echo Hello, World!:

sh: echo Hello, World!

The output of this test case is stdout produced by the shell command. To record expected output, run pbbt in training mode again.

Built-in Test Types

Out of the box, PBBT supports a small set of predefined test types:

  • test Python code;
  • test a shell command;
  • file manipulation tests.

Also available are special test types:

  • test suite;
  • include;
  • conditional variables.
  • gateway to other test systems.

Each test type defines the structure of input and output records, that is, the set of mandatory and optional fields and the type of field values. In this section, we list all available test types and describe their input fields.

Common Fields

The following optional fields are available for all test types where they make sense:

skip: true or false
On true, skip this test case.
if: variable, list of variables or Python expression

On a variable name, run this test case only if the variable is defined and not false.

On a list of variables, run this test case only if at least one variable is defined and not false.

On a Python expression, run this test case if the expression evaluates to true. You can use any conditional variables in the expression.

unless: variable, list of variables or Python expression

On a variable name, skip this test case if the variable is defined and not false.

On a list of variables, skip this test case if at least one variable is defined and not false.

On a Python expression, skip this test case if the expression evaluates to true. You can use any conditional variables in the expression.

ignore: true, false or regular expression

On true, permit the actual and expected output to be unequal. The test case must still execute without any errors.

On a regular expression, pre-process the actual and expected output before comparing them:

  1. Run the regular expression against the output data and find all matches.
  2. If the regular expression does not contain () subgroups, erase all the matches from the output.
  3. If the regular expression contains one or more () subgroups, erase the content of the subgroups from the output.

The regular expression is compiled with MULTILINE and VERBOSE flags.

Example:

title: Integration with MySQL
if: has_mysql
tests:
- set:
    MYSQL_HOST: localhost
    MYSQL_PORT: 3306
  unless: [MYSQL_HOST, MYSQL_PORT]
- read: /etc/mysql/my.cnf
  if: MYSQL_HOST == 'localhost'
- py: test-scalar-types.py
  ignore: |
    ^Today:.(\d\d\d\d)-(\d\d)-(\d\d)$
- py: test-array-types.py
  skip: true    # No array type in MySQL

Test Suite

A test suite is a collection of test cases.

A suite may contain other suites and thus all test suites form a tree structure. A path formed from suite identifiers can uniquely locate any suite. We use file-system notation for suite paths (e.g. /path/to/suite).

Fields:

title: text
The title of the suite.
suite: identifier (optional)
The identifier of the suite. If not set, generated from the title.
tests: list of input records
The content of the suite.
output: path (optional)
If set, the expected output of the suite is loaded from the given file.

Example:

title: All Tests
suite: all
output: output.yaml
tests:
- py: ./test/core.py
- py: ./test/ext.py
- title: Database Tests
  tests:
  - py: ./test/sqlite.py
  - py: ./test/pgsql.py
  - py: ./test/mysql.py

In this example, the path to the Database Tests suite is /all/database-tests.

Conditional Variables

This test case defines a conditional variable.

Variables could be used in if and unless clauses to conditionally enable or disable a test case. Variables could also be set or read in Python tests via a global dictionary __pbbt__.

Conditional variables could also be set from command line using -D option.

Setting a conditional variable affects all subsequent test cases within the same test suite. Variable values are reset on exit from the suite.

Fields:

set: variable or dictionary of variables

On a variable name, set the value of the given variable to True.

On a dictionary, set the values of the given variables.

Example:

title: MySQL Tests
tests:
- set: MYSQL
- set:
    MYSQL_HOST: localhost
    MYSQL_PORT: 3306
  unless: [MYSQL_HOST, MYSQL_PORT]
- py: |
    # Determine the version of the MySQL server
    import MySQLdb
    connection = MySQLdb.connect(host=__pbbt__['MYSQL_HOST'],
                                 port=int(__pbbt__['MYSQL_PORT']),
                                 db='mysql')
    cursor = connection.cursor()
    cursor.execute("""SELECT VERSION()""")
    version_string = cursor.fetchone()[0]
    version = tuple(map(int, version_string.split('-')[0].split('.')))
    __pbbt__['MYSQL_VERSION'] = version
- py: test-ddl.py
- py: test-dml.py
- py: test-select.py
- py: test-new-features.py
  if: MYSQL_VERSION >= (5, 5)

Include Test

This test case loads and executes a test case from a file.

Fields:

include: path
The file to load. The file should contain an input test record in YAML format.

Example:

title: All Tests
tests:
- include: test/core.yaml
- include: test/ext.yaml
- include: test/sqlite.yaml
- include: test/pgsql.yaml
- include: test/mysql.yaml

Python Code

This test case executes Python code and produces stdout.

Fields:

py: path or Python code

On Python code, the source code to execute.

On a file name, the file which contains source code to execute.

stdin: text (optional)
Content of the standard input.
except: exception type (optional)
If set, indicates that the code is expected to raise an exception of the given type.

Example:

title: Python tests
tests:
- py: hello.py
- py: &sum |
    # Sum of two numbers
    import sys
    a = int(sys.stdin.readline())
    b = int(sys.stdin.readline())
    c = a+b
    sys.stdout.write("%s\n" % c)
  stdin: |
    2
    2
- py: *sum
  stdin: |
    1
    -5
- py: *sum
  stdin: |
    one
    three
  except: ValueError

Note that we use a YAML anchor (denoted by &sum) and aliases (denoted by *sum) to use the same piece of code in several tests.

Shell Command

This test case executes a shell command and produces stdout.

Fields:

sh: command or executable with a list of parameters
The shell command to execute.
stdin: text (optional)
Content of the standard input.
cd: path (optional)
Change the current working directory to the given path before executing the command.
environ: dictionary of variables (optional)
Add the given variables to the command environment.
exit: integer (optional)
The expected exit code; 0 by default.

Example:

title: Shell tests
tests:
- sh: echo Hello, World!
- sh: cat
  stdin: |
    Hello, World!
- sh: [cat, /etc/shadow]
  exit: 1   # Permission denied

Write to File

This test case creates a file with the given content.

Fields:

write: path
The file to create.
data: text
The file content.

Example:

write: test/tmp/data.txt
data: |
    Hello, World!

Read from File

The output of this test is the content of a file.

Fields:

read: path
The file to read.

Example:

read: test/tmp/data.txt

Remove File

This test case removes a file. It is not an error if the file does not exist.

Fields:

rm: path or list of paths
File(s) to remove.

Example:

rm: test/tmp/data.txt

Make Directory

This test case creates a directory.

Parent directories are also created if necessary. It is not an error if the directory already exists.

Fields:

mkdir: path
The directory to create.

Example:

mkdir: test/tmp

Remove Directory

This test case removes a directory with all its content.

It is not an error if the directory does not exist.

Fields:

rmdir: path
The directory to delete.

Example:

rmdir: test/tmp

Doctest

This test case executes doctest on a set of files.

Fields:

doctest: path pattern
Files with doctest sessions.

Example:

doctest: test/test_*.rst

Unittest

This test case executes unittest test suite.

Fields:

unittest: path pattern
Files with unittest tests.

Example:

unittest: test/test_*.py

Pytest

This test case executes py.test test suite.

Package pytest from http://pytest.org/ must be installed.

Fields:

pytest: path pattern
Files with py.test tests.

Example:

pytest: test/test_*.py

Coverage

This test case starts coverage of Python code.

Package coverage from http://nedbatchelder.com/code/coverage/ must be installed.

Fields:

coverage: file name or None
Path to the configuration file.
data_file: file name
Where to save coverage data.
timid: false or true (optional)
Use a simpler trace function.
branch: false or true (optional)
Enable branch coverage.
source: file paths of package names (optional)
Source files or packages to measure.
include: file patterns
Files to measure.
omit: file patterns
Files to omit.

Example:

coverage:
source: pbbt
branch: true

Coverage check

This test case stops coverage and reports measurement summary.

Fields:

coverage-check: float
Expected coverage percentage.

Example:

coverage-check: 99.0

Coverage report

This test case stops coverage and saves coverage report to a file.

Fields:

coverage-report: directory
Where to save the report.

Example:

coverage-report: coverage

Custom Test Types

In this section, we discuss how to add custom test types to PBBT.

Suppose we want to test a SQL database by running a series of SQL queries and validating that we get expected output. To implement this testing scheme, we need a way to:

  • open a connection to the database;
  • execute a SQL statement and produce a sequence of records.

The input file may look like this:

title: Database tests
tests:
# Remove the database file left after the last testing session.
- rm: test.sqlite
# Create a new SQLite database.
- connect: test.sqlite
# Run a series of SQL commands.
- sql: |
    SELECT 'Hello, World!';
- sql: |
    CREATE TABLE student (
        id      INTEGER PRIMARY KEY,
        name    TEXT NOT NULL,
        gender  CHAR(1) NOT NULL
                CHECK (gender in ('f', 'i', 'm')),
        dob     DATE NOT NULL
    );
- sql: |
    INSERT INTO student (id, name, gender, dob)
    VALUES (1001, 'Linda Wright', 'f', '1988-10-03'),
           (1002, 'Beth Thompson', 'f', '1988-01-24'),
           (1003, 'Mark Melton', 'm', '1984-06-05'),
           (1004, 'Judy Flynn', 'f', '1986-09-02'),
           (1005, 'Jonathan Bouchard', 'm', '1982-02-12');
- sql: |
    SELECT *
    FROM student
    ORDER BY dob;
- sql: |
    SELECT name
    FROM student
    WHERE id = 1003;

In this input file, we use two new types of test cases:

connect
Specifies the connection to a SQLite database.
sql
Specifies a SQL statement to execute.

We will write a PBBT extension implementing these test types.

Create a file sql.py with the following content:

from pbbt import Test, Field, BaseCase, MatchCase, listof
import sqlite3, traceback, csv, StringIO

@Test
class ConnectCase(BaseCase):

    class Input:
        connect = Field(str)

    def check(self):
        self.state['connect'] = None
        try:
            self.state['connect'] = sqlite3.connect(self.input.connect)
        except:
            self.ui.literal(traceback.format_exc())
            self.ui.warning("exception occurred while"
                            " connecting to the database")
            self.ctl.failed()
        else:
            self.ctl.passed()

@Test
class SQLCase(MatchCase):

    class Input:
        sql = Field(str)

    class Output:
        sql = Field(str)
        rows = Field(listof(listof(object)))

    def render(self, output):
        stream = StringIO.StringIO()
        writer = csv.writer(stream, lineterminator='\n')
        writer.writerows(output.rows)
        return stream.getvalue()

    def run(self):
        connection = self.state.get('connect')
        if not connection:
            self.ui.warning("database connection is not defined")
            return
        rows = []
        try:
            cursor = connection.cursor()
            cursor.execute(self.input.sql)
            for row in cursor.fetchall():
                rows.append(list(row))
        except:
            self.ui.literal(traceback.format_exc())
            self.ui.warning("exception occurred while"
                            " executing a SQL query")
            connection.rollback()
            new_output = None
        else:
            connection.commit()
            new_output = self.Output(sql=self.input.sql, rows=rows)
        return new_output

To use this extension, add parameter -E sql.py to all PBBT invocations. For example:

$ pbbt -E sql.py input.yaml output.yaml --train

Now we will explain the content of sql.py line by line.

The first line imports some classes and decorators we will use:

from pbbt import Test, Field, BaseCase, MatchCase

To register a test type, use test decorator. Here is a most general template:

@Test
class CustomCase(object):

    class Input:
        some_field = Field(...)
        [...]

    class Output:
        some_field = Field(...)
        [...]

    def __init__(self, ctl, input, output):
        self.ctl = ctl
        self.input = input
        self.output = output

    def __call__(self):
        [...]
        return new_output

The argument of the decorator must be a class that adheres the following rules:

  • The structure of input and output records is described with nested classes Input and Output. Record fields are specified using Field descriptor.

  • To create a test case, the test harness makes an instance of the class. The class constructor accepts three arguments:

    ctl

    Test harness controller. It is used for user interaction, reporting test success or failure, and as a storage for conditional variables.

    input

    The input record.

    output

    The expected output record or None.

  • The test case is executed by calling its __call__() method. This method must run the test case, generate a new output record, and compare it with the given expected output record.

    If the expected and actual output record do not coincide:

    • In the check mode, the method must report a failure.
    • In the train mode, the method may ask the user for permission to update expected output. If expected output is to be updated, the method should return the new output record.

PBBT also provides two mixin classes: BaseCase and MatchCase. These classes implement most of the necessary functionality for common types of tests.

Let's review connect test type, which is implemented by ConnectCase class:

@Test
class ConnectCase(BaseCase):

    class Input:
        connect = Field(str)

    def check(self):
        [...]

ConnectCase is inherited from BaseCase, which is suitable for test which produce no output data and are executed for their side effects. The nested Input class definition is used to declare the fields of the input record. In this case, the input record has just one text field connect, which contains the name of the database.

Test types inherited from BaseCase must override abstract method check():

import sqlite3, traceback
[...]

@Test
class ConnectCase(BaseCase):
    [...]

    def check(self):
        self.state['connect'] = None
        try:
            self.state['connect'] = sqlite3.connect(self.input.connect)
        except:
            self.ui.literal(traceback.format_exc())
            self.ui.warning("exception occurred while"
                            " connecting to the database")
            self.ctl.failed()
        else:
            self.ctl.passed()

This code attempts to create a new database connection and store it as a conditional variable connect. If the attempt fails, it calls ui.literal() to display the exception traceback and ctl.failed() to report test failure. Otherwise, ctl.passed() is called to indicate that the test succeeded.

Next, let's review sql test type:

@Test
class SQLCase(MatchCase):

    class Input:
        sql = Field(str)

    class Output:
        sql = Field(str)
        rows = Field(listof(listof(object)))

    def run(self):
        [...]

    def render(self, output):
        [...]

This test type has both input and output records, which are described with Input and Output nested classes. The input record contains one field sql, a SQL query to execute. The output record contains two fields: sql and rows. Field sql contains the same SQL query and is used for matching the output record with the complementary input record. Field rows contains a list of output rows generated by the SQL query.

SQLCase is inherited from MatchCase, which is a mixin class for test types that produce text output. Test types inherited from MatchCase must override two methods:

run()
Executes the test case and returns the produced output record.
render(output)
Convert an output record to printable form.

In SQLCase, render() is implemented by converting output rows to CSV format:

import csv, StringIO
[...]

@Test
class SQLCase(MatchCase):
    [...]

    def render(self, output):
        stream = StringIO.StringIO()
        writer = csv.writer(stream, lineterminator='\n')
        writer.writerows(output.rows)
        return stream.getvalue()

Method run() is implemented as follows:

@Test
class SQLCase(MatchCase):
    [...]

    def run(self):
        connection = self.state.get('connect')
        if not connection:
            self.ui.warning("database connection is not defined")
            return
        rows = []
        try:
            cursor = connection.cursor()
            cursor.execute(self.input.sql)
            for row in cursor.fetchall():
                rows.append(list(row))
        except:
            self.ui.literal(traceback.format_exc())
            self.ui.warning("exception occurred while"
                            " executing a SQL query")
            connection.rollback()
            new_output = None
        else:
            connection.commit()
            new_output = self.Output(sql=self.input.sql, rows=rows)
        return new_output

Command-line Interface

Usage:

pbbt [<options>] INPUT [OUTPUT]

Here, INPUT and OUTPUT are files which contain input and output test data respectively.

The following options are available:

-h, --help
Display usage information and exit.
-q, --quiet
Show only warnings and errors.
-T, --train
Run tests in the training mode.
-P, --purge
Purge stale output data.
-M N, --max-errors N
Halt after N tests failed; 0 means "never".
-D VAR, -D VAR=VALUE, --define VAR, --define VAR=VALUE
Set a conditional variable.
-E FILE, -E MODULE, --extend FILE, --extend MODULE
Load a PBBT extension from a file or a Python module.
-S ID, --suite ID
Run a specific test suite.

PBBT can also read configuration from the following files:

setup.cfg
This file is in INI format with PBBT settings defined in section [pbbt]. The following parameters are recognized: extend, input, output, define, suite, train, purge, max_errors, quiet.
pbbt.yaml
This file must be a YAML file with the following keys: extend, input, output, define, suite, train, purge, max-errors, quiet.

PBBT can also be executed as a Distutils command:

python setup.py pbbt

In this case, PBBT configuration could be specified in setup.cfg or via command-line parameters.

API Reference

pbbt.maybe(T)
Pseudo-type for isinstance(X, ...): checks if X is an instance of T or equal to None.
pbbt.oneof(T1, T2, ...)
Pseudo-type for isinstance(X, ...): checks if X is an instance of T1 or an instance of T2, etc.
pbbt.choiceof(values)
Pseudo-type for isinstance(X, ...): checks if X is equal to one of the given values.
pbbt.listof(T, length=None)
Pseudo-type for isinstance(X, ...): checks if X is a list of elements of T.
pbbt.tupleof(T1, T2, ...)
Pseudo-type for isinstance(X, ...): checks if X is a tuple with fields of types T1, T2, etc.
pbbt.dictof(T1, T2)
Pseudo-type for isinstance(X, ...): checks if X is a dictionary with keys of type T1 and values of type T2.
pbbt.raises(E)
Use with with clause to verify that the nested block raises an exception of the given type.
pbbt.raises(E, fn, *args, **kwds)
Verifies that the function call fn(*args, **kwds) raises an exception of the given type.
pbbt.Test(CaseType)
Registers the given class as a test type.
pbbt.Field(check=None, default=REQUIRED, order=None, hint=None)

Describes a field of an input or an output record.

check
The type of the field value.
default
Default value if the field is missing. If not set, the field is mandatory.
order
If set, allows you to override the default field order.
hint
Short description of the field.
pbbt.Record(*p_fields, **kv_fields)

Base class for all input and output records.

The Test decorator converts nested Input and Output classes to Record subclasses.

The following methods could be used or overridden:

classmethod __recognizes__(keys)

Checks if the set of keys is compatible with the record type.

The default implementation checks if the set of keys contains all mandatory record fields.

classmethod __load__(mapping)
Generates a record instance from a mapping of record keys and values.
__dump__()
Generates a list of field keys and values.
__complements__(other)
Checks if two records are complementary input and output records for the same test case.
__clone__(**kv_fields)
Makes a copy of the record with new values for the given fields.
__str__()
Generates a printable representation.
pbbt.Control(...)

Test harness.

The test harness object is passed to test cases as the first argument of the constructor. The following methods and attributes could be used by test case objects.

Attributes:

ui
Provides user interaction services.
training
If set, the harness is in training mode.
purging
If set, the harness is in purging mode.
quiet
If set, only warnings and errors are displayed.
halted
If set, the test process has been halted.
state
Conditional variables.
selection
Selected suites.

Methods:

passed(text=None)
Attests that the current test case has passed.
failed(text=None)
Attests that the current test case has failed.
updated(text=None)
Attests that output data for the current test case has been updated.
halt(text=None)
Halts testing process.
load_input(path)
Loads input test data from the given file.
load_output(path)
Loads output test data from the given file.
dump_output(path, data)
Saves output test data to the given file.
run(case)
Executes a test case.
__call__(input_path, output_path)
Runs testing process with the given input and output.
pbbt.locate(record)
Finds the location of the given record.
pbbt.Location(filename, line)
Position of an input or output record in the YAML document.
pbbt.run(input_path, output_path=None, **configuration)

Loads test data from the given files and runs the tests.

Configuration:

ui
User interaction controller.
variables
Predefined conditional variables.
targets
Selected suites.
training
Set the harness into training mode.
purging
Purge stale output data.
max_errors
Maximum permitted number of failures before the harness halts.
quiet
Display only warnings and errors.
pbbt.main()
Implements the pbbt command-line utility.
pbbt.BaseCase

Template class suitable for most test types.

Subclasses need to override the following methods:

check()
Runs the test case in check mode.
train()
Runs the test case in train mode. The default implementation simply calls check().
pbbt.MatchCase

Template class for test types which produce output.

Subclasses need to override the following methods:

run()
Executes the case; returns produced output.
render(output)
Converts output record to text.
pbbt.UI

Abstract class for user interaction services.

Methods:

part()
Starts a new section.
section()
Starts a subsection.
header(text)
Shows a section header.
notice(text)
Shows a notice.
warning(text)
Shows a warning.
error(text)
Shows an error.
literal(text)
Shows a literal text.
choice(text, *choices)
Asks a single-choice question.
pbbt.ConsoleUI(stdin=None, stdout=None, stderr=None)
Implements UI for console input/output.
pbbt.SilentUI(backend)
Implements UI for use with --quiet option.