This HOWTO describes how documentation of the Erlang/OTP project works. If you want to know how to document a project built using Erlang/OTP see Documentation.
For details on how to build the Erlang/OTP documentation see How to Build the Documentation and Writing and building documentation.
The documentation for Erlang/OTP can be divided into three major sections:
- Erlang module reference documentation.
- User's Guides and other reference manuals (config files, command line tools, etc).
- Release notes and other project metadata (deprecations, removals etc)
- Internal documentation.
The first three are user facing documentation that describe how to use Erlang/OTP. The last is documentation describing how various part of Erlang/OTP works, for example the document you are reading now is internal documentation.
All user facing documentation, and some internal documentation, is rendered by ex_doc and follows the style described in Documentation. To get a full understanding on how to write documentation for Erlang/OTP you needs to read both the Documentation docs, ex_doc's docs, GFM docs, and earmark_parser docs. For most use cases though you don't need a full understanding and it is enough to just read Documentation.
Internal documentation is mostly written using GFM or just plain text.
All user facing documentation should follow the Documentation style rules described later in this document.
All user facing documentation is written using Markdown, specifically the Markdown flavor used by earmark_parser. The flavor is described in detail in the earmark_parser README, but it is basically GFM with some extra extensions, such as IAL and Mathematical Expressions.
The documentation is built using ex_doc and $ERL_TOP/make/ex_doc.exs, which is
a top level ex_doc configuration file describing how to build the docs for
Erlang/OTP. Each application then has a local docs.exs
(for example
stdlib) describing details specific for that application.
ex_doc divides documentation into two separate areas, "modules" and "extra pages". Modules is documentation about Erlang modules, and "extra pages" is Markdown files describing anything else.
In moduledoc
and extra pages it is also allowed to use mermaid diagrams to
complement the documentation text. For example
```mermaid
sequenceDiagram
Alice->>John: Hello John, how are you?
John->>Alice: Great!
Alice-)John: See you later!
```
will be rendered as:
sequenceDiagram
Alice->>John: Hello John, how are you?
John->>Alice: Great!
Alice-)John: See you later!
Any images or other static assets needed by the documentation should be placed
in the assets
folder.
All modules in applications except edoc
are documented using inline markdown as
described in Documentation.
edoc
modules are still documented using edoc
in order to test that it works.
However, ex_doc is still used to generate HTML docs for the edoc
application.
It is possible to create groups of types and/or functions within a module using documentation metadata. To do that you need to do two things:
- Add the key
titles
to themoduledoc
metadata with a list containing which groups you want. - Add the key
title
to each type and/or function you want to be listed within that group.
Example:
-module(test).
-moduledoc "A test module".
-moduledoc(#{ titles => [{function, <<"Obsolete API">>}, {type, <<"Obsolete API">>}] }).
-export([old_function/0]).
-doc "An old type".
-doc(#{ title => <<"Obsolete API">> }).
-type old_type() :: term().
-doc "An old function".
-doc(#{ title => <<"Obsolete API">> }).
old_function() -> very_old.
For an example in practice see the string module.
If you don't want to manually annotate each function, you can also create groupings programmatically. See Groups in the ex_doc documentation for details on how to do that. See the erts application docs.exs for an example.
Erlang/OTP by default divides "extra pages" into four groups, "User's Guides",
"References", "Command line tools" and "Internal Documentation". Which group a
guide is placed within depends on which sub-folder in doc
that the file is
placed in.
guides
-> "User's Guides"references/*_cmd.md
-> "Command line tools"references
-> "References"internal_docs
-> "Internal documentation"
For an examples, see the kernel application and erts application.
It is also possible to customize the groups for an application by adding a groups_for_extras
in
docs.exs
. See the system application's docs.exs for an example on how
to do that.
When writing documentation for the Erlang/OTP project, the following rules should be followed. The documentation has been written by a lot of different people, with their own writing style, that may or may not be following this style guide. When in doubt, follow the style that is used in the document you are currently editing in order to keep the same style throughout that document.
American English is to be used in all documents.
-
behavior - Should be spelled this way because it is American English. However, the module attribute is named "-behaviour.".
-
back-end - Not "backend"
-
boot script - ".script file", a script specifying which code to load and which applications to start. The binary version is called a boot file, see below.
-
boot file - ".boot file", an argument for the command line flag -boot.
-
code path - Not "search path"
-
command line arguments - Everything after
erl
on the command line. -
command line flags - Flags after
erl
on the command line, for example,-boot
. Subset of "command line arguments" -
(application) configuration parameter - Values that can be set in a
.config
file, or from Erlang shell or command line. Not "(application) variable". -
Erlang, Erlang/OTP, OTP - "Erlang" is the language. "Erlang/OTP" is the product and should normally be referred to as such. "OTP" is the project. "OTP" can also in some cases be used to refer to the parts of the product that are completely decoupled from the Erlang part. Example: 'OTP Design Principles', a part of Erlang/OTP, but you can write Erlang programs without ever caring about them.
Release versions is written as, for example, "Erlang/OTP R10B.", or only "OTP 26.0".
-
Erlang runtime system - What is started by the command
erl
. -
Erlang system - A system written in Erlang. Not the same thing as a runtime system.
-
function, function call - A function can be "local" (in the same module) or "external" (in another module). A function call can be "implicitly qualified" (without module name) or "fully qualified" (with module name).
-
node - Normally an Erlang runtime system that is part of a distributed Erlang system, but can also be used as an overarching concept (thus meaning a runtime system, distributed or not). A node can be "alive" or possibly "distributed", otherwise it is "non-distributed". "This" node is "the local node" other nodes are "remote".
-
port - A port can be "open" or "closed". It can also possibly be "existing" or "non-existing" (that is, be closed or never having existed).
-
process - A process can be "alive" or "dead". It can also possibly be "existing" or "non-existing" (that is, be dead or never having existed). A process is "local" (on "this" node) or "remote" (on another node). "This" process is "the calling process", not "current", "executing", "currently executing", or similar.
-
system configuration file - ".config file" an argument for the command line flag
-config
.
Refer to the Errors and Error Handling,
"Terminology" for
terminology related to errors and error handling. Note that it is no
longer appropriate to write that a process "exits", as this now (since
try
..catch
and the concept of "exception class" were introduced)
implies that the process either calls exit/1
or terminates normally.
Avoid contractions such as "can't", "it's" in written text. Write "cannot", "it is", and so on.
Do not write "can not", but almost always "cannot", except when "not" belongs with the words to the right (binds to the right).
Avoid "e.g.", "i.e.", and "etc.". Instead write "for example", "that is", and "and so on", respectively.
Use the Oxford comma. That is, in a series of three or more terms, place a comma immediately after the penultimate term.
Example:
"Common data types are lists, tuples, and maps."
Use punctuation properly around the word "however".
Examples:
-
"However, do not try this at home".
-
"They, however, did not agree."
-
"The new feature worked; however, it was far too slow."
However, when "however" is used in the sense of "no matter how" or "in whatever way", it should not be followed by a comma.
Example:
- "I am going fix this bug however long it takes."
Introduce the code example in the preceding text and end with a colon, not a period. Example: “The compiler gives a warning for the following code fragment:”
For code examples with several steps and for long examples:
- Ensure that it is obvious for the reader what belongs to the example.
- Use either a subheading
#### Example
or use a separate paragraph with the text*Example*:
.
Example:
The compiler gives a warning for some code fragments.
*Example*:
```erlang
> erlc test.erl
test.erl:3:6: Warning: variable 'Data' is unused
% 3| main(Data) ->
% | ^
```
Use capital letters at the beginning of each word, except for small words like "and" and "of," following American English language tradition.
Example: Principles of Writing Titles and Captions
Name for applications in text should be considered proper names and should be written beginning with capital letter.
Except for when they are acronyms or some other tradition exists. Example:
asn1
Common Test
Compiler
Crypto
Debugger
Dialyzer
Diameter
EDoc
Eldap
Erl_interface
Event Tracer
EUnit
FTP
Inets
Jinterface
Kernel
Megaco
Mnesia
Observer
ODBC
OS_Mon
Parsetools
Public_Key
Reltool
Runtime_Tools
SASL
SNMP
SSH
SSL
STDLIB
Syntax_Tools
TFTP
Tools
Wx
Xmerl
When documenting a set of options or items you would normally use a definition list. However, Markdown does not support them so a convention using un-ordered lists should be used:
- **tag**{: #tag-anchor } - Description
It is recommended, but not required, to place an anchor on each definition list item. Placing an anchor on an item will make it possible for search to find that specific item.
For example the erl_cmd.md has this item:
- **`+c true | false`**{: #+c } - Enables or disables
[time correction](time_correction.md#time-correction):
- **`true`** - Enables time correction. This is the default if time correction
is supported on the specific platform.
- **`false`** - Disables time correction.
For backward compatibility, the boolean value can be omitted. This is
interpreted as `+c false`.
The above example has nested definition lists using an anchor for the top item.
When a function call is the same as another, you should strive to use
the argument form of the equiv
doc
metadata.
Example:
-doc #{equiv => atom_to_binary(Atom, utf8)}).
-spec atom_to_binary(Atom) -> binary() when
Atom :: atom().
If it is not possible to use equiv
, then you should use:
-doc """
Equivalent to [atom_to_binary(Atom, utf8)](`atom_to_binary/2`).
""".
Sometimes it is useful to be able to concisely refer to multiple arities of the same function. This should be done like this:
See [`disk_log:blog/1,2`](`disk_log:blog/2`).
See [`disk_log:blog/*`](`disk_log:blog/2`).
The link should point the to arity with the most relevant documentation, which usually is the largest arity function.
Diagrams and images are powerful tools to describe complicated concepts. They should not be used in function/type/callback documentation as that documentation is used by editors to display help and typically cannot render any images.
For simple diagrams it is recommended to use
dia and then commit both the .dia
source
and the .png
generated. See the emulator internal docs
Makefile for an
example on how it is done.
When writing since
annotations for functions/types/callbacks the
convention is that the -doc
attribute shows when a specific MFA
was added, and then in text describe when specific options or
behaviors are added.
Example:
-doc #{since => ~"OTP 26.2"}.
-spec info([key1, key2]) -> list(term()).
-doc """
Get information about a thing.
- **`key1`** - Fetch `key1` data.
- **`key2`** - Fetch `key2` data.
Since: OTP 26.3
""".
As the actual version is generally not yet certain at the time when a
change gets merged, the an internal ticket number (for example
OTP-12345
) to use as a placeholder in the respective since
tags,
such as -doc #{since => ~"OTP OTP-12345"}.
. When your new
functions are released with an Erlang/OTP release, this placeholder
will get replaced with the actual OTP version, leading to something
like "OTP 26.0".