erb_lint
is a tool to help lint your ERB or HTML files using the included linters or by writing your own.
- Ruby 2.3.0+
- This is due to use of the safe navigation operator (
&.
) - This is also due to the use of the tilde-heredoc
<<~
syntax in some tests.
gem install erb_lint
...or add the following to your Gemfile
and run bundle install
:
gem 'erb_lint', require: false
Create a .erb_lint.yml
file in your project, with the following structure:
---
EnableDefaultLinters: true
linters:
ErbSafety:
enabled: true
better_html_config: .better-html.yml
Rubocop:
enabled: true
rubocop_config:
inherit_from:
- .rubocop.yml
See below for linter-specific configuration options.
This gem provides a command-line interface which can be run like so:
- Run
erb_lint [options]
if the gem is installed standalone. - Run
bundle exec erb_lint [options]
if the gem is installed as a Gemfile dependency for your app.
For example, erb_lint --lint-all --enable-all-linters
will run all available
linters on all ERB files in the current directory or its descendants (**/*.html{+*,}.erb
).
If you want to change the glob & exclude that is used, you can configure it by adding it to your config file as follows:
---
glob: "**/*.{html,text,js}{+*,}.erb"
exclude:
- '**/vendor/**/*'
- '**/node_modules/**/*'
linters:
ErbSafety:
enabled: true
better_html_config: .better-html.yml
Rubocop:
enabled: true
rubocop_config:
inherit_from:
- .rubocop.yml
Make sure to add **/
to exclude patterns; it matches the target files' absolute paths.
EnableDefaultLinters
: enables or disables default linters. Default linters are enabled by default.
You can disable a rule by placing a disable comment in the following format:
Comment on offending lines
<hr /> <%# erblint:disable SelfClosingTag %>
To raise an error when there is a useless disable comment, enable NoUnusedDisable
.
To disable inline comments and report all offenses, set --disable-inline-configs
option.
You can specify the exclude patterns both of global and lint-local.
---
exclude:
- '**/global-lib/**/*'
linters:
ErbSafety:
exclude:
- '**/local-lib/**/*'
Available Linters | Default | Description |
---|---|---|
AllowedScriptType | Yes | prevents the addition of <script> tags that have type attributes that are not in a white-list of allowed values |
ClosingErbTagIndent | Yes | |
CommentSyntax | Yes | detects bad ERB comment syntax |
ExtraNewline | Yes | |
FinalNewline | Yes | warns about missing newline at the end of a ERB template |
NoJavascriptTagHelper | Yes | prevents the usage of Rails' javascript_tag |
ParserErrors | Yes | |
PartialInstanceVariable | No | detects instance variables in partials |
RequireInputAutocomplete | Yes | warns about missing autocomplete attributes in input tags |
RightTrim | Yes | enforces trimming at the right of an ERB tag |
SelfClosingTag | Yes | enforces self closing tag styles for void elements |
SpaceAroundErbTag | Yes | enforces a single space after <% and before %> |
SpaceIndentation | Yes | |
SpaceInHtmlTag | Yes | |
TrailingWhitespace | Yes | |
DeprecatedClasses | No | warns about deprecated css classes |
ErbSafety | No | detects unsafe interpolation of ruby data into various javascript contexts and enforce usage of safe helpers like .to_json . |
Rubocop | No | runs RuboCop rules on ruby statements found in ERB templates |
RequireScriptNonce | No | warns about missing Content Security Policy nonces in script tags |
HardCodedString | No | warns if there is a visible hardcoded string in the DOM (does not check for a hardcoded string nested inside a JavaScript tag) |
DeprecatedClasses will find all classes used on HTML elements and report any classes that violate the rule set that you provide.
A rule_set
is specified as a list, each with a set of deprecated
classes and a corresponding suggestion
to use as an alternative.
Example configuration:
---
linters:
DeprecatedClasses:
enabled: true
exclude:
- 'app/views/shared/deprecated/**'
addendum: "See UX wiki for help."
rule_set:
- deprecated: ['badge[-_\w]*']
suggestion: "Use the ui_badge() component instead."
You can specify an addendum
to be added to the end of each violation.
The error message format is: "Deprecated class ... #{suggestion}"
or "Deprecated class ... #{suggestion} #{addendum}"
if an addendum
is present.
Linter-Specific Option | Description |
---|---|
rule_set |
A list of rules, each with a deprecated and suggestion option. |
deprecated |
A list of regular expressions which specify the classes deprecated by this rule. |
suggestion |
A string to be included in the rule's error message. Make this informative and specific to the rule that it is contained in. |
addendum |
A string to be included at the end of every error message of the rule set. (Optional) |
Files must have a final newline. This results in better diffs when adding lines to the file, since SCM systems such as git won't think that you touched the last line.
You can customize whether or not a final newline exists with the present
option.
Example configuration:
---
linters:
FinalNewline:
enabled: true
Linter-Specific Option | Description |
---|---|
present |
Whether a final newline should be present (default true) |
Runs the checks provided by better-html's erb safety test helper.
When using ERB interpolations in javascript contexts, this linter enforces the usage of safe helpers such as .to_json
.
See better-html's readme for more information.
Any ERB statement that does not call a safe helper is deemed unsafe and a violation is shown.
For example:
Not allowed ❌
<a onclick="alert(<%= some_data %>)">
Allowed ✅
<a onclick="alert(<%= some_data.to_json %>)">
Not allowed ❌
<script>var myData = <%= some_data %>;</script>
Allowed ✅
<script>var myData = <%= some_data.to_json %>;</script>
Example configuration:
---
linters:
ErbSafety:
enabled: true
better_html_config: .better-html.yml
Linter-Specific Option | Description |
---|---|
better_html_config |
Name of the configuration file to use for better-html . Optional. Valid options and their defaults are described in better-html's readme. |
Runs RuboCop on all ruby statements found in ERB templates. The RuboCop configuration that erb_lint
uses can inherit from
the configuration that the rest of your application uses. erb_lint
can be configured independently however, as it will often
be necessary to disable specific RuboCop rules that do not apply to ERB files.
Note: Each ruby statement (between ERB tags <% ... %>
) is parsed and analyzed independently of each other. Any rule that requires a broader context can trigger false positives (e.g. Lint/UselessAssignment
will complaint for an assignment even if used in a subsequent ERB tag).
Example configuration:
---
linters:
Rubocop:
enabled: true
rubocop_config:
inherit_from:
- .rubocop.yml
Layout/InitialIndentation:
Enabled: false
Layout/LineLength:
Enabled: false
Layout/TrailingEmptyLines:
Enabled: false
Layout/TrailingWhitespace:
Enabled: false
Naming/FileName:
Enabled: false
Style/FrozenStringLiteralComment:
Enabled: false
Lint/UselessAssignment:
Enabled: false
Rails/OutputSafety:
Enabled: false
The cops disabled in the example configuration above provide a good starting point.
Linter-Specific Option | Description |
---|---|
rubocop_config |
A valid rubocop configuration hash. Mandatory when this cop is enabled. See rubocop's manual entry on Configuration |
only |
Only run cops listed in this array instead of all cops. |
config_file_path |
A path to a valid rubocop configuration file. When this is provided, rubocop_config will be ignored. |
This linter prevents the usage of certain types of HTML <input>
without an autocomplete
argument: color
, date
, datetime-local
, email
, month
, number
, password
, range
, search
, tel
, text
, time
, url
, or week
.
The HTML autocomplete helps users to complete filling in forms by using data stored in the browser. This is particularly useful for people with motor disabilities or cognitive impairment who may have difficulties filling out forms online.
Bad ❌
<input type="email" ...>
Good ✅
<input type="email" autocomplete="nope" ...>
<input type="email" autocomplete="email" ...>
Trimming at the right of an ERB tag can be done with either =%>
or -%>
, this linter enforces one of these two styles.
Example configuration:
---
linters:
RightTrim:
enabled: true
enforced_style: '-'
Linter-Specific Option | Description |
---|---|
enforced_style |
Which style to enforce, can be either - or = . Optional. Defaults to - . |
Enforce a single space after <%
and before %>
in the ERB source.
This linter ignores opening ERB tags (<%
) that are followed by a newline,
and closing ERB tags (%>
) that are preceded by a newline.
Bad ❌
<%foo%>
<%=foo-%>
Good ✅
<% foo %>
<%
foo
%>
Example configuration:
---
linters:
SpaceAroundErbTag:
enabled: true
This linter prevents the usage of
Rails' javascript_tag
helper in ERB templates.
The html parser used in this gem knows not to look for html tags within certain other tags
like script
, style
, and others. The html parser does this to avoid confusing javascript
expressions like if (1<a || b>1)
for a malformed html tag. Using the javascript_tag
in
a ERB template prevents the parser from recognizing the change of parsing context and may
fail or produce erroneous output as a result.
Bad ❌
<%= javascript_tag(content, defer: true) %>
Good ✅
<script defer="true"><%== content %></script>
Bad ❌
<%= javascript_tag do %>
alert(1)
<% end %>
Good ✅
<script>
alert(1)
</script>
The autocorrection rule adds //<![CDATA[
and //]]>
markers to the existing script, as this is the default
behavior for javascript_tag
. This can be disabled by changing the correction_style
linter option
from cdata
to plain
.
Example configuration:
---
linters:
NoJavascriptTagHelper:
enabled: true
correction_style: 'plain'
Linter-Specific Option | Description |
---|---|
correction_style |
When configured with cdata , adds CDATA markers. When configured with plain , don't add makers. Defaults to cdata . |
This linter prevents the usage of HTML <script>
, Rails javascript_tag
, javascript_include_tag
and javascript_pack_tag
without a nonce
argument. The purpose of such a check is to ensure that when content securty policy is implemented in an application, there is a means of discovering tags that need to be updated with a nonce
argument to enable script execution at application runtime.
Bad ❌
<script>
alert(1)
</script>
Good ✅
<script nonce="<%= request.content_security_policy_nonce %>" >
alert(1)
</script>
Bad ❌
<%= javascript_tag do -%>
alert(1)
<% end -%>
Good ✅
<%= javascript_tag nonce: true do -%>
alert(1)
<% end -%>
Bad ❌
<%= javascript_include_tag "script" %>
Good ✅
<%= javascript_include_tag "script", nonce: true %>
Bad ❌
<%= javascript_pack_tag "script" %>
Good ✅
<%= javascript_pack_tag "script", nonce: true %>
This linter enforces self closing tag styles for void elements.
The void elements are area
, base
, br
, col
, embed
, hr
, img
, input
, keygen
, link
, menuitem
, meta
, param
, source
, track
, and wbr
.
If enforced_style
is set to always
(XHTML style):
Bad ❌
<link rel="stylesheet" type="text/css" href="styles.css">
Good ✅
<img src="someimage.png" alt="Some Image" />
If enforced_style
is set to never
(HTML5 style):
Bad ❌
<hr />
Good ✅
<meta charset="UTF-8">
Example configuration:
---
linters:
SelfClosingTag:
enabled: true
enforced_style: 'always'
Linter-Specific Option | Description |
---|---|
enforced_style |
If we should always or never expect self closing tags for void elements. Defaults to never . |
This linter prevent the addition of <script>
tags that have type
attributes that are not in a white-list of allowed values.
It is common practice for web developers to use <script>
tags with non-executable
type
attributes, such as application/json
or text/html
to pass arbitrary data into an html page.
Despite not being executable, these tags are subject to the same parsing quirks as executable script tags, and
it is therefore more difficult to prevent security issues from creeping in. Consider for instance an application
where it is possible to inject the string </script><script>
unescaped into a text/html
tag, the application
would be vulnerable to XSS.
This pattern can easily be replaced by <div>
tags with data
attributes that can just as easily be read
from javascript, and have the added benefit of being safer. When content_tag(:div)
or tag.div()
is used
to pass arbitrary user data into the html document, it becomes much harder to inadvertently introduce a
security issue.
It may also be desirable to avoid typos in type
attributes.
Bad ❌
<script type="text/javacsrïpt"></script>
Good ✅
<script type="text/javascript"></script>
By default, this linter allows the type
attribute to be omitted, as the behavior in browsers is to
consider <script>
to be the same as <script type="text/javascript">
. When the linter is configured with
allow_blank: false
, instances of <script>
tags without a type will be auto-corrected
to <script type="text/javascript">
.
It may also be desirable to disallow <script>
tags from appearing anywhere in your application.
For instance, Rails applications can benefit from serving static javascript code from the asset
pipeline, as well as other security benefits.
The disallow_inline_scripts: true
config option may be used for that purpose.
Example configuration:
---
linters:
AllowedScriptType:
enabled: true
allowed_types:
- 'application/json'
- 'text/javascript'
- 'text/html'
allow_blank: false
disallow_inline_scripts: false
Linter-Specific Option | Description |
---|---|
allowed_types |
An array of allowed types. Defaults to ["text/javascript"] . |
allow_blank |
True or false, depending on whether or not the type attribute may be omitted entirely from a <script> tag. Defaults to true . |
disallow_inline_scripts |
Do not allow inline <script> tags anywhere in ERB templates. Defaults to false . |
HardCodedString
warns if there is a visible hardcoded string in the DOM. It does not check for a hardcoded string nested inside a JavaScript tag.
Example configuration:
---
linters:
HardCodedString:
enabled: true
corrector:
path: path/to/corrector_file.rb
name: I18nCorrector
i18n_load_path: config/locales/en.yml
To leverage this linter's autocorrect feature, you must define the configuration for its corrector
, such that its path, class name, and i18n_load_path
are resolvable.
Linter-Specific Option | Description |
---|---|
path | a string pointing to the path to the corrector file |
name | the name of the corrector class (can either be I18nCorrector or Rubocop::I18nCorrector ) |
i18n_load_path | a string pointing to the path of the file(s) to be translated |
Below is an example corrector file. For your project, the actual details of the autocorrect
method are left up to how you want to correct those offenses.
class I18nCorrector
attr_reader :node
def initialize(node, filename, i18n_load_path, range)
end
def autocorrect(tag_start:, tag_end:)
->(corrector) do
node
end
end
end
This linter enforces the use of the correct ERB comment syntax, since Ruby comments (<% # comment %>
with a space) are not technically valid ERB comments.
Bad ❌
<% # This is a Ruby comment %>
Good ✅
<%# This is an ERB comment %>
Bad ❌
<% # This is a Ruby comment; it can fail to parse. %>
Good ✅
<%# This is an ERB comment; it is parsed correctly. %>
Good ✅
<%
# This is a multi-line ERB comment.
%>
erb_lint
allows you to create custom linters specific to your project. It will load linters from the .erb_linters
directory in the root of your
repository. See the linters directory for examples of how to write linters.
# .erb_linters/custom_linter.rb
module ERBLint
module Linters
class CustomLinter < Linter
include LinterRegistry
class ConfigSchema < LinterConfig
property :custom_message, accepts: String
end
self.config_schema = ConfigSchema
def run(processed_source)
unless processed_source.file_content.include?('this file is fine')
add_offense(
processed_source.to_source_range(0 ... processed_source.file_content.size),
"This file isn't fine. #{@config.custom_message}"
)
end
end
end
end
end
By default, this linter would be disabled. You can enable it by adding an entry to .erb_lint.yml
:
---
linters:
CustomLinter:
enabled: true
custom_message: We suggest you change this file.
Test your linter by running erb_lint
's command-line interface:
bundle exec erb_lint --enable-linters custom_linter --lint-all
Running this on a random project might yield this output:
Linting 15 files with 1 linters...
This file isn't fine. We suggest you change this file.
In file: app/views/layouts/application.html.erb:1
Errors were found in ERB files
To write a linter that can autocorrect offenses it detects, simply add an
autocorrect
method that returns a callable. The callable is called with an instance of
RuboCop::Cop::Corrector
as argument, and therefore erb_lint correctors work exactly as RuboCop correctors do.
def autocorrect(_processed_source, offense)
lambda do |corrector|
corrector.insert_after(offense.source_range, "this file is fine")
end
end
You can change the output format of ERB Lint by specifying formatters with the -f/--format
option.
$ erb_lint
Linting 8 files with 12 linters...
Remove multiple trailing newline at the end of the file.
In file: app/views/users/show.html.erb:95
Remove newline before `%>` to match start of tag.
In file: app/views/subscriptions/index.html.erb:38
2 error(s) were found in ERB files
erb_lint --format compact
Linting 8 files with 12 linters...
app/views/users/show.html.erb:95:0: Remove multiple trailing newline at the end of the file.
app/views/users/_graph.html.erb:27:37: Extra space detected where there should be no space
2 error(s) were found in ERB files
erb_lint --format junit
<?xml version="1.0" encoding="UTF-8"?>
<testsuite name="erb_lint" tests="2" failures="2">
<properties>
<property name="erb_lint_version" value="%{erb_lint_version}"/>
<property name="ruby_engine" value="%{ruby_engine}"/>
<property name="ruby_version" value="%{ruby_version}"/>
<property name="ruby_patchlevel" value="%{ruby_patchlevel}"/>
<property name="ruby_platform" value="%{ruby_platform}"/>
</properties>
<testcase name="app/views/subscriptions/_loader.html.erb" file="app/views/subscriptions/_loader.html.erb" lineno="1">
<failure message="SpaceInHtmlTag: Extra space detected where there should be no space." type="SpaceInHtmlTag">
<![CDATA[SpaceInHtmlTag: Extra space detected where there should be no space. at app/views/subscriptions/_loader.html.erb:1:7]]>
</failure>
</testcase>
<testcase name="app/views/application/index.html.erb" file="app/views/subscriptions/_menu.html.erb"/>
</testsuite>
Used by GitLab Code Quality.
[
{
"description":"Extra space detected where there should be no space.",
"check_name":"SpaceInHtmlTag",
"fingerprint":"5a259c7cafa2c9ca229dfd7d21536698",
"severity":"info",
"location":{
"path":"app/views/subscriptions/_loader.html.erb",
"lines":{
"begin":1,
"end":1
}
}
},
{
"description":"Remove newline before `%\u003e` to match start of tag.",
"check_name":"ClosingErbTagIndent",
"fingerprint":"60b4ed2120c7abeebebb43fba4a19559",
"severity":"warning",
"location":{
"path":"app/views/subscriptions/_loader.html.erb",
"lines":{
"begin":52,
"end":54
}
}
}
]
The cache is currently opt-in - to turn it on, use the --cache
option:
erb_lint --cache ./app
Cache mode is on
Linting 413 files with 15 linters...
File names pruned from the cache will be logged
No errors were found in ERB files
Cached lint results are stored in the .erb_lint_cache
directory by default, though a custom directory can be provided
via the --cache-dir
option. Cache filenames are computed with a hash of information about the file and erb_lint
settings.
These files store instance attributes of the CachedOffense
object, which only contain the Offense
attributes
necessary to restore the results of running erb_lint
for output. The cache also automatically prunes outdated files each time it's run.
You can also use the --clear-cache
option to delete the cache file directory.
This project is released under the MIT license.