Cercis /ˈsɜːrsɪs/ is a Python code formatter that is more configurable than Black (a popular Python code formatter).

Cercis is also the name of a deciduous tree that boasts vibrant pink to purple-hued flowers, which bloom in early spring.

This code repository is forked from and directly inspired by Black. The original license of Black is included in this repository (see LICENSE_ORIGINAL).

Cercis inherited Black's very comprehensive test cases, which means we are confident that our configurability addition does not introduce any undesirable side effects. We also thoroughly tested every configurable options that we added.

In particular, via its configurable options, Cercis can completely fall back to Black. See Section 4.5 below for more details.

While we like the idea of auto-formatting and code readability, we take issue with some style choices and the lack of configurability of the Black formatter. Therefore, Cercis aims at providing some configurability beyond Black's limited offering.

Cercis can be installed by running pip install cercis. It requires Python 3.7+ to run. If you want to format Jupyter Notebooks, install with pip install "cercis[jupyter]".

To get started right away with sensible defaults:

cercis {source_file_or_directory}

You can run Cercis as a package if running it as a script doesn't work:

python -m cercis {source_file_or_directory}

The commands above reformat entire file(s) in place.

To format Python files (.py), put the following into your .pre-commit-config.yaml file. Remember to replace <VERSION> with your version of this tool (such as v0.1.0):

- repo: https://github.com/jsh9/cercis
  rev: <VERSION>
  hooks:
    - id: cercis
      args: [--line-length=88]

To format Jupyter notebooks (.ipynb), put the following into your .pre-commit-config.yaml file:

- repo: https://github.com/jsh9/cercis
  rev: <VERSION>
  hooks:
    - id: cercis-jupyter
      args: [--line-length=88]

See pre-commit for more instructions. In particular, here is how to specify arguments in pre-commit config.

Cercis's code style is largely consistent with the style of of Black.

The main difference is that Cercis provides several configurable options that Black doesn't. Configurability is our main motivation behind creating Cercis.

Cercis offers the following configurable options:

1. Line length
2. Single quote vs double quote
3. Tabs vs spaces
4. Base indentation spaces
5. Extra indentation at line continuation
   1. At function definition
   2. In other line continuations
   3. At closing brackets
6. "Simple" lines with long strings
7. Collapse nested brackets
8. Wrap pragma comments

The next section (How to configure Cercis) contains detailed instructions of how to configure these options.

Cercis uses 79 characters as the line length limit, instead of 88 (Black's default).

You can override this default if necessary.

Cercis uses single quotes (') as the default for strings, instead of double quotes (") which is Black's default.

You can override this default if necessary.

Cercis offers users the ability to use tabs rather than spaces.

There are two associated options:

• --use-tabs (bool): whether to use tabs or spaces to format the code

• --tab-width (int): when calculating line length (to determine whether to wrap lines), how wide shall Cercis treat each tab. Only effective when --use-tabs is set to True.

This option defines the number of spaces that each indentation level adds. This option has no effect when --use-tabs is set to True.

For example, if you set it to 2, contents within a for block is indented 2 spaces:

for i in (1, 2, 3, 4, 5):
  print(i)

There are three associated options:

• --function-definition-extra-indent
• --other-line-continuation-extra-indent
• --closing-bracket-extra-indent

They control whether we add an additional indentation level in some situations. Note that these options can work well with tabs (--use-tabs=True).

# Cercis's default style
def some_function(
        arg1_with_long_name: str,
        arg2_with_longer_name: int,
        arg3_with_longer_name: float,
        arg4_with_longer_name: bool,
) -> None:
    ...

# Black's style (not configurable)
def some_function(
    arg1_with_long_name: str,
    arg2_with_longer_name: int,
    arg3_with_longer_name: float,
    arg4_with_longer_name: bool,
) -> None:
    ...

We choose to add an extra indentation level when wrapping a function signature line. This is because def␣ happens to be 4 characters, so when the base indentation is 4 spaces, it can be difficult to visually distinguish the function name and the argument list if we don't add an extra indentation.

If you set --base-indentation-spaces to other values than 4, this visual separation issue will disappear, and you may not need to turn this option on.

This style is encouraged in PEP8.

"Other line continuations" are cases other than in function definitions, such as:

var = some_function(
    arg1_with_long_name,
    arg2_with_longer_name,
)

var2 = [
    'something',
    'something else',
    'something more',
]

So if you set this option (--other-line-continuation-extra-indent) to True, you can add an extra level of indentation in these cases:

var = some_function(
        arg1_with_long_name,
        arg2_with_longer_name,
)

var2 = [
        'something',
        'something else',
        'something more',
]

This option lets people customize where the closing bracket should be. Note that both styles are OK according to PEP8.

# --closing-bracket-extra-indent=False

def function(
        arg1: int,
        arg2: float,
        arg3_with_long_name: list,
) -> None:
    print('Hello world')


result = func2(
    12345,
    3.1415926,
    [1, 2, 3],
)


something = {
    'a': 1,
    'b': 2,
    'c': 3,
}

# --closing-bracket-extra-indent=True

def function(
        arg1: int,
        arg2: float,
        arg3_with_long_name: list,
        ) -> None:
    print('Hello world')


result = func2(
    12345,
    3.1415926,
    [1, 2, 3],
    )


something = {
    'a': 1,
    'b': 2,
    'c': 3,
    }

By default, Black wraps lines that exceed length limit. But for very simple lines (such as assigning a long string to a variable), line wrapping is not necessary.

# Cercis's default style
# (Suppose line length limit is 30 chars)

# Cercis doesn't wrap slightly long lines
var1 = 'This line has 31 chars'



# Cercis doesn't wrap longer lines
var2 = 'This line has 43 characters_______'


# Falls back to Black when comments present
var3 = (
    'shorter line'  # comment
)

# Black's style (not configurable)
# (Suppose line length limit is 30 chars)

# Black wraps slightly long lines
var1 = (
    "This line has 31 chars"
)

# But Black doesn't wrap longer lines
var2 = "This line has 43 characters_______"


# Black wraps comments like this:
var3 = (
    "shorter line"  # comment
)

Cercis by default collapses nested brackets to make the code more compact.

# Cercis's default style

# If line length limit is 30
value = np.array([
    [1, 2, 3, 4, 5],
    [6, 7, 8, 9, 0],
])



# If line length limit is 10
value = function({
    1,
    2,
    3,
    4,
    5,
})

# Black's style (not configurable)

# If line length limit is 30
value = np.array(
    [
        [1, 2, 3, 4, 5],
        [6, 7, 8, 9, 0],
    ]
)

# If line length limit is 10
value = function(
    {
        1,
        2,
        3,
        4,
        5,
    }
)

The code implementation of this option comes from Pyink, another forked project from Black.

"Pragma comments", in this context, mean the directives for Python linters usually to tell them to ignore certain errors. Pragma comments that Cercis currently recognizes include:

• noqa: # noqa: E501
• type: ignore: # type: ignore[no-untyped-def]
• pylint: # pylint: disable=protected-access
• pytype: # pytype: disable=attribute-error

# Cercis's default style
# (Suppose line length limit is 30)

# This line has 30 characters
var = some_func(some_long_arg)  # noqa:F501

# This line has 31 characters
var_ = some_func(
    some_long_arg
)  # type: ignore

# Cercis doesn't wraps a line if its main
# content (without the comment) does not
# exceed the line length limit.

# Black's style (not configurable)
# (Suppose line length limit is 30)

# Black doesn't wrap lines, no matter
# how long, if the line has
# a "# type: ignore..." comment.
# (This line has 31 characters.)
var_ = some_func(some_long_arg)  # type: ignore

# Black does not recognize "# type:ignore",
# even though mypy recognizes it.
var_ = some_func(
    some_long_arg
)  # type:ignore

# Black only recognizes "# type: ignore"
var_ = some_func(
    some_long_arg
)  # noqa:F501

Here are some examples:

• cercis --single-quote=True myScript.py to format files to single quotes
• cercis --function-definition-extra-indent=False myScript.py to format files without extra indentation at function definition
• cercis --line-length=79 myScript.py to format files with a line length of 79 characters

You can specify the options under the [tool.cercis] section of the file:

[tool.cercis]
line-length = 88
function-definition-extra-indent = true
single-quote = false

You can specify the options under the args section of your .pre-commit-config.yaml file.

For example:

repos:
  - repo: https://github.com/jsh9/cercis
    rev: 0.1.0
    hooks:
      - id: cercis
        args: [--function-definition-extra-indent=False, --ling-length=79]
  - repo: https://github.com/jsh9/cercis
    rev: 0.1.0
    hooks:
      - id: cercis-jupyter
        args: [--function-definition-extra-indent=False, --line-length=79]

The value in rev can be any Cercis release, or it can be main, which means to always use the latest (including unreleased) Cercis features.

Currently, Cercis does not support a config section in tox.ini. Instead, you can specify the options in pyproject.toml.

Here are the configuration options to fall back to Black's behavior. Put them in pyproject.toml:

[tool.cercis]
line-length = 88
single-quote = false
use-tabs = false
base-indentation-spaces = 4
function-definition-extra-indent = false
other-line-continuation-extra-indent = false
closing-bracket-extra-indent = false
wrap-line-with-long-string = true
collapse-nested-brackets = false
wrap-pragma-comments = true