What are modules

A box with functions written on it

Working with the interactive shell is often considered a vital part of using dynamic programming languages. It is useful to test all kinds of code and programs. Most of the basic data types of Erlang were used without even needing to open a text editor or saving files. You could drop your keyboard, go play ball outside and call it a day, but you would be a terrible Erlang programmer if you stopped right there. Code needs to be saved somewhere to be used!

This is what modules are for. Modules are a bunch of functions regrouped in a single file, under a single name. Additionally, all functions in Erlang must be defined in modules. You have already used modules, perhaps without realizing it. The BIFs mentioned in the previous chapter, like hd or tl, actually belong to the erlang module, as well as all of the arithmetic, logic and Boolean operators. BIFs from the erlang module differ from other functions as they are automatically imported when you use Erlang. Every other function defined in a module you will ever use needs to be called with the form Module:Function(Arguments).

You can see for yourself:

1> erlang:element(2, {a,b,c}).
2> element(2, {a,b,c}).
3> lists:seq(1,4).
4> seq(1,4).
** exception error: undefined shell command seq/2

Here, the seq function from the list module was not automatically imported, while element was. The error 'undefined shell command' comes from the shell looking for a shell command like f() and not being able to find it. There are some functions from the erlang module which are not automatically imported, but they're not used too frequently.

Logically, you should put functions about similar things inside a single module. Common operations on lists are kept in the lists module, while functions to do input and output (such as writing to the terminal or in a file) are regrouped in the io module. One of the only modules you will encounter which doesn't respect that pattern is the aforementioned erlang module that has functions which do math, conversions, deal with multiprocessing, fiddle with the virtual machine's settings, etc. They have no point in common except being built-in functions. You should avoid creating modules like erlang and instead focus on clean logical separations.

Module Declaration

A scroll with small text on it

When writing a module, you can declare two kinds of things: functions and attributes. Attributes are metadata describing the module itself such as its name, the functions that should be visible to the outside world, the author of the code, and so on. This kind of metadata is useful because it gives hints to the compiler on how it should do its job, and also because it lets people retrieve useful information from compiled code without having to consult the source.

There is a large variety of module attributes currently used in Erlang code across the world; as a matter of fact, you can even declare your own attributes for whatever you please. There are some pre-defined attributes that will appear more frequently than others in your code. All module attributes follow the form -Name(Attribute).. Only one of them is necessary for your module to be compilable:

This is always the first attribute (and statement) of a file, and for good reason: it's the name of the current module, where Name is an atom. This is the name you'll use to call functions from other modules. The calls are made with the M:F(A) form, where M is the module name, F the function, and A the arguments.

It's time to code already! Our first module will be very simple and useless. Open your text editor and type in the following, then save it under useless.erl:


This line of text is a valid module. Really! Of course it's useless without functions. Let's first decide what functions will be exported from our 'useless' module. To do this, we will use another attribute:

-export([Function1/Arity, Function2/Arity, ..., FunctionN/Arity]).
This is used to define what functions of a module can be called by the outside world. It takes a list of functions with their respective arity. The arity of a function is an integer representing how many arguments can be passed to the function. This is critical information, because different functions defined within a module can share the same name if and only if they have a different arity. The functions add(X,Y) and add(X,Y,Z) would thus be considered different and written in the form add/2 and add/3 respectively.

Note: Exported functions represent a module's interface. It is important to define an interface revealing strictly what is necessary for it to be used and nothing more. Doing so lets you fiddle with all the other [hidden] details of your implementation without breaking code that might depend on your module.

Our useless module will first export a useful function named 'add', which will take two arguments. The following -export attribute can be added after the module declaration:


And now write the function:

add(A,B) ->
    A + B.

The syntax of a function follows the form Name(Args) -> Body., where Name has to be an atom and Body can be one or more Erlang expressions separated by commas. The function is ended with a period. Note that Erlang doesn't use the 'return' keyword. 'Return' is useless! Instead, the last logical expression of a function to be executed will have its value returned to the caller automatically without you having to mention it.

Add the following function (why yes, every tutorial needs a 'Hello world' example! Even at the fourth chapter!), without forgetting to add it to the -export attribute.

%% Shows greetings.
%% io:format/1 is the standard function used to output text.
hello() ->
    io:format("Hello, world!~n").

What we see from this function is that comments are single-line only and begin with a % sign (using %% is purely a question of style.) The hello/0 function also demonstrates how to call functions from foreign modules inside yours. In this case, io:format/1 is the standard function to output text, as written in the comments.

A last function will be added to the module, using both functions add/2 and hello/0:

greet_and_add_two(X) ->
A box being put in another one

Do not forget to add greet_and_add_two/1 to the exported function list. The calls to hello/0 and add/2 don't need to have the module name prepended to them because they were declared in the module itself.

Had you wanted to be able to call io:format/1 in the same manner as add/2 or any other function defined within the module, you could have added the following module attribute at the beginning of the file: -import(io, [format/1]).. Then you could have called format("Hello, World!~n"). directly. More generally, the -import attribute follows this recipe:

-import(Module, [Function1/Arity, ..., FunctionN/Arity]).

Importing a function is not much more than a shortcut for programmers when writing their code. Erlang programmers are often discouraged from using the -import attribute as some people find it reduces the readability of code. In the case of io:format/2, the function io_lib:format/2 also exists. Finding which one is used means going to the top of the file to see from which module it was imported. Consequently, leaving the module name in is considered good practice. Usually, the only functions you'll see imported come from the lists module: its functions are used with a higher frequency than those from most other modules.

Your useless module should now look like the following file:

-export([add/2, hello/0, greet_and_add_two/1]).

add(A,B) ->
    A + B.

%% Shows greetings.
%% io:format/1 is the standard function used to output text.
hello() ->
    io:format("Hello, world!~n").

greet_and_add_two(X) ->

We are done with the "useless" module. You can save the file under the name useless.erl. The file name should be the module name as defined in the -module attribute, followed by '.erl', which is the standard Erlang source extension.

Before showing how to compile the module and finally try all its exciting functions, we will see how to define and use macros. Erlang macros are really similar to C's '#define' statements, mainly used to define short functions and constants. They are simple expressions represented by text that will be replaced before the code is compiled for the VM. Such macros are mainly useful to avoid having magic values floating around your modules. A macro is defined as a module attribute of the form: -define(MACRO, some_value). and is used as ?MACRO inside any function defined in the module. A 'function' macro could be written as -define(sub(X,Y), X-Y). and used like ?sub(23,47), later replaced by 23-47 by the compiler. Some people will use more complex macros, but the basic syntax stays the same.

Compiling the code

Erlang code is compiled to bytecode in order to be used by the virtual machine. You can call the compiler from many places: $ erlc flags file.erl when in the command line, compile:file(FileName) when in the shell or in a module, c() when in the shell, etc.

It's time to compile our useless module and try it. Open the Erlang shell, type in:

1> cd("/path/to/where/you/saved/the-module/").
"Path Name to the directory you are in"

By default, the shell will only look for files in the same directory it was started in and the standard library: cd/1 is a function defined exclusively for the Erlang shell, telling it to change the directory to a new one so it's less annoying to browse for our files. Windows users should remember to use forward slashes. When this is done, do the following:

2> c(useless).

If you have another message, make sure the file is named correctly, that you are in the right directory and that you've made no mistake in your module. Once you successfully compile code, you'll notice that a useless.beam file was added next to useless.erl in your directory. This is the compiled module. Let's try our first functions ever:

3> useless:add(7,2).
4> useless:hello().
Hello, world!
5> useless:greet_and_add_two(-3).
Hello, world!
6> useless:not_a_real_function().
** exception error: undefined function useless:not_a_real_function/0

The functions work as expected: add/2 adds numbers, hello/0 outputs "Hello, world!", and greet_and_add_two/1 does both! Of course, you might be asking why hello/0 returns the atom 'ok' after outputting text. This is because Erlang functions and expressions must always return something, even if they would not need to in other languages. As such, io:format/1 returns 'ok' to denote a normal condition, the absence of errors.

Expression 6 shows an error being thrown because a function doesn't exist. If you have forgotten to export a function, this is the kind of error message you will have when trying it out.

Note: If you were ever wondering, '.beam' stands for Bogdan/Björn's Erlang Abstract Machine, which is the VM itself. Other virtual machines for Erlang exist, but they're not really used anymore and are history: JAM (Joe's Abstract Machine, inspired by Prolog's WAM and old BEAM, which attempted to compile Erlang to C, then to native code. Benchmarks demonstrated little benefits in this practice and the concept was given up.

There are a whole lot of compilation flags existing to get more control over how a module is compiled. You can get a list of all of them in the Erlang documentation. The most common flags are:

Erlang tools such as debuggers, code coverage and static analysis tools will use the debug information of a module in order to do their work.
By default, the Erlang compiler will create the 'beam' files in the current directory. This will let you choose where to put the compiled file.
Will ignore the -export module attribute and will instead export all functions defined. This is mainly useful when testing and developing new code, but should not be used in production.
-{d,Macro} or {d,Macro,Value}
Defines a macro to be used in the module, where Macro is an atom. This is more frequently used when dealing when unit-testing, ensuring that a module will only have its testing functions created and exported when they are explicitly wanted. By default, Value is 'true' if it's not defined as the third element of the tuple.

To compile our useless module with some flags, we could do one of the following:

7> compile:file(useless, [debug_info, export_all]).
8> c(useless, [debug_info, export_all]).

You can also be sneaky and define compile flags from within a module, with a module attribute. To get the same results as from expressions 7 and 8, the following line could be added to the module:

-compile([debug_info, export_all]).

Then just compile and you'll get the same results as if you manually passed flags. Now that we're able to write down functions, compile them and execute them, it's time to see how far we can take them!

Note: another option is to compile your Erlang module to native code. Native code compiling is not available for every platform and OS, but on those that support it, it can make your programs go faster (about 20% faster, based on anecdotal evidence). To compile to native code, you need to use the hipe module and call it the following way: hipe:c(Module,OptionsList). You could also use c(Module,[native]). when in the shell to achieve similar results. Note that the .beam file generated will contain both native and non-native code, and the native part will not be portable across platforms.

More About Modules

Before moving on to learning more about writing functions and barely useful snippets of code, there are a few other miscellaneous bits of information that might be useful to you in the future that I'd like to discuss.

The first one concerns metadata about modules. I mentioned in the beginning of this chapter that module attributes are metadata describing the module itself. Where can we find this metadata when we don't have an access to the source? Well the compiler plays nice with us: when compiling a module, it will pick up most module attributes and store them (along with other information) in a module_info/0 function. You can see the metadata of the useless module the following way:

9> useless:module_info().
10> useless:module_info(attributes).

The snippet above also shows an additional function, module_info/1 which will let you grab one specific piece of information. You can see exported functions, imported functions (none in this case!), attributes (this is where your custom metadata would go), and compile options and information. Had you decided to add -author("An Erlang Champ"). to your module, it would have ended up in the same section as vsn. There are limited uses to module attributes when it comes to production stuff, but they can be nice when doing little tricks to help yourself out: I'm using them in my testing script for this book to annotate functions for which unit tests could be better; the script looks up module attributes, finds the annotated functions and shows a warning about them.

Note: vsn is an automatically generated unique value differentiating each version of your code, excluding comments. It is used in code hot-loading (upgrading an application while it runs, without stopping it) and by some tools related to release handling. You can also specify a vsn value yourself if you want: just add -vsn(VersionNumber) to your module.

A small graph with three nodes: Mom, Dad and You. Mom and Dad are parents of You, and You is brother of Dad. Text under: 'If circular dependencies are digusting in real life, maybe they should be disgusting in your programs too'

Another point that would be nice to approach regards general module design: avoid circular dependencies! A module A should not call a module B that also calls module A. Such dependencies usually end up making code maintenance difficult. In fact, depending on too many modules even if they're not in a circular dependency can make maintenance harder. The last thing you want is to wake up in the middle of the night only to find a maniac software engineer or computer scientist trying to gouge your eyes out because of terrible code you have written.

For similar reasons (maintenance and fear for your eyes), it is usually considered a good practice to regroup functions that have similar roles close together. Starting and stopping an application or creating and deleting a record in some database are examples of such a scenario.

Well, that's enough for the pedantic moralizations. How about we explore Erlang a little more?