• Stars
    star
    110
  • Rank 316,770 (Top 7 %)
  • Language
    Erlang
  • License
    MIT License
  • Created about 5 years ago
  • Updated 25 days ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Erlang Formatter for Rebar3

rebar3_format

Build Status Hex pm

A rebar plugin for code formatting.

Build

$ rebar3 compile

Use

Note About OTP25+

As #314 and #315 show, if you use OTP25 or newer, you need to enable all features in your Erlang VM when you're running the formatter. This is because rebar3_format uses katana_code and, to be able to parse code containing new features, katana_code is compiled with those features enabled. So, even if you don't use the new features yourself, you need them enabled when running the formatter on your code. To achieve that, you can set this variable on your development/CI environment:

ERL_AFLAGS="-enable-feature all"

Format your Code

Add the plugin to your rebar config:

{project_plugins, [rebar3_format]}

Then just call your plugin directly in an existing application:

$ rebar3 format

By default, this will format every Erlang file (including hrls, erls and app.srcs) under src, include and test, together with every .config file, in your root directory and any nested Erlang apps' directories within your project. You can specify the directory/file to format as following:

$ rebar3 format --files 'src/my_subdir/*.erl'
$ rebar3 format --files src/other_subdir/my_file.erl
$ rebar3 format --files 'test/**/*.erl' --files 'include/*.hrl'

To save the formatted files in a different directory you have to pass it as a parameter:

$ rebar3 format --output formatted/

Configuration

The plugin supports the following configuration options in the format section of rebar.config:

  • formatter (module()):
    • This is the module that will dictate the style in which all the code will be formatted. It must implement the rebar3_formatter behavior. This project itself provides 2 formatters:
      • otp_formatter: Based on the default formatter that comes with Erlang/OTP (erl_prettypr), we only fixed some bugs but then respected the original format it produced in its entirety. This formatter only recognizes 2 options:
        • paper(pos_integer()):
          • Specifies the preferred maximum number of characters on any line, including indentation.
          • The default value is 80.
        • ribbon(pos_integer()):
          • Specifies the preferred maximum number of characters on any line, not counting indentation.
          • The default value is 56.
      • default_formatter: Our own default formatter, defining our personal criteria for how to stylize Erlang code. It admits all the options listed below.
    • The default value is default_formatter.
  • options (#{atom() => term()}):
    • A map with a list of options that should be interpreted by the chosen formatter. The available keys are:
      • encoding(none | epp:source_encoding()):
        • Encoding to use when writing files.
        • The default value is none.
      • paper(pos_integer()):
        • Specifies the preferred maximum number of characters on any line, including indentation.
        • The default value is 100.
      • ribbon(pos_integer()):
        • Specifies the preferred maximum number of characters on any line, not counting indentation.
        • The default value is 90.
      • break_indent(pos_integer()):
        • Specifies the preferred number of characters to use to indent a line that "breaks" from the previous one (for instance, a clause body after a clause head).
        • The default value is 4.
      • sub_indent(pos_integer()):
        • Specifies the preferred number of characters to use to indent a guard that "follows" a longer-than-paper clause head.
        • This value only affects the line that starts with when
        • The default value is the value of break_indent.
      • unquote_atoms (boolean()):
        • Specifies whether the formatter should remove quotes from atoms that don't need them (e.g. 'this_one') or not.
        • The default value is true, i.e. the formatter won't preserve your quotes if they're not needed, unless you explicitely ask for.
      • truncate_strings (boolean()):
        • Specifies whether the formater should accomodate long strings into multiple lines using Erlang's consecutive string syntax. This is what the otp_formatter does, turning a long string into multiple same-length strings one per row to respect paper and ribbon.
        • The default value is false, i.e. the formatter will keep the strings as it finds them.
      • inline_attributes (inlining()):
        • Specifies the desired behavior for inlining attributes with lists, like -export, -export_type and -optional_callbacks.
        • See inlining() type definition below for a list of options.
        • The default value is all, i.e. always put as many functions/types on each row as possible.
      • inline_fields (inlining()):
        • Specifies the desired behavior for inlining lists of map and record fields.
        • See inlining() type definition below for a list of options.
        • The default value is {when_under, 3}, i.e. always place each field in its own line, unless there is only 1 or 2 of them.
      • inline_items (inlining()):
        • Specifies the desired behavior when the formatter needs to use multiple lines for a multi-item structure (i.e. tuple, list, map, etc.).
        • NOTE: If the formatter can put all items in the same row, it will do it, regardless of this configuration. This is to prevent short lists, binaries or tuples to be spread out in multiple lines.
        • See inlining() type definition below for a list of options.
        • The default value is {when_over, 25} to properly accommodate large binaries or lists.
      • inline_simple_funs (boolean()):
        • Specifies if anonymous function bodies should be placed in the same line as the function clause head in case for anonymous functions with just one clause if paper and ribbon allows it or if these simple funs should be indented as all the others.
        • The default value is true.
      • inline_qualified_function_composition (boolean()):
        • Specifies if composed qualified function calls (e.g. module1:function1(module2:function2(...) should stay in the same line if they fit or if the formatter should always put the internal function call in the next line.
        • Because of how OTP's prettypr is built (which is the tool we're using to finally print the formatted code) we can't indent these function calls only if it doesn't fit in a line, at least not without adding an extra space to the right of ( for all function applications. That's why this switch is all-or-none. You can a more detailed explanation of this behaviour here.
        • The non-inlining only applies when both functions that are composed are fully qualified (we're using fully-qualified as a proxy for has a long name); e.g. in this case d(f:f(g:g(h(…)))) the formatter will always write g:g(...) in the next row, but not h(...) nor f:f(...) will be moved to a new row.
        • The default value is false.
      • inline_clause_bodies (boolean()):
        • Specifies if clause bodies (for case, function, etc. statements) should be placed in the same line as the clause heads if paper and ribbon allows it or if all bodies should be placed in the next line after their clause heads.
        • The default value is false.
      • inline_expressions (boolean()):
        • Specifies if sequential expressions in a clause should be placed in the same line if paper and ribbon allows it or if each expression should be placed in its own line.
        • The default value is false.
      • parenthesize_infix_operations (boolean()):
        • Specifies if parentheses should be added around composed infix operations to avoid confusion around precedence.
        • The default value is false.
      • spaces_around_arguments (boolean()):
        • Specifies if an expression such as a_function:call("with", "arguments") should be formatted as a_function:call( "with", "argments" ).
        • This parameter doesn't apply to every expression with arguments (e.g. attributes, function types, etc.). It only applies to function calls (i.e. applications).
        • Although this configuration doesn't override the value of inline_qualified_function_composition, we strongly recommend you to use inline_qualified_function_composition => true if you use spaces_within_parentheses => true.
        • The default value is false.
      • spaces_around_fields (boolean()):
        • Specifies if an expression such as #{a => map, "with" => "fields"} should be formatted as #{ a => map, "with" => "fields" }.
        • This parameter works on map and record expressions and types, but it doesn't affect tuples.
        • The default value is false.
      • preserve_empty_lines (boolean()):
        • Specifies if blank lines between statements should be preserved when formatting.
        • Keep in mind that blank lines between clauses, between items in tuples, lists, etc, between attributes, and so on will not be affected by this configuration and therefore they'll be unconditionally removed.
        • This option is only used when inline_expressions is false.
        • If this option is true, one empty line will preserved for each group of empty lines that are placed between expressions in a clause.
        • The default value is true.
      • parse_macro_definitions (boolean()):
        • ktn_dodger (the module we use to parse the code) doesn't parse macro definitions by default. That's to prevent removing parentheses where they're actually meaningful in the context where the macro is used, but not in the context where it's defined.
        • With this option in true, the formatter will instruct ktn_dodger to actually parse the macros.
        • The default value is true.
        • The idea is for users to turn it to false only for the module that contain macros that would be broken otherwise.
      • sort_arity_qualifiers (true | false)
        • Specifies if the arity qualifiers should be sorted alphabetically (note that if two functions share the name, they will be sorted by arity).
        • Currently, the formatter only supports sorting of arity qualifiers in module attributes that contain pure lists of said elements. These attributes are:
          • -export
          • -export_type
          • -optional_callback
        • The default value is false.
  • files ([file:filename_all()]):
    • List of wildcard patterns representing the files that will be formatted by default (i.e. when not using --files on command line).
    • The default value is ["src/**/*.?rl"]
  • ignore ([file:filename_all()]):
    • List of wildcard patterns representing the files that the formatter will ignore when formatting.
    • Note that it will ignore the files set for formatting either with the files option or using --files in the command line if they match one of the given wildcards.
    • You can also ignore a specific file adding the attribute -format(ignore). in it or the comment % @format ignore.

Configuration Types

  • inlining() :: all | none | {when_over, pos_integer()} | {when_under, pos_integer()}:
    • Options that allow this type of configuration will work this way:
      • When this option is all, the formatter will try to fit as many items in each line as permitted by paper and ribbon.
      • When the flag is none, the formatter will place each item in its own line.
      • When the flag is {when_over, N} the formatter will work as none for lists with up to N elements, and it will inline longer lists.
      • When the flag is {when_under, N} the formatter will work as none for lists with more than N elements, and it will inline shorter lists.

Per-File Configuration

You can tweak any of the formatter options for a particular file, using the format attribute in it, like this:

-format(#{paper => 80}).

You can also achieve the same effect using @format in a comment, like this:

% @format #{paper => 80}.

We're very strict with the parsing of comments, tho. You can use multiple % signs if you want, but you have to place the @ sign exactly one space after the last % and you have to place the whole map with options in a single line (although you can have multiple @format comments per file) that has to end in period (.).

Test

To test the plugin just run rebar3 test. It will essentially run rebar3 format inside test_app. Add modules with any "tricky" formatting you want to test_app/src, and push them to github including the after results. The after results can be tought as the expected output behaviour.


Proposed Workflow

When we created this tool, we envisioned a workflow for teams where each member can use their preferred style for code formatting. The idea is to take advantage of rebar3 profiles and write the following on your rebar.config file:

%% The canonical format used when pushing code to the central repository
{format, [
    {files, ["src/*.erl", "include/*.hrl", "test/*.erl"]},
    {formatter, default_formatter},
    {options, #{paper => 100}}
]}.
{profiles, [
    {brujo, [
        {format, [
            {files, ["src/*.erl", "include/*.hrl", "test/*.erl"]},
            {formatter, rok_formatter}, % I prefer comma-first formatting
            {options, #{paper => 100}}
        ]}
    ]},
    {miriam, [
        {format, [
            {files, ["src/*.erl", "include/*.hrl", "test/*.erl"]},
            {formatter, default_formatter},
            {options, #{
                inline_clause_bodies => false, % she doesn't like one-liners
                inline_simple_funs => false, % and she's adamant about it
                inline_items => all % but she does like long lists of items
            }}
        ]}
    ]}
]}

Then whenever you're about to work on something, follow this ritual:

git checkout main
git checkout -b my-branch
rebar3 as brujo format
# Work on your code...
rebar3 format # This can be a git hook for commits
git commit -am "Apply my changes"
git push origin my-branch --set-upstream

Other developers do the same but using as $THEIR_NAME instead of as brujo.

That way each developer can read code in the way they understand it better, write code exactly how they like to write it, etc. Then push it to the central repository in a consistent way that matches the style of the rest of the project.


Using External Formatters

Through rebar3 format, you can use other formatters that are not included in this repository. That way you can follow our proposed workflow and allow each developer to format the code with their favorite formatter using rebar3 plugins while still maintaining an unique canonical formatter when pushing to your central git repository. You also get -format attribute compliance (including -format ignore.) for free, since they're respected when using any formatter.

Steamroller

If you want to use @old-reliable's steamroller, you just need to add the following things to your rebar.config file:

{project_plugins, [rebar3_format, steamroller]}.

{format, [
{files, ["src/*.erl", "include/*.hrl"]},
{ignore, ["src/*_ignore.erl", "src/ignored_file_config.erl"]},
{formatter, sr_formatter}, %% The steamroller formatter.
{options, #{line_length => 80}}
]}.

erlfmt

If you want to use @whatsapp's erlfmt, you just need to add the following things to your rebar.config file:

{project_plugins, [rebar3_format, erlfmt]}.
{format, [
    {files, ["src/*.erl", "include/*.hrl"]},
    {ignore, ["src/*_ignore.erl", "src/ignored_file_config.erl"]},
    {formatter, erlfmt_formatter}, %% The erlfmt formatter interface.
    {options, #{print_width => 100, ignore_pragma => true}} %% ...or no options at all.
]}.

Compatibility Note

erlfmt_formatter is compatible with version v0.7.0 and v0.8.0 of erlfmt, which are currently available at hex.pm.

Implementing your own Formatter

To create a new formatter, you need to implement the rebar3_formatter behaviour. It defines just one callback:

-callback format(file:filename_all(), opts()) -> result().

That means you need to write a function that receives a filename and a map with options (some of them are specified in the rebar3_formatter module, but you can add as many others as you want) and returns a result (either changed or unchanged). It's expected for your formatter to honor the predefined options as described below:

  • output_dir:
    • none: Don't produce any output.
    • current: Replace files when formatting.
    • file:filename_all(): Drop files in this folder, preserving their current names.
  • encoding:
    • none: Preserve/guess original encoding of files.
    • epp:source_encoding(): Use this encoding when parsing files.
  • action:
    • verify: Only return the result without actually modifying any files.
    • format: Do format the files.

It's a good practice, although not enforced by the formatter itself to respect -format attributes in files as the formatters provided in this repo do. To remove the need for parsing and writing files, you can use the rebar3_ast_formatter module/behaviour as default_formatter and otp_formatter do.

Editor Integration

Visual Studio Code

You can use rebar3_format from Visual Studio Code with the Erlang Formatter extension.

Helpers

Git commit hooks

In the scripts folder you'll find two scripts that work really well as pre and post commit git hooks, in case you want to slowly format your huge repos with a myriad of modules :)

Contribute

To contribute to rebar3_format, please refer to CONTRIBUTING.

More Repositories

1

hologram

Easy, painless AWS credentials on developer laptops.
Go
802
star
2

goamz

Fork of the GOAMZ version developed within Canonical with additional functionality with DynamoDB
Go
401
star
3

rustenstein

Wolfenstein 3D port written in Rust
Rust
229
star
4

baker

Baker is a high performance, composable and extendable data-processing pipeline for the big data era
Go
169
star
5

batchiepatchie

Go
147
star
6

AirControl

Control AirPlay Mirroring from the command line.
Shell
94
star
7

cantor

Cantor provides utilities for estimating the cardinality of large sets.
Java
83
star
8

rebar3_hank

The Erlang Dead Code Cleaner
Erlang
68
star
9

mero

scalable and lightweight OTP Erlang client for memcached
Erlang
48
star
10

backbone.d3

Reusable D3 visualizations using Backbone.js views
JavaScript
46
star
11

exmld

elixir interface to kinesis client library via MultiLangDaemon
Elixir
44
star
12

habanero

Load Testing framework in erlang with real-time feedback
JavaScript
43
star
13

gulp-react-docs

Gulp plugin for generating Markdown documentation for React components
JavaScript
41
star
14

erlmld

erlang interface to kinesis client library via MultiLangDaemon
Erlang
39
star
15

kinetic

Erlang Kinesis Client
Erlang
38
star
16

meandro

Finds dead code in Elixir applications
Elixir
34
star
17

spillway

An Erlang OTP application for load shedding
Erlang
27
star
18

rebar3_depup

Dependency Updater for rebar3-based Erlang Projects
Erlang
21
star
19

python-hll

python-hll
Python
18
star
20

privacy

Collection of privacy specs and discussion
16
star
21

rebar3_typer

Rebar3 plugin wrapper around TypEr
Erlang
12
star
22

slackask

Ask people anonymous questions on Slack.
Python
10
star
23

erliam

erlang library for caching credentials and signing AWS API requests
Erlang
10
star
24

lbfgs-d

A native D implementation of Limited Memory Broyden–Fletcher–Goldfarb–Shanno optimizer
D
9
star
25

plunger

A linter utility for Luigi jobs
Haskell
8
star
26

getbox

Playground EC2 instance manager
Python
8
star
27

canary

Dashboard publishing for Erlang's folsom metrics project.
Erlang
8
star
28

ecs-ship

Ship your code to your ECS infraestructure with ecs-ship. Unless it's Friday in which case, don't ship it.
Go
4
star
29

lambda-examples

Shell
4
star
30

fledge_lab

A small dockerized lab to learn and experiment with FLEDGE.
Python
3
star
31

TERN

2
star
32

terraform-container-datadog

Terraform module to generate well-formed JSON documents (container definitions) that are passed to the aws_ecs_task_definition Terraform resource for datadog
HCL
1
star
33

zendesk-community-post-to-slack

Python
1
star
34

terraform-container-fluentbit

Terraform module to generate well-formed JSON documents (container definitions) that are passed to the aws_ecs_task_definition Terraform resource for fluentbit
HCL
1
star