Crate forge_fmt 

Formatter (fmt)

Solidity formatter that respects (some parts of) the Style Guide and is tested on the Prettier Solidity Plugin cases.

Architecture

The formatter is built on top of Solar, and the architecture is based on a Wadler-style pretty-printing engine. The formatting process consists of two main steps:

1. Parsing: The Solidity source code is parsed using solar into an Abstract Syntax Tree (AST). The AST is a tree representation of the code’s syntactic structure.
2. Printing: The AST is traversed by a visitor, which generates a stream of abstract tokens that are then processed by a pretty-printing engine to produce the final formatted code.

The Pretty Printer (pp)

The core of the formatter is a pretty-printing engine inspired by Philip Wadler’s algorithm, and adapted from the implementations in rustc_ast_pretty and prettyplease. Its goal is to produce an optimal and readable layout by making intelligent decisions about line breaks.

The process works like this:

1. AST to Abstract Tokens: The formatter’s State object walks the solar AST. Instead of directly writing strings, it translates the AST nodes into a stream of abstract formatting “commands” called Tokens. This decouples the code’s structure from the final text output. The primary tokens are:

   • String: An atomic, unbreakable piece of text, like a keyword (function), an identifier (myVar), or a literal (42).
   • Break: A potential line break. This is the core of the engine’s flexibility. The Printer later decides whether to render a Break as a single space or as a newline with appropriate indentation.
   • Begin/End: These tokens define a logical group of tokens that should be formatted as a single unit. This allows the printer to decide how to format the entire group at once.

2. Grouping and Breaking Strategy: The Begin and End tokens create formatting “boxes” that guide the breaking strategy. There are two main types of boxes:

   • Consistent Box (cbox): If any Break inside this box becomes a newline, then all Breaks inside it must also become newlines. This is ideal for lists like function parameters or struct fields, ensuring they are either all on one line or neatly arranged with one item per line.
   • Inconsistent Box (ibox): Breaks within this box are independent. The printer can wrap a long line at any Break point without forcing other breaks in the same box to become newlines. This is useful for formatting long expressions or comments.

3. The Printer Engine: The Printer consumes this stream of tokens and makes the final decisions:

   • It maintains a buffer of tokens and tracks the remaining space on the current line based on the configured line_length.
   • When it encounters a Begin token for a group, it calculates whether the entire group could fit on the current line if all its Breaks were spaces.
   • If it fits, all Breaks in that group are rendered as spaces.
   • If it doesn’t fit, Breaks are rendered as newlines, and the indentation level is adjusted accordingly based on the box’s rules (consistent or inconsistent).

Crucially, this entire process is deterministic. Because the formatter completely rebuilds the code from the AST, it discards all original whitespace, line breaks, and other stylistic variations. This means that for a given AST and configuration, the output will always be identical. No matter how inconsistently the input code is formatted, the result is a single, canonical representation, ensuring predictability and consistency across any codebase.

Debug Mode: To visualize the debug output, and understand how the pretty-printer makes its decisions about boxes and breaks, see the Debug section in Testing.

Comments

Comment handling is a critical aspect of the formatter, designed to preserve developer intent while restructuring the code.

1. Categorization: Comments are parsed and categorized by their position and style: Isolated (on its own line), Mixed (on a line with code), and Trailing (at the end of a line).

2. Blank Line Handling: Blank lines in the source code are treated as a special BlankLine comment type, allowing the formatter to preserve vertical spacing that separates logical blocks of code. However, to maintain a clean and consistent vertical rhythm, any sequence of multiple blank lines is collapsed into a single blank line. This prevents excessive empty space in the formatted output.

3. Integration with Printing: During the AST traversal, the formatter queries for comments that appear before the current code element. These comments, including blank lines, are then strategically inserted into the Printer’s token stream. The formatter inserts Break tokens around comments to ensure they are correctly spaced from the surrounding code, and emits one or two hardbreaks for blank lines to maintain the original vertical rhythm.

This approach allows the formatter to respect both the syntactic structure of the code and the developer’s textual annotations and spacing, producing a clean, readable, and intentional layout.

Example

Source Code

pragma   solidity ^0.8.10 ;
contract  HelloWorld {
    string   public message;
    constructor(  string memory initMessage) { message = initMessage;}
}



event    Greet( string  indexed  name) ;

Abstract Syntax Tree (AST) (simplified)

SourceUnit
 ├─ PragmaDirective("solidity", "^0.8.10")
 ├─ ItemContract("HelloWorld")
 │   ├─ VariableDefinition { name: "message", ty: "string", visibility: "public" }
 │   └─ ItemFunction {
 │         kind: Constructor,
 │         header: FunctionHeader {
 │            parameters: [
 │               VariableDefinition { name: "initMessage", ty: "string", data_location: "memory" }
 │            ]
 │         },
 │         body: Block {
 │            stmts: [
 │               Stmt { kind: Expr(Assign {lhs: Ident("message"), rhs: Ident("initMessage")}) }
 │            ]
 │         }
 │      }
 └─ ItemEvent { name: "Greet", parameters: [
       VariableDefinition { name: "name", ty: "string", indexed: true }
    ] }

Formatted Source Code The code is reconstructed from the AST using the pretty-printer.

pragma solidity ^0.8.10;

contract HelloWorld {
    string public message;

    constructor(string memory initMessage) {
        message = initMessage;
    }
}

event Greet(string indexed name);

Configuration

The formatter supports multiple configuration options defined in foundry.toml.

Option	Default	Description
line_length	120	Maximum line length where the formatter will try to wrap the line.
tab_width	4	Number of spaces per indentation level. Ignored if style is tab.
style	space	The style of indentation. Options: space, tab.
bracket_spacing	false	Print spaces between brackets.
int_types	long	Style for uint256/int256 types. Options: long, short, preserve.
multiline_func_header	attributes_first	The style of multiline function headers. Options: attributes_first, params_always, params_first_multi, all, all_params.
quote_style	double	The style of quotation marks. Options: double, single, preserve.
number_underscore	preserve	The style of underscores in number literals. Options: preserve, remove, thousands.
hex_underscore	remove	The style of underscores in hex literals. Options: preserve, remove, bytes.
single_line_statement_blocks	preserve	The style of single-line blocks in statements. Options: preserve, single, multi.
override_spacing	false	Print a space in the override attribute.
wrap_comments	false	Wrap comments when line_length is reached.
docs_style	preserve	Enforces the style of doc (natspec) comments. Options: preserve, line, block.
ignore	[]	Globs to ignore.
contract_new_lines	false	Add a new line at the start and end of contract declarations.
sort_imports	false	Sort import statements alphabetically in groups. A group is a set of imports separated by a newline.
pow_no_space	false	Suppress spaces around the power operator (**).

Check FormatterConfig for a more detailed explanation.

Inline Configuration

The formatter can be instructed to skip specific sections of code using inline comments. While the tool supports fine-grained control, it is generally more robust and efficient to disable formatting for entire AST items or statements.

This approach is preferred because it allows the formatter to treat the entire disabled item as a single, opaque unit. It can simply copy the original source text for that item’s span instead of partially formatting a line, switching to copy mode, and then resuming formatting. This leads to more predictable output and avoids potential edge cases with complex, partially-disabled statements.

Disable Line

These directives are best used when they apply to a complete, self-contained AST statement, as shown below. In this case, uint x = 100; is a full statement, making it a good candidate for a line-based disable.

To disable the next line:

// forgefmt: disable-next-line
uint x = 100;

To disable the current line:

uint x = 100; // forgefmt: disable-line

Disable Block

This is the recommended approach for complex, multi-line constructs where you want to preserve specific formatting. In the example below, the entire function definition is disabled. This is preferable to trying to disable individual lines within the signature, because lines like uint256 b /* a comment that goes inside the comma */, do not correspond to a complete AST item or statement on their own. Disabling the whole item is cleaner and more aligned with the code’s structure.

// forgefmt: disable-start
function fnWithManyArguments(
    uint a,
    uint256 b /* a comment that goes inside the comma */,
    uint256      c
) external returns (bool) {
// forgefmt: disable-end

Contributing

Check out the foundry contribution guide.

Guidelines for contributing to forge fmt:

Opening an issue

1. Create a short, concise title describing the issue.
   • Bad: Forge fmt does not work
   • Good: bug(forge-fmt): misplaces postfix comment on if-statement
2. Fill in the issue template fields, including Foundry version, platform, and component info.
3. Provide code snippets showing the current and expected behaviors.
4. If it’s a feature request, explain why the feature is needed.
5. Add the C-forge and Cmd-forge-fmt labels.

Fixing a Bug or Developing a Feature

1. Specify the issue being addressed in the PR description.
2. Add a note on your solution in the PR description.
3. Ensure the PR includes comprehensive acceptance tests under fmt/testdata/, covering:
   • The specific case being fixed/added.
   • Behavior with different kinds of comments (isolated, mixed, trailing).
   • If it’s a new config value, tests covering all available options.

Testing

Tests are located in the fmt/testdata folder. Each test consists of an original.sol file and one or more expected output files, named *.fmt.sol.

The default configuration can be overridden from within an expected output file by adding a comment in the format // config: {config_key} = {config_value}. For example:

// config: line_length = 160

The testing process for each test suite is as follows:

1. Read original.sol and the corresponding *.fmt.sol expected output.
2. Parse any // config: comments from the expected file to create a test-specific configuration.
3. Format original.sol and assert that the output matches the content of *.fmt.sol.
4. To ensure idempotency, format the content of *.fmt.sol again and assert that the output does not change.

Debug

The formatter includes a debug mode that provides visual insight into the pretty-printer’s decision-making process. This is invaluable for troubleshooting complex formatting issues and understanding how the boxes and breaks described in The Pretty Printer section work.

To enable it, run the formatter with the FMT_DEBUG environment variable set:

FMT_DEBUG=1 cargo test -p forge-fmt --test formatter Repros

When enabled, the output will be annotated with special characters representing the printer’s internal state:

• Boxes:

  • « and »: Mark the start and end of a consistent box (cbox).
  • ‹ and ›: Mark the start and end of an inconsistent box (ibox).

• Breaks:

  • ·: Represents a Break token, which could be a space or a newline.

For example, running debug mode on the HelloWorld contract from earlier would produce an output like this:

pragma solidity ^0.8.10;·
·
«‹«contract HelloWorld »{›·
‹‹    string· public· message››;·
·
«    constructor«(«‹‹string memory initMessage››»)» {»·
«‹        message = ·initMessage›·;·
»    }·
»}·
·
event Greet(««‹‹string indexed name››»»);·

This annotated output allows you to see exactly how the printer is grouping tokens and where it considers inserting a space or a newline. This makes it much easier to diagnose why a certain layout is being produced.

Modules

pp 🔒

Adapted from rustc_ast_pretty and prettyplease.

state 🔒

Structs

FormatterConfig

Contains the config and rule set

Enums

DiagnosticsResult

The result of the formatter.

DocCommentStyle

Style of doc comments

HexUnderscore

Style of underscores in hex literals

IndentStyle

Style of indent

IntTypes

Style of integer types.

MultilineFuncHeaderStyle

Style of function header in case it doesn’t fit

NumberUnderscore

Style of underscores in number literals

QuoteStyle

Style of string quotes

SingleLineBlockStyle

Style of single line blocks in statements

Constants

DEBUG 🔒

DEBUG_INDENT 🔒

Functions

diff 🔒

format

Format a string input with the default compiler.

format_ast

format_file

format_inner 🔒

format_once 🔒

format_source

parse_inline_config 🔒

Type Aliases

FormatterResult

The result of the formatter.