UniExpect

UniExpect (UE) offers inline testing for any REPL language, making the notion of doctests more universal and even more versatile. There are two immediate benefits to using the UE utility:

1. With this, code in all supported languages can be placed directly below executable, easily-checked tests that help to explain a function or class's basic functionality.

2. The "unittest" approach to testing antiquates manual, sandbox testing. Writing code in SQL could, for example, could benefit immensely. Each accidental table modification means you have to restart the session, copy in the data, and then try your code again. With UE, it's just one command: expect test.sql.

See below for how to get started.

Installation

UniExpect is now installable via pypi:

pip install uniexpect

How to Use

The standard expect syntax matches that of Python doctests. By default, the test input prefix is >>>. In other words, tests following the following format.

>>> <input>
<output>

Any language with block comments, such as /* ... */ or """ ... """, can use the following syntax:

/*
>>> <test input>
<expected output>
>>> <test input>
<expected output>
*/

For languages that only support inline comments, UE default configuration files use the following convention to denote test suites:

###
# >>> <test input>
# <expected output>
# >>> <test input>
# <expected output>
###

The comment symbol is simply repeated three times before and after the test suite. For more concrete examples, see the samples/ folder.

How to Run

Usage is simple. To run expect on samples/scheme.scm:

expect samples/scheme.scm

Settings

The following is an abridged list of more commonly-used settings. For a full list, run expect --help.

--language=<language>

If the language is not specified, UE will (1) assume the file extension is the language name and, if no such configuration file exists, will (2) search all preference files with the same first letter as the extension.

Example: expect samples/sql.sql --language=sql

--verbose

Just as Python doctests do, UniExpect only reports incorrect outputs by default. To view all output, pass the verbose flag. Add vs to increase verbosity, with -vvv for maximum whining.

Examples:

• expect samples/python.py --verbose
• expect samples/python.py -vvv

Support

UniExpect can be used for nearly any programming language. Here are a few that I've already setup configurations for:

• python2.7

expect samples/python.py --language=python2

• python3

expect samples/python.py --language=python3

• scheme

expect samples/scheme.scm

version: chibi-scheme

• sqlite3

expect samples/sqlite.sql

UE has also been setup to work with custom programming languages.

• berkeleyscheme

expect samples/berkeleyscheme.scm --language=berkeleyscheme

How to Add a Language

Take the Scheme specification as an example. The first several sections are self-explanatory: we describe the shell basics, language information, and comment styles. As for the tests, each dictionary represents a different test type. The input_prefix and output_prefix denote the question and expected response, respectively. block_comments indicates whether or not we can expect this test type into a comment block. The inline_comments likewise indicates whether or not we can expect this test type in inline comments. See more examples in the samples/ directory.

shell = {
    'command': 'scheme',
    'prompt': '> ',
    'continuation': '>> ',
    '_load_file': '(load "{filename}")'
}

language = 'scheme'
extension = 'scm'

# doesn't exist in scheme, but we will use ;;; to denote multi-line tests
block_comments = [(';;;', ';;;')]

inline_comments = [';']

tests = [
    {
        'input_prefix': '; >>>',  # prefix for test input
        'output_prefix': ';',  # prefix for test output
        'block_comments': True,  # just like standard doctests
        'inline_comments': False
    },
    {
        'input_prefix': '> ',
        'output_prefix': '=>',
        'one-liner': True,
        'block_comments': True,
        'inline_comments': True
    }
]