Hello, world
Last updated
Last updated
It is traditional for your first program in a new language to be .
Create a folder wherever you like
Put a new file in it called hello.ex and put the following code inside it
Type in iex hello.ex followed by Greetings.say_hello() to be greeted by your program!
When you write a program in Elixir, you will have an Elixir module defined with a function inside it. Modules are ways of grouping up related Elixir code together.
The defmodule keyword is how you define a module with a name and a body.
The def keyword is how you define a function with a name and a body.
Functions in Elixir implicitly return the last expression the function without the need specify the return keyword like in other languages.
The @spec keyword is used to specify the types of the arguments and the return value of a function.
We are expecting say_hello/0 to return the :ok atom
The /0 besides say_hello in the above sentence stands for the amount of argumens say_hello takes in.
IO.puts prints a line to stdout and returns an :ok atom when it has finished exectuting successfully.
Atoms are constants that represent a name or a string. They commonly used as message names in Elixir's process-oriented programming model. The :ok atom is a common convention in Elixir to indicate that a function or operation has completed successfully.
How do you test this? It is good to separate your "domain" code from the outside world (side-effects). The IO.puts is a side effect (printing to stdout) and the string we send in is our domain.
So let's separate these concerns so it's easier to test
We have created a new function again with def. This time we've added a type spec of String.t() in the definition. This means this function returns a string.
Now create a new file called hello_test.exs where we are going to write a test for our hello/0 function using ExUnit
Run elixirc hello.ex to first compile the app.
Then, run elixir hello_test.exs in your terminal. It should've passed! Just to check, try deliberately breaking the test by changing the "Hello, world" string.
Notice the file extensions .exs which hello_test.exs uses and the .ex our hello.ex file uses. The main difference between .ex and .exs files in Elixir is the way they are compiled and executed. .ex files are compiled ahead-of-time (AOT) into bytecode and are intended to be used in production environments, while .exs files are interpreted at runtime and are typically used for development, testing, and other tasks that don't require the performance benefits of AOT compilation.
Writing a test is just like writing a function, with a few rules
It needs to be in a file with a name like xxx_test.exs
The test module must end with the word Test(ex. GreetingsTest)
The test function must start with the word test
ExUnit.Case needs to be imported and callbacks invoked using the keyword use
ExUnit.start()-Starts ExUnit and automatically runs tests right before the VM terminates.
ExUnit.CaseExUnit.Case is a module for defining test cases. It provides various functions for defining and running tests, as well as for setting up and tearing down test fixtures. When a module uses the ExUnit.Case macro, ExUnit will treat the module as a container for test functions and make the test functions defined inside it available to the testing framework.
In Elixir, a case is a control structure that allows you to match the value of an expression against a series of patterns, and execute different code depending on which pattern matches. The ExUnit.Case module is named after this control structure because it is used to define a series of test cases that can be run to verify that your code is working as expected.
A macro in Elixir can be thought of as a special kind of function that operates at compile time, rather than at runtime. However, instead of returning a value, a macro generates code that is then inserted into the program at the point where the macro is called.
testThe test macro, which is defined in the ExUnit.Case module, defines a test function. A test function must be defined inside a module that uses the ExUnit.Case macro. It takes a string describing the test as an argument.
assertThe assert function is used to make assertions about the expected behavior of your code. When you write a test case, you typically specify the input to your code, the expected output, and any other conditions that should be met. The assert function allows you to check that the actual output of your code matches the expected output and that the other conditions are met.
elixircelixirc is used to compile Elixir source code files into bytecode that can be run on the Erlang Virtual Machine (VM). In order to run tests on your Elixir application, you need to compile the code first. Tests are run on the compiled version of the code, which has been transformed into bytecode that can be run on the Erlang Virtual Machine (VM).
Now that we have a test we can iterate on our software safely.
In the last example we wrote the test after the code had been written just so you could get an example of how to write a test and declare a function. From this point on we will be writing tests first.
Our next requirement is to let us specify the recipient of the greeting.
Let's start by capturing these requirements in a test. This is basic test driven development and allows us to make sure our test is actually testing what we want. When you retrospectively write tests there is the risk that your test may continue to pass even if the code doesn't work as intended.
Now run elixir hello_test.exs, ExUnit's test runner should give you an error
The first line is the name of the test
The actual error: ** (UndefinedFunctionError)
code: the code of the function that was being executed when the error or exception occurred.
stacktrace: the sequence of function calls that led to the error.
The final line tells us how many tests there are and how many of them failed.
It is important to listen to the ExUnit Test Runner when running into errors because it provides valuable information about what went wrong with the tests and how to fix the issue.
Here, the test runner is telling you to change the function hello/0 to accept an argument of one. Do so now.
Then, edit the type spec to reflect this. Since we know name will be a string, we'll add String.t() to hello/0's argument field and thus make it hello/1.
If you try and run your program again with elixirc, your hello.ex will fail to compile because you're not passing in an argument to the say_hello/0 function.
It is still important to listen to the compiler in a dynamically typed language like Elixir. While Elixir does not require explicit type declarations, the compiler still performs various checks and optimizations on the code. Listening to it can help catch errors and improve the performance of your code.
Send in "world" to make hello.ex compile.
It is important to run elixirc after making changes to your code, to recompile you app so that your tests are testing your latest code.
Now when you run your tests you should see something like
The right: represent want we are trying to assert, while left: represents the actual value returned by the function call.
Let's make the test pass by using the name argument and concatenate it with "Hello, "
When you run the tests they should now pass. Normally as part of the TDD cycle we should now refactor.
At this point, if you are using source control (which you should!) I would commit the code as it is. We have working software backed by a test.
I wouldn't push to master though, because I plan to refactor next. It is nice to commit at this point in case you somehow get into a mess with refactoring - you can always go back to the working version.
There's not a lot to refactor here, but we can introduce another language feature, module attributes.
Module attributes are variables that are associated with a specific module and can be accessed from anywhere within that module. They are defined using the "@" symbol, followed by the attribute name and value.
Module attributes are defined like so
We can now refactor our code
After refactoring, re-run your tests to make sure you haven't broken anything.
Module Attributes should improve performance of your application as it saves you creating the "Hello, " string instance every time hello/1 is called.
To be clear, the performance boost is incredibly negligible for this example! But it's worth thinking about creating module attributes to capture the meaning of values and sometimes to aid performance.
The next requirement is when our function is called with an empty string it defaults to printing "Hello, world", rather than "Hello, ".
Start by writing a new failing test
Here we are introducing another tool in our testing arsenal, the describe macro.
The describe, macro is used to define a test case in the ExUnit testing framework. It allows you to group related tests together and give them a descriptive name
Now that we have a well-written failing test, let's fix the code, using pattern matching.
If we run our tests we should see it satisfies the new requirement and we haven't accidentally broken the other functionality.
In Elixir, pattern matching is a feature that allows you to specify patterns that data should conform to, and then check if the data matches those patterns. It can be used to analyze input and determine the behavior of a function based on the form of the input.
The name_with_default/1 function is defined using two separate clauses, each with its own pattern in the function head. The first clause has an empty string pattern, and the second clause has a variable pattern.
When the name_with_default/1 function up above is called, Elixir will attempt to match the argument passed to it against the patterns in the function head of each clause. If the argument is an empty string, the first clause will be selected and the body of that clause will be executed, returning the string "world". If the argument is any other value, the second clause will be selected and the body of that clause will be executed, returning the value of the name argument.
The defp keyword is how you define a private function with a name and a body. They are functions that are defined and only visible within a module. They also help to reduce the complexity of the module's public API by hiding implementation details that are not relevant to external code.
Now we are happy with the code I would amend the previous commit so we only check in the lovely version of our code with its test.
Let's go over the cycle again
Write a test
Run the test, see that it fails and check the error message is meaningful
Write enough code to make the test pass
Refactor
On the face of it this may seem tedious but sticking to the feedback loop is important.
Not only does it ensure that you have relevant tests, it helps ensure you design good software by refactoring with the safety of tests.
Seeing the test fail is an important check because it also lets you see what the error message looks like. As a developer it can be very hard to work with a codebase when failing tests do not give a clear idea as to what the problem is.
By ensuring your tests are fast and setting up your tools so that running tests is simple you can get in to a state of flow when writing your code.
By not writing tests you are committing to manually checking your code by running your software which breaks your state of flow and you won't be saving yourself any time, especially in the long run.
Goodness me, we have more requirements. We now need to support a second parameter, specifying the language of the greeting. If a language is passed in that we do not recognise, just default to English.
We should be confident that we can use TDD to flesh out this functionality easily!
Write a test for a user passing in Spanish. Add it to the existing suite.
Remember not to cheat! Test first. When you try and run the test, the test runner should complain because you are calling hello/1 with two arguments rather than one.
Fix the test error by adding another string argument to hello/1.
When you try and run the test again it will complain about not passing through enough arguments to hello/2 in your other tests and in hello.exs
Fix them by passing through empty strings. Now all your tests should pass, apart from our new scenario
We can use pattern matching here to check the language is equal to :spanish and if so change the message
The tests should now pass.
Now it is time to refactor. You should see some problems in the code, "magic" strings, some of which are repeated. Try and refactor it yourself, with every change make sure you re-run the tests to make sure your refactoring isn't breaking anything.
Write a test asserting that if you pass in :french you get "Bonjour, "
See it fail, check the error message is easy to read
Do the smallest reasonable change in the code
You may have written something that looks roughly like this
caseWhen you have lots of statements checking a particular value it is common to use a case statement instead. We can use case to refactor the code to make it easier to read and more extensible if we wish to add more language support later
The case statement in Elixir is similar to the switch statement in other languages except that it can match against any expression, not just simple values like integers or strings.
You can use the case statement to match against complex data structures, such as lists or maps, using pattern matching.
In other languages, the switch statement will continue to execute the next case clause if you do not include a break statement, whereas in Elixir, the case statement will stop executing as soon as it finds a matching clause.
Write a test to now include a greeting in the language of your choice and you should see how simple it is to extend our amazing function.
You could argue that maybe our function is getting a little big. The simplest refactor for this would be to extract out some functionality into another function.
A few new concepts:
Functions in Elixir are first-class citizens. We can bind functions to variables, store them in other data types, pass them as arguments to other functions; such as we did with prefix.
Functions are just data as far as our programs are concerned. We can work with them like we would work with any data structure.
The _ symbol is used as a catch-all pattern in a case statement. In our code it is used as the final arm of the case statement, and it matches any value that is not matched by the previous arms.
The defp keyword is used to define a private function. They are accessible and only visible internally within a module. They help to reduce the complexity of the module's public API by hiding implementation details that are not relevant to external code.
Who knew you could get so much out of Hello, world?
By now you should have some understanding of:
Writing tests
Declaring functions, with arguments and return types
pattern matching and case
Declaring variables and module attributes
Write a failing test and see it fail so we know we have written a relevant test for our requirements and seen that it produces an easy to understand description of the failure
Writing the smallest amount of code to make it pass so we know we have working software
Then refactor, backed with the safety of our tests to ensure we have well-crafted code that is easy to work with
In our case we've gone from hello() to hello("name"), to hello("name", :french) in small, easy to understand steps.
This is of course trivial compared to "real world" software but the principles still stand. TDD is a skill that needs practice to develop, but by breaking problems down into smaller components that you can test, you will have a much easier time writing software.