true-formatter 0.2.0

The uncompromising Python formatter.

True — The Uncompromising Python Formatter

True is an opinionated Python source-code formatter. It enforces a consistent style so you never have to think about formatting again.

---

Table of Contents

• Features
• Installation
• Quick Start
• CLI Reference
• API Reference
• PEP 8 Rules Applied
• Architecture
• Contributing

---

Features

Category	What True fixes
Whitespace	Spaces around operators, after commas, inside brackets
Indentation	Any indent style → 4 spaces
Blank lines	2 before top-level def/class, 1 between methods, 2 after last def
Imports	Split import os, sys into separate lines
Strings	Normalise 'hello' → "hello"
Semicolons	x = 0; y = 0 → two lines
Compound statements	if x: pass → split onto two lines
Lambda assignment	f = lambda x: x → def f(x): return x
Class names	class my_class → class MyClass
Variable names	xValue → x_value, calcArea → calc_area
Default params	def f(x =1) → def f(x=1)
Annotated params	def f(x: int=1) → def f(x: int = 1)
Keyword args	foo(key =1) → foo(key=1)
Power operator	2 ** 10 → 2**10
Extra spaces	for i in range → for i in range
Trailing whitespace	Stripped from every line
Final newline	Exactly one trailing newline

Zero dependencies — pure stdlib.

---

Installation

pip install true-formatter

Or from source:

git clone https://github.com/yourname/true-formatter
cd true-formatter
pip install -e .[dev]

---

Quick Start

CLI

# Format a file in-place
true my_script.py

# Check without modifying (for CI)
true --check my_script.py

# Show a unified diff
true --diff my_script.py

# Format an entire directory
true src/

# Read from stdin
echo "x=1" | true -

Library

import true_formatter

code = "x=1\ny=  'hello'\n"
result = true_formatter.format_str(code, mode=true_formatter.Mode())
print(result)
# x = 1
# y = "hello"

---

CLI Reference

Synopsis

true [OPTIONS] [FILES|DIRECTORIES]...

Pass - as a file path to read from stdin and write to stdout.

Options

Flag	Default	Description
--check	off	Don't write files; exit 1 if any would change
--diff	off	Print a unified diff instead of rewriting
-l, --line-length INT	88	Maximum line length
-S, --skip-string-normalization	off	Leave string quotes as-is
-q, --quiet	off	Suppress all non-error output

Exit Codes

Code	Meaning
0	All files already formatted (or successfully reformatted)
1	--check mode: at least one file would be reformatted

Examples

# Format a single file
true app.py

# Format all Python files in a directory
true src/

# CI — fail if anything needs reformatting
true --check src/ tests/

# Preview changes before applying
true --diff app.py

# Pipe from stdin
cat messy.py | true - > clean.py

# Use a shorter line length
true -l 79 legacy_module.py

# Preserve existing string quotes
true -S my_file.py

Integration

pre-commit:

repos:
  - repo: local
    hooks:
      - id: true-formatter
        name: true
        entry: true
        language: python
        types: [python]

Makefile:

fmt:
	true src/ tests/

check:
	true --check src/ tests/

GitHub Actions:

- name: Check formatting
  run: |
    pip install true-formatter
    true --check src/ tests/

---

API Reference

Top-level Functions

format_str(src_contents, *, mode) → str

Format a Python source string and return the result.

import true_formatter

result = true_formatter.format_str("x=1", mode=true_formatter.Mode())
# → "x = 1\n"

Parameter	Type	Description
src_contents	str	Python source code to format
mode	Mode	Formatting configuration

Raises TypeError if src_contents is not a str. Raises TrueFormattingError if the source cannot be tokenized.

---

format_file_contents(src_contents, *, mode) → str

Like format_str but additionally guarantees a single trailing newline. Use this when writing back to a file.

formatted = true_formatter.format_file_contents(src, mode=mode)
Path("my_file.py").write_text(formatted, encoding="utf-8")

---

check_format(src_contents, *, mode) → bool

Return True if src_contents is already correctly formatted. No files are written.

ok = true_formatter.check_format("x = 1\n", mode=true_formatter.Mode())
# → True

ok = true_formatter.check_format("x=1\n", mode=true_formatter.Mode())
# → False

---

Mode

Dataclass that holds all formatting options.

from true_formatter import Mode

mode = Mode(
    line_length=79,
    string_normalization=True,
    skip_string_normalization=False,
    magic_trailing_comma=True,
)

Field	Type	Default	Description
line_length	int	88	Maximum line length
string_normalization	bool	True	Rewrite '…' as "…"
skip_string_normalization	bool	False	Disable string normalisation
magic_trailing_comma	bool	True	Preserve trailing commas
target_versions	set[TargetVersion]	set()	Python versions to target
rules	RuleSet	DEFAULT_RULES	Active rule set

Properties:

• mode.quote_char — returns '"' or "'" based on normalisation settings.

---

TargetVersion

Enum of supported Python versions.

from true_formatter import TargetVersion, Mode

mode = Mode(target_versions={TargetVersion.PY311, TargetVersion.PY312})

Values: PY38, PY39, PY310, PY311, PY312

---

Rules

Rule (abstract base class)

Extend this to create custom formatting rules that operate on the token stream.

import tokenize
from true_formatter import Rule, Mode

class BanPassRule(Rule):
    name = "ban-pass"

    def apply(self, tokens, mode):
        return [t for t in tokens if t.string != "pass"]

Attribute	Type	Description
name	str	Unique identifier shown in --diff output

Required method: apply(tokens, mode) → list[TokenInfo]

---

RuleSet

An ordered, immutable-style collection of rules.

from true_formatter import DEFAULT_RULES, Mode
from my_rules import BanPassRule

custom_rules = DEFAULT_RULES.add(BanPassRule())
mode = Mode(rules=custom_rules)

Method	Returns	Description
add(rule)	RuleSet	New set with rule appended
remove(name)	RuleSet	New set without the named rule
__len__()	int	Number of rules
__iter__()	iterator	Iterate over rules in order

---

Exceptions

TrueError
├── TrueFormattingError   ← source cannot be parsed / formatted
└── TrueConfigError       ← invalid Mode configuration

from true_formatter.exceptions import TrueFormattingError

try:
    true_formatter.format_str("def (:", mode=true_formatter.Mode())
except TrueFormattingError as e:
    print(f"Bad source: {e}")

---

PEP 8 Rules Applied

Code	Rule	Example
E101	Indentation → 4 spaces	tab/2-space → 4 space
E201	No space after ( [ {	foo( x ) → foo(x)
E202	No space before ) ] }	foo(x ) → foo(x)
E203	No space before , ; :	x , y → x, y
E211	No space before ( in call	foo (x) → foo(x)
E225	Spaces around operators	x==1 → x == 1
E231	Space after , : ;	[1,2] → [1, 2]
E251	No spaces around = in default params	def f(x =1) → def f(x=1)
E252	Spaces around = in annotated params	def f(x: int=1) → def f(x: int = 1)
E261	Two spaces before inline comment	x=1 # note → x = 1 # note
E262	Space after # in comment	#note → # note
E271	Single space after keyword	for i → for i
E302	Two blank lines before top-level def/class
E301	One blank line between methods in class
E305	Two blank lines after last def before code
E401	One import per line	import os, sys → two lines
E702	No semicolons between statements	x=0; y=0 → two lines
E711	Compound statements split onto separate lines	if x: pass → two lines
E731	Lambda assignment → def	f = lambda x: x → def f(x): return x
W291	No trailing whitespace
W292	File ends with newline
N801	Class names in PascalCase	class my_class → class MyClass
N802	Function/variable names in snake_case	calcArea → calc_area

---

Architecture

Formatting Pipeline

Source string
     │
     ▼
 Tokenizer              (stdlib tokenize)
     │
     ▼
 Rule pipeline          (true_formatter.rules)
     │  each Rule: token list → token list
     ▼
 Emitter                (tokenize.untokenize)
     │
     ▼
 Text-level passes      (true_formatter.transforms)
     │
     ▼
 Formatted string

Text-level Passes (in order)

Pass	PEP 8 codes
fix_indentation	E101, W191
fix_multiple_imports	E401
split_semicolons	E702, E711
fix_lambda_assignment	E731
fix_extra_whitespace	E271, E272
fix_class_names	N801
fix_variable_names	N802
fix_pep8_whitespace	E201, E202, E203, E211, E225, E231, E251, E252, E261, E262
fix_operator_priority_spacing	E226
normalise_strings	W605
fix_trailing_whitespace	W291, W293
fix_blank_lines	E302, E301
fix_blank_lines_after_def	E305
ensure_final_newline	W292

Module Map

true_formatter/
├── __init__.py       Public API — re-exports from submodules
├── core.py           format_str / format_file_contents / check_format
├── transforms.py     Text-level formatting passes
├── rules.py          Rule base class, RuleSet, built-in rules
├── exceptions.py     TrueError hierarchy
└── cli.py            argparse CLI entry point

---

Contributing

Setup

git clone https://github.com/yourname/true-formatter
cd true-formatter
pip install -e .[dev]

Running the Tests

pytest                          # all tests
pytest tests/test_transforms.py # one module
pytest -k "test_single"         # filter by name
pytest --tb=short               # compact tracebacks

Writing a New Rule

Rules live in true_formatter/rules.py and operate on the token stream:

import tokenize
from true_formatter.rules import Rule

class NoSemicolonRule(Rule):
    name = "no-semicolons"

    def apply(self, tokens, mode):
        return [t for t in tokens if t.string != ";"]

Add it to DEFAULT_RULES and write tests in tests/test_rules.py.

Writing a New Text-level Transform

Transforms live in true_formatter/transforms.py and operate on strings:

def remove_semicolons(src: str) -> str:
    return src.replace(";", "")

Wire it into _text_level_format in core.py in the correct position in the pipeline.

Test Guidelines

• One test class per module under test
• Each test method tests exactly one behaviour
• Use pytest.raises for exception assertions
• Use tmp_path fixture for file I/O tests
• Name tests test_<thing_being_tested>

Code Style

True is formatted with itself:

true true_formatter/ tests/

---

License

MIT © True Contributors