EditorConfig Specification ¶

This is version 0.17.2 of this specification.

Introduction (informative)¶

All content in this document is normative unless marked “(informative)”.

EditorConfig helps maintain consistent coding styles for multiple developers working on the same project across various editors and IDEs. The EditorConfig project consists of a file format for defining coding styles and a collection of text editor plugins that enable editors to read the file format and adhere to defined styles. EditorConfig files are easily readable and they work nicely with version control systems.

Terminology ¶

Changed in version 0.15.1.

In EditorConfig:

“EditorConfig files” (named .editorconfig in lowercase) include section(s) storing key-value pairs. EditorConfig files must conform to this specification.
“Cores” parse files conforming to this specification, and provide key-value pairs to plugins.
“Plugins” receive key-value pairs from cores and update an editor’s settings based on the key-value pairs.
“Editors” permit editing files, and use plugins to update settings for files being edited.

A conforming core or plugin must pass the tests in the core-tests repository or plugin-tests repository, respectively.

(informative) Some plugins include or bundle their own cores, and some rely on external cores. Some editors include or bundle plugin or core functionality. Editors, plugins, and cores may all come from different people. Those people may or may not have any direct interaction with the EditorConfig organization.

File Format ¶

Changed in version 0.17.2.

EditorConfig files are in an INI-like file format. To read an EditorConfig file, take one line at a time, from beginning to end. For each line:

Remove all leading and trailing whitespace.
Process the remaining text as specified for its type below.

The types of lines are:

Blank: Contains nothing. Blank lines are ignored.
Comment: starts with a ; or a #. Comment lines are ignored.
Section Header: starts with a [ and ends with a ]. These lines define globs; see Glob Expressions.
- May contain any characters between the square brackets (e.g., [ and ] and even spaces and tabs are allowed).
- Forward slashes (/) are used as path separators.
- Backslashes (\\) are not allowed as path separators (even on Windows).
Key-Value Pair (or Pair): contains a key and a value, separated by an =. See Supported Pairs.
- Key: The part before the first = on the line.
- Value: The part, if any, after the first = on the line.
- Keys and values are trimmed of leading and trailing whitespace, but include any whitespace that is between non-whitespace characters.
- If a value is not provided, then the value is an empty string (i.e., "" in C or Python).

Any line that is not one of the above is invalid.

EditorConfig files must be UTF-8 encoded, with LF or CRLF line separators.

No inline comments ¶

Changed in version 0.15.0.

A ; or # anywhere other than at the beginning of a line does not start a comment, but is part of the text of that line. For example:

[*.txt]
foo = editorconfig ;)

gives variable foo the value editorconfig ;) in *.txt files, not the value editorconfig.

This specification does not define any “escaping” mechanism for ; or # characters.

Compatibility

The EditorConfig file format formerly allowed the use of ; and # after the beginning of the line to mark the rest of a line as comment. This led to confusion how to parse values containing those characters. Old EditorConfig parsers may still allow inline comments.

Parts of an EditorConfig file ¶

The parts of an EditorConfig file are:

Preamble: the lines that precedes the first section. The preamble is optional and may contain key-value pairs, comments and blank lines.
Section Name: the string between the beginning [ and the ending ].
Section: the lines starting from a Section Header until the beginning of the next Section Header or the end of the file.

Glob Expressions ¶

Section names in EditorConfig files are filepath globs, similar to the format accepted by .gitignore. They support pattern matching through Unix shell-style wildcards. These filepath globs recognize the following as special characters for wildcard matching:

Special Characters	Matching
`*`	any string of characters, except path separators (`/`)
`**`	any string of characters
`?`	any single character, except path separators (`/`)
`[seq]`	any single character in seq. Any character inside those brackets is considered literally. It means, that pattern `[abc{1..2}]` is considered literally: either ‘a’, or ‘b’, or ‘’, or ‘c’, or ‘{’, or ‘1’, or ‘.’, or ‘2’, or ‘}’.
`[!seq]`	any single character not in seq. Any character inside those brackets is considered literally as well (see example above).
`{s1,s2,s3}`	any of the strings given (separated by commas, can be nested) (But `{s1}` only matches `{s1}` literally.)
`{num1..num2}`	any integer numbers between `num1` and `num2`, where `num1` and `num2` can be either positive or negative. `num1` is required to be less than `num2`. For instance, `{1..3}`, `{-1..4}` are valid, but `{-4..-5}`, `{3..1}` are not.

If the glob contains a path separator (a / not inside square brackets), then the glob is relative to the directory level of the particular .editorconfig file itself. Otherwise the pattern may also match at any level below the .editorconfig level. For example, *.c matches any file that ends with .c in the directory of .editorconfig or any other directory below one that stores this .editorconfig. However, the glob subdir/*.c only matches files that end with .c in the subdir directory in the directory of .editorconfig.

Therefore, a leading slash is not relevant if there is already a slash in the middle of the pattern. Thus, the globs /subdir/*.c and subdir/*.c must yield the same result.

As a corollary, a section name ending with / does not match any file.

The backslash character (\\) can be used to escape a character so it is not interpreted as a special character.

Cores must accept section names with length up to and including 1024 characters. Beyond that, each implementation may choose to define its own upper limit or no explicit upper limit at all.

File Processing ¶

When a filename is given to EditorConfig a search is performed in the directory of the given file and all parent directories for an EditorConfig file. An EditorConfig file is named “.editorconfig”, all lowercased. Non-existing directories are treated as if they exist and are empty. All found EditorConfig files are searched for sections with section names matching the given filename. The search shall stop if an EditorConfig file is found with the root key set to true in the preamble or when reaching the root filesystem directory.

Files are read top to bottom and the most recent pairs found take precedence. Thus, in case a given file matches multiple sections within a single .editorconfig file, the pairs defined in the section that comes later in the .editorconfig file take precedence over pairs defined in the section that comes earlier in the same .editorconfig file. If multiple EditorConfig files have matching sections, the pairs from the closer EditorConfig file are read last, so pairs in closer files take precedence.

Capitalization of the File Name ¶

As noted above, the .editorconfig filename should be lowercased. On some platforms, opening a file with a different capitalization results in opening the same file with lowercased file names. On such a platform, in addition to the all lowercased .editorconfig file name, a Core may choose to also accept files with a different capitalization as if it were all lowercased.

(informative) Such platform is common with a case-insensitive filesystem. For example, a file named .editorConfig exists in the filesystem, but opening a file named .editorconfig via a file-opening API still opens the differently capitalized .editorConfig file. The behavior of the Core as described in the previous paragraph is to prevent the need of the additional operation of specifically retrieving the filename, which can be relatively expensive in the context of EditorConfig.

Supported Pairs ¶

Changed in version 0.17.1.

EditorConfig file sections contain key-value pairs separated by an equal sign (=). With the exception of the root key, all pairs MUST be located under a section to take effect.

EditorConfig cores shall accept and report all syntactically valid key-value pairs, even if the key is not defined in this specification.
EditorConfig plugins shall ignore unrecognized keys and invalid/unsupported values.

Here is the list of all keys defined by this version of this specification, and the supported values associated with them:

Key	Supported values
`indent_style`	Set to `tab` or `space` to use tabs or spaces for indentation, respectively. Option `tab` implies that an indentation is to be filled by as many hard tabs as possible, with the rest of the indentation filled by spaces. A non-normative explanation can be found in the indentation_ section. The values are case-insensitive.
`indent_size`	Set to a whole number defining the number of columns used for each indentation level and the width of soft tabs (when supported). If this equals `tab`, the `indent_size` shall be set to the tab size, which should be `tab_width` (if specified); else, the tab size set by the editor. The values are case-insensitive.
`tab_width`	Set to a whole number defining the number of columns used to represent a tab character. This defaults to the value of `indent_size` and should not usually need to be specified.
`end_of_line`	Set to `lf`, `cr`, or `crlf` to control how line breaks are represented. The values are case-insensitive.
`charset`	Set to `latin1`, `utf-8`, `utf-8-bom`, `utf-16be` or `utf-16le` to control the character set. Use of `utf-8-bom` is discouraged.
`spelling_language`	Sets the natural language that should be used for spell checking. Only one language can be specified. There is no default value. The format is `ss` or `ss-TT`, where `ss` is an ISO 639 two-letter language code and `TT` is an ISO 3166 two-letter territory identifier. (Therefore `spelling_language` must be either two or five characters long.) Note: This property does not specify the charset to be used. The charset is in separate property `charset`.
`trim_trailing_whitespace`	Set to `true` to remove all whitespace characters preceding newline characters in the file and `false` to ensure it doesn’t.
`insert_final_newline`	Set to `true` ensure file ends with a newline when saving and `false` to ensure it doesn’t. Editors must not insert newlines in empty files when saving those files, even if `insert_final_newline = true`.
`root`	Must be specified in the preamble. Set to `true` to tell the core not to check any higher directory for EditorConfig settings for on the current filename. The value is case-insensitive.

For any pair, a value of unset removes the effect of that pair, even if it has been set before. For example, add indent_size = unset to undefine the indent_size pair (and use editor defaults).

Pair keys are case-insensitive. All keys are lowercased after parsing.

Cores must accept keys and values with lengths up to and including 1024 and 4096 characters respectively. Beyond that, each implementation may choose to define its own upper limits or no explicit upper limits at all.

Indentation (Non-Normative)¶

The indentation related options (indent_style, indent_size and tab_width) require a special documentation section to specify their behavior. Consider the following code snippet:

def execute():
    source = "indentation is important"
    for i in source.split(" "):
        print(i)

The indent_size setting for this code snippet equals 4, because indent_size means how many columns are required to indent the next line in relation to previous (if indentation, of course, is applicable for this line). Then the next question is how this indentation of 4 columns is achieved. It may be 4 consequent spaces/soft tabs, a single tab with width equal to 4, or two tabs with width equal to 2.

This is when indent_style comes into picture. It specifies what character should be used whenever possible in order to achieve the indentation size specified in indent_size. To fully understand what “whenever possible” actually means, let’s consider the following EditorConfig file living in the same directory as the file above:

root = true
[example_file.py]
indent_style = tab
indent_size = 4
tab_width = 3

The indent_size of 4 is not achievable by placing 1 or 2 consequent tabs, because tab_width = 3. Therefore, in order to comply with this EditorConfig configuration, the new lines (where indentation is applicable) must be precisely indented with one tab, and one space. That is because by placing one tab we’re not achieving the indent_size required, but by placing the 2 consequent tabs we’re overreaching. Therefore, although indent_style is tab, we still have to supplement with one space character to fulfill the requirement.

For another example, if we have the following EditorConfig file:

root = true
[another_file.py]
indent_style = tab
indent_size = 8
tab_width = 4

One MUST expect that spaces will not be used at all for indentation, since all the indentation can be achieved via tabs only.

Additionally, it is possible to have indent_size less than the tab_width.

root = true
[another_file.py]
indent_style = tab
indent_size = 4
tab_width = 8

To understand the way it works, let’s look at the following example:

def func():
    if True:
        return True

In this case, the line where the if statement condition is specified is indented with 4 spaces, because the indent_size = 4 and the tab cannot fit in. On the other hand, the line with return statement must be indented with one tab, because the indentation level for this line is 8 columns, and a tab can fit in.

Suggestions for Plugin Developers ¶

TODO. For now please read the Plugin Guidelines on GitHub wiki.

Versioning ¶

This section applies beginning with version 0.14.0 of this specification.

This specification has a version, tagged in the specification repository. Each specification version corresponds to the same version in the core-tests repository.

The version numbering of the specification follows Semantic Versioning 2.0.0 (“SemVer”). The version numbering of the core-tests repository also follows SemVer.

Each EditorConfig core, to pass the core tests, must process version numbers given with the -b switch, and must report version numbers when given -v or --version. The version numbers used for -b, -v, and --version are versions of this specification. For example, the Vimscript core might respond to -v with:

EditorConfig Vimscript core v1.0.0 - Specification Version 0.14.0

Cores, plugins, or editors supporting EditorConfig have their own version numbers. Those version numbers are independent of the version number of this specification.