pytest-csv-params

A pytest plugin to parametrize data-driven tests by CSV files.

Requirements

• Python 3.8, 3.9 or 3.10
• pytest >= 7.1

There's no operating system dependent code in this plugin, so it should run anywhere where pytest runs.

Installation

Simply install it with pip...

pip install pytest-csv-params

... or poetry ...

poetry add --dev pytest-csv-params

Usage: Command Line Argument

Argument	Required	Description	Example
--csv-params-base-dir	no (optional)	Define a base dir for all relative-path CSV data files (since 0.1.0)	pytest --csv-params-base-dir /var/testdata

Usage: Decorator

Simply decorate your test method with @csv_params and the following parameters:

Parameter	Type	Description	Example
data_file	str	The CSV file to use, relative or absolute path	"/var/testdata/test1.csv"
base_dir	str (optional)	Directory to look up relative CSV files (see data_file); overrides the command line argument	join(dirname(__file__), "assets")
id_col	str (optional)	Column name of the CSV that contains test case IDs	"ID#"
dialect	csv.Dialect (optional)	CSV Dialect definition (see Python CSV Documentation)	csv.excel_tab
data_casts	dict (optional)	Cast Methods for the CSV Data (see "Data Casting" below)	{ "a": int, "b": float }
header_renames	dict (optional)	Replace headers from the CSV file, so that they can be used as parameters for the test function (since 0.3.0)	{ "Annual Amount of Bananas": "banana_count", "Cherry export price": "cherry_export_price" }

CSV File Lookup Order

CSV files are looked up following this rules:

• If the data_file parameter is an absolute path, this is used, regardless of the base_dir parameter or command line argument.
• If the data_file parameter is relative:
  • If the base_dir parameter is set, the file is looked up there, regardless of the command line argument
  • If the base_dir parameter is not set (or None):
    • If the command line argument is set, the file is looked up there
    • If the command line argument is not set, the file is looked up in the current working directory

Data Casting

When data is read from CSV, they are always parsed as str. If you need them in other formats, you can set a method that should be called with the value.

These methods can also be lambdas, and are also good for further transformations.

Data Casting Example

from pytest_csv_params.decorator import csv_params

def normalize(x: str) -> str:
    return x.strip().upper()

@csv_params(
    data_file="/test/data.csv",
    data_casts={
        "col_x": normalize,
        "col_y": float,
    },
)
def test_something(col_x, col_y):
    # Test something...
    ...

CSV Format

The default CSV format is:

• \r\n as line ending
• All non-numeric fields are surrounded by "
• If you need a " in the value, use "" (double quote)
• Fields are separated by comma (,)

The first line must contain the row names. Row names must match the parameters of the test method (except for an ID column that is configured as such -- see id_col decorator parameter).

Example CSV

"ID#", "part_a", "part_b", "expected_result"
"first", 1, 2, 3
"second", 3, 4, 7
"third", 10, 11, 21

Headers

The header line is very important, as it maps the values to parameters of the test function. The plugin supports you with that. The following rules apply:

• Every character that is not valid in a variable name is replaced by an underscore (_)
• If the character at the start is not a letter or an underscore, it is replaced by an underscore(_)
• If the name is still invalid then, because it's a keyword or a builtin name, an exception is raised (CsvHeaderNameInvalid)

If you don't want to change your CSV file, you can use the header_renames parameter to the decorator to rename headers as needed.

Headers must be unique, and an Exception is raised if not (CsvHeaderNameInvalid).

The header handling was heavily improved in Version 0.3.0.

Usage Example

This example uses the CSV example from above.

from pytest_csv_params.decorator import csv_params

@csv_params(
    data_file="/data/test-lib/cases/addition.csv",
    id_col="ID#",
    data_casts={
        "part_a": int,
        "part_b": int,
        "expected_result": int,
    },
)
def test_addition(part_a, part_b, expected_result):
    assert part_a + part_b == expected_result

Shorthand example (no ID col, only string values):

from pytest_csv_params.decorator import csv_params

@csv_params("/data/test-lib/cases/texts.csv")
def test_texts(text_a, text_b, text_c):
    assert f"{text_a}:{text_b}" == text_c

More complex example

This example features nearly all things the plugin has to offer. You find this example also in the test cases, see tests/test_complex_example.py.

The CSV file (tests/assets/example.csv):

"Test ID","Bananas shipped","Single Banana Weight","Apples shipped","Single Apple Weight","Container Size"
"Order-7","1503","0.5","2545","0.25","1500"
"Order-15","101","0.55","1474","0.33","550"

The Test (tests/test_complex_example.py):

from math import ceil
from os.path import join, dirname

from pytest_csv_params.decorator import csv_params


@csv_params(
    data_file="example.csv",
    base_dir=join(dirname(__file__), "assets"),
    id_col="Test ID",
    header_renames={
        "Bananas shipped": "bananas_shipped",
        "Single Banana Weight": "banana_weight",
        "Apples shipped": "apples_shipped",
        "Single Apple Weight": "apple_weight",
        "Container Size": "container_size",
    },
    data_casts={
        "bananas_shipped": int,
        "banana_weight": float,
        "apples_shipped": int,
        "apple_weight": float,
        "container_size": int,
    },
)
def test_container_size_is_big_enough(
    bananas_shipped: int, banana_weight: float, apples_shipped: int, apple_weight: float, container_size: int
) -> None:
    """
    This is just an example test case for the documentation.
    """

    gross_weight = (banana_weight * bananas_shipped) + (apple_weight * apples_shipped)
    assert ceil(gross_weight) <= container_size

If you decide not to rename the columns, the test would look like this:

@csv_params(
    data_file="example.csv",
    base_dir=join(dirname(__file__), "assets"),
    id_col="Test ID",
    data_casts={
        "Bananas_Shipped": int,
        "Single_Banana_Weight": float,
        "Apples_Shipped": int,
        "Single_Apple_Weight": float,
        "Container_Size": int,
    },
)
def test_container_size_is_big_enough(
    Bananas_Shipped: int, Single_Banana_Weight: float, Apples_Shipped: int, Single_Apple_Weight: float, Container_Size: int
) -> None:
    ...

Breaking Changes

Version 0.3.0

• Column header names that are reserved keywords or builtin names are no longer accepted. You should have been in trouble already if you used them, so nothing should go wrong with this change and existing tests.

Version 0.2.0

• The parameter order for pytest_csv_params.decorator.csv_params changed to allow the shorthand usage with only a data_file as positional argument. If you used keyword arguments only (like the docs recommend), you will not run into trouble.

Contributing

Build and test

You need Poetry in order to build this project.

Tests are implemented with pytest, and tox is used to orchestrate them for the supported python versions.

• Checkout this repo
• Run poetry install
• Run poetry run tox (for all supported python versions) or poetry run pytest (for your current version)

Bugs etc.

Please send your issues to csv-params_issues (at) jued.de. Please include the following:

• Plugin Version used
• Pytest version
• Python version with operating system

It would be great if you could include example code that clarifies your issue.

Pull Requests

Pull requests are always welcome. Since this Gitea instance is not open to public, just send an e-mail to discuss options.

Any changes that are made are to be backed by tests. Please give me a sign if you're going to break the existing API and let us discuss ways to handle that.

Quality Measures

Backed with pytest plugins:

• Pylint (static code analysis and best practises)
• black (code formatting standards)
• isort (keep imports sorted)
• Bandit (basic static security tests)
• mypy (typechecking)

Please to a complete pytest run (poetry run pytest), and consider running it on all supported platforms with (poetry run tox).

License

Code is under MIT license. See LICENSE.txt for details.