elvis_core is the core library for the elvis Erlang style reviewer. It is also used by rebar3_lint for easier integration into your Erlang libraries or applications.

It includes the mechanisms to apply rules to your Erlang code, as well as their implementation.

It implements pre-defined rules, but also supports user-defined ones.

The elvis command-line tool uses elvis_core extensively, so do check that project for a concrete example on how you could use it for your own purposes.

The rebar3_lint plugin eases integration of the style reviewer into your application or library. Be sure to check that out for further information.

After adding elvis_core as a dependency to your project and starting a shell, you will need to make sure the application is started:

1> {ok, _} = application:ensure_all_started(elvis_core).
{ok,[zipper,katana_code,elvis_core]}
2>

Once this is done you can apply the style rules in the following ways.

1> ElvisConfig = elvis_config:from_file("elvis.config").
<loaded_config>
2> elvis_core:rock(ElvisConfig).
# src/elvis_core.erl [OK]
# src/elvis_result.erl [OK]
# src/elvis_style.erl [OK]
# src/elvis_utils.erl [OK]
ok
3>

This will load the configuration, specified in file elvis.config, from the current directory. If no configuration is found {invalid_config, _} is thrown.

Another option for using elvis_core from the shell is to explicitly provide the configuration as an argument to elvis_core:rock/1:

1> ElvisConfig = [#{dirs => ["src"], filter => "*.erl", rules => []}].
[#{dirs => ["src"],filter => "*.erl",rules => []}]
2> elvis_core:rock(ElvisConfig).
Loading src/elvis_core.erl
# src/elvis_core.erl [OK]
Loading src/elvis_result.erl
# src/elvis_result.erl [OK]
Loading src/elvis_style.erl
# src/elvis_style.erl [OK]
Loading src/elvis_utils.erl
# src/elvis_utils.erl [OK]
ok
3>

We have only presented results where all files were well-behaved (i.e. they respect all the rules), so here's an example of how the output looks when files break some of the rules:

# ../../test/examples/fail_line_length.erl [FAIL]
  - line_length
    - Line 14 is too long: "    io:format(\"This line is 81 characters long and should be detected, yeah!!!\").".
    - Line 20 is too long: "    io:format(\"This line is 90 characters long and should be detected!!!!!!!!!!!!!!!!!!\").".
# ../../test/examples/fail_no_tabs.erl [FAIL]
  - no_tabs
    - Line 6 has a tab at column 0.
    - Line 15 has a tab at column 0.
# ../../test/examples/small.erl [OK]

An elvis.config file looks something like this:

[{elvis, [
    {config, [
        #{ dirs    => ["src"]
         , filter  => "*.erl"
         , ruleset => erl_files }
      , #{ dirs    => ["include"]
         , filter  => "*.hrl"
         , ruleset => hrl_files }
      , #{ dirs    => ["."]
         , filter  => "rebar.config"
         , ruleset => rebar_config }
      , #{ dirs    => ["."]
         , filter  => "elvis.config"
         , ruleset => elvis_config }
    ]}
    % output_format (optional): how to format the output.
    % Possible values are 'plain', 'colors' or 'parsable' (default='colors').
  , {output_format, colors}
    % verbose (optional): when 'true' more information will
    % be printed (default=false).
  , {verbose, true}
    % no_output (optional): when 'true' nothing will be printed
    % (default=false).
  , {no_output, false}
    % parallel: determine how many files will be
    % analyzed in parallel (default=1).
  , {parallel, 1}
]}].

The dirs key is a list that tells elvis_core where it should look for the files that match filter, which will be run through each of the pre-defined rules in the specified ruleset. filter can contain ** for further matching (it uses filelib:wildcard/1 under the hood).

If you want to override the pre-defined rules, for a given ruleset, you need to specify them in a rules key which is a list of items with the following structure {Module, Function, RuleConfig}, or {Module, Function} - if the rule takes no configuration values. You can also disable certain rules if you want to, by specifying them in the rules key and passing disable as a third parameter.

IMPORTANT: disable will only work if you also provided a ruleset as shown above.

Let's say you like your files to have a maximum of 90 characters per line and you also like to use tabs instead of spaces. In that case, you need to override erl_files's ruleset pre-defined rules as follows:

#{ dirs => ["src"]
 , filter => "*.erl"
 , rules => [
       {elvis_text_style, line_length, #{ limit => 90 }} % change line_length from 100 to 90
     , {elvis_text_style, no_tabs, disable} % disable no_tabs
   ]
 , ruleset => erl_files
 }.

You can also ignore modules at a check level or at a ruleset (group of checks) level:

• at a check level, you set the ignore option in the rule you want to ignore, e.g.:

{elvis_style, no_debug_call, #{ ignore => [elvis, elvis_utils] }}

(we are telling elvis to ignore the elvis and elvis_utils modules when executing the no_debug_call check.

• at a ruleset (group of checks) level, you set the ignore option for the group you want to ignore, e.g.:

#{ dirs => ["src"]
 , filter => "*.erl"
 , ruleset => erl_files
 , ignore => [module1, module4]
}.

With this configuration, none of the checks for erl_files is applied to module1 or module4.

Option output_format allows you to configure the output format. Possible values are colors, plain and parsable. The latter could be used for automated parsing and has a format very close to the one presented by dialyzer, like <file>:<line>:<rule>:<message>:

src/example.erl:1:god_modules:This module has too many functions (56). Consider breaking it into a number of modules.
src/example_a.erl:341:no_debug_call:Remove the debug call to io:format/2 on line 341.
src/example_a.erl:511:used_ignored_variable:Ignored variable is being used on line 511 and column 54.
src/example_a.erl:1252:used_ignored_variable:Ignored variable is being used on line 1252 and column 21.

The default value for the output_format option is colors.

It is possible to tell elvis_core to produce a more verbose output, using the verbose option. The value provided is a boolean, either true or false.

The default value for the verbose option is false.

On the other hand, if no output is desired then the value for the no_output option should be true.

The default value for the no_output option is false.

In order to speed up the analysis process, you can use the parallel option.

Its value indicates how many processes to use at the same time to apply the style rules to all the files gathered. The provided number should be less than or equal to the available cores, since any value higher than that won't report any speedup benefits.

The default value for parallel is 1.

You can find examples for configuration files in this project's config directory.

Options output_format, verbose, no_output, and parallel can also be set as application-level environment variables, i.e. as they would be found by application:get_env/2,3.

A reference to all pre-defined rules (and some other information) implemented in elvis_core can be found in this repository's RULES.md.

The implementation of a new rule is a function that takes 3 arguments in the following order:

1. t:elvis_config:config(): the value of option config as found in the configuration,
2. t:elvis_file:file(): the file to be analyzed,
3. t:erlang:map(): a configuration map specific to your user-defined rule.

This means you can define rules of your own (user-defined rules) as long as the functions that implement them respect this interface.

If you find any bugs or have other problems using this library, open an issue in this repository (or even a pull request 😃).

Inspired by HoundCI.