Skip to content

Instantly share code, notes, and snippets.

@miekg miekg/learninggo.txt
Created Aug 26, 2018

Embed
What would you like to do?
Go Working Group R. Gieben
Internet-Draft 25 August 2018
Intended status: Informational
Expires: 26 February 2019
Learning Go
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on 26 February 2019.
Copyright Notice
Copyright (c) 2018 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (http://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Simplified BSD License text
as described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Simplified BSD License.
Table of Contents
1. Preface
2. Learning Go
3. Introduction
3.1. How to Read this Book
3.2. Official Documentation
4. Basics
4.1. Hello World
4.2. Compiling and Running Code
4.3. Variables, Types and Keywords
4.3.1. Boolean Types
4.3.2. Numerical Types
4.3.3. Constants
4.3.4. Strings
4.3.5. Runes
4.3.6. Complex Numbers
4.3.7. Errors
4.4. Operators and Built-in Functions
4.5. Go Keywords
4.6. Control Structures
4.6.1. If-Else
4.6.2. Goto
4.6.3. For
4.6.4. Break and Continue
4.6.5. Range
4.6.6. Switch
4.7. Built-in Functions
4.8. Arrays, Slices, and Maps
4.8.1. Arrays
4.8.2. Slices
4.8.3. Maps
4.9. Exercises
4.9.1. For-loop
4.9.2. Answer
4.9.3. Average
4.9.4. Answer
4.9.5. FizzBuzz
4.9.6. Answer
5. Functions
5.1. Scope
5.2. Functions as values
5.3. Callbacks
5.4. Deferred Code
5.5. Variadic Parameter
5.6. Panic and recovering
5.7. Exercises
5.7.1. Average
5.7.2. Answer
5.7.3. Bubble sort
5.7.4. Answer
5.7.5. For-loop II
5.7.6. Answer
5.7.7. Fibonacci
5.7.8. Answer
5.7.9. Var args
5.7.10. Answer
5.7.11. Functions that return functions
5.7.12. Answer
5.7.13. Maximum
5.7.14. Answer
5.7.15. Map function
5.7.16. Answer
5.7.17. Stack
5.7.18. Answer
6. Packages
6.1. Identifiers
6.2. Documenting packages
6.3. Testing packages
6.4. Useful packages
6.5. Exercises
6.5.1. Stack as package
6.5.2. Answer
6.5.3. Calculator
6.5.4. Answer
7. Beyond the basics
7.1. Allocation
7.1.1. Allocation with new
7.1.2. Allocation with make
7.1.3. Constructors and composite literals
7.2. Defining your own types
7.2.1. More on structure fields
7.2.2. Methods
7.3. Conversions
7.3.1. User defined types and conversions
7.4. Exercises
7.4.1. Map function with interfaces
7.4.2. Answer
7.4.3. Pointers
7.4.4. Answer
7.4.5. Linked List
7.4.6. Answer
7.4.7. Cat
7.4.8. Answer
7.4.9. Method calls
7.4.10. Answer
8. Interfaces
8.1. Which is what?
8.2. Empty interface
8.3. Methods
8.4. Methods on interface types
8.5. Interface names
8.6. A sorting example
8.7. Listing interfaces in interfaces
8.8. Introspection and reflection
8.9. Exercises
8.9.1. Answer
8.9.2. Pointers and reflection
8.9.3. Answer
9. Concurrency
9.1. Make it run in parallel
9.2. More on channels
9.3. Exercises
9.3.1. Channels
9.3.2. Answer
9.3.3. Fibonacci II
9.3.4. Answer
10. Communication
10.1. io.Reader
10.2. Some examples
10.3. Command line arguments
10.4. Executing commands
10.5. Networking
10.6. Exercises
10.6.1. Finger daemon
10.6.2. Answer
10.6.3. Echo server
10.6.4. Answer
10.6.5. Word and Letter Count
10.6.6. Answer
10.6.7. Uniq
10.6.8. Answer
10.6.9. Quine
10.6.10. Answer
10.6.11. Processes
10.6.12. Answer
10.6.13. Number cruncher
10.6.14. Answer
11. Informative References
Author's Address
1. Preface
The source of this book (https://github.com/miekg/learninggo) is
written in mmark (https://github.com/mmarkdown/mmark) and is
converted from the original LaTeX source (https://github.com/miekg/
gobook).
_All example code used in this book is hereby licensed under the
Apache License version 2.0._
<blockquote>:
This work is licensed under the Attribution-NonCommercial-ShareAlike
3.0 Unported License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/3.0/
(http://creativecommons.org/licenses/by-nc-sa/3.0/) or send a letter
to Creative Commons, 171 Second Street, Suite 300, San Francisco,
California, 94105, USA.
The following people made large or small contributions to earlier
versions of this book:
Adam J. Gray, Alexander Katasonov, Alexey Chernenkov, Alex Sychev,
Andrea Spadaccini, Andrey Mirtchovski, Anthony Magro, Babu Sreekanth,
Ben Bullock, Bob Cunningham, Brian Fallik, Cecil New, Cobold, Damian
Gryski, Daniele Pala, Dan Kortschak, David Otton, Fabian Becker,
Filip Zaludek, Hadi Amiri, Haiping Fan, Iaroslav Tymchenko, Jaap
Akkerhuis, JC van Winkel, Jeroen Bulten, Jinpu Hu, John Shahid,
Jonathan Kans, Joshua Stein, Makoto Inoue, Marco Ynema, Mayuresh
Kathe, Mem, Michael Stapelberg, Nicolas Kaiser, Olexandr Shalakhin,
Paulo Pinto, Peter Kleiweg, Philipp Schmidt, Robert Johnson, Russel
Winder, Simoc, Sonia Keys, Stefan Schroeder, Thomas Kapplet, T.J.
Yang, Uriel"\dagger", Vrai Stacey, Xing Xing.
"Learning Go" has been translated into (note that this used the
original LaTeX source).
* Chinese, by Xing Xing, 这里是中文译本: http://www.mikespook.com/learning-
go/ (http://www.mikespook.com/learning-go/)
I hope this book is useful.
Miek Gieben, London, 2015.
This book still sees development, small incremental improvements
trickle in from Github.
Miek Gieben, London, 2017.
Learning Go's source has been rewritten in mmark2
(https://github.com/mmarkdown/mmark), but did not see any other
changes.
Miek Gieben, London, 2018.
2. Learning Go
3. Introduction
<blockquote>:
Is Go an object-oriented language? Yes and no.
The Go programming language is an open source project language to
make programmers more productive.
According to the website [go_web] "Go is expressive, concise, clean,
and efficient". And indeed it is. My initial interest was piqued
when I read early announcements about this new language that had
built-in concurreny and a C-like syntax (Erlang also has built-in
concurrency, but I could never get used to its syntax). Go is a
compiled statically typed language that feels like a dynamically
typed, interpreted language. My go to (scripting!) language Perl has
taken a back seat now that Go is around.
The unique Go language is defined by these principles:
Clean and Simple
Go strives to keep things small and beautiful. You should be able
to do a lot in only a few lines of code.
Concurrent
Go makes it easy to "fire off" functions to be run as _very_
lightweight threads. These threads are called goroutines
Channels
Communication with these goroutines is done, either via shared
state or via [csp].
Fast
Compilation is fast and execution is fast. The aim is to be as
fast as C. Compilation time is measured in seconds.
Safe
Explicit casting and strict rules when converting one type to
another. Go has garbage collection. No more "free()" in Go: the
language takes care of this.
Standard format
A Go program can be formatted in (almost) any way the programmers
want, but an official format exists. The rule is very simple: The
output of the filter "gofmt" _is the officially endorsed format_.
Postfix types
Types are given _after_ the variable name, thus "var a int",
instead of "int a".
UTF-8
UTF-8 is everywhere, in strings _and_ in the program code.
Finally you can use "\Phi = \Phi + 1" in your source code.
Open Source
The Go license is completely open source.
Fun
Programming with Go should be fun!
As I mentioned Erlang also shares some features of Go. A notable
difference between Erlang and Go is that Erlang borders on being a
functional language, while Go is imperative. And Erlang runs in a
virtual machine, while Go is compiled.
3.1. How to Read this Book
I've written this book for people who already know some programming
languages and how to program. In order to use this book, you (of
course) need Go installed on your system, but you can easily try
examples online in the Go playground. All exercises in this book
work with Go 1, the first stable release of Go -- if not, it's a bug.
The best way to learn Go is to create your own programs. Each
chapter therefore includes exercises (and answers to exercises) to
acquaint you with the language. Each exercise is either _easy_,
_intermediate_, or _difficult_. The answers are included after the
exercises on a new page. Some exercises don't have an answer; these
are marked with an asterisk.
Here's what you can expect from each chapter:
Section 4
We'll look at the basic types, variables, and control
structures available in the language.
Section 5
Here we look at functions, the basic building blocks of
Go programs.
Section 6
We'll see that functions and data can be grouped together
in packages. We'll also see how to document and test our
packages.
Section 7
We'll create our own types. We'll also look at memory
allocations in Go.
Section 8
We'll learn how to use interfaces. Interfaces are the
central concept in Go, as Go does not support object
orientation in the traditional sense.
Section 9
We'll learn the "go" keyword, which can be used to start
function in separate routines (called goroutines).
Communication with those goroutines is done via channels.
Section 10
Finally we'll see how to interface with the rest of the
world from within a Go program. We'll see how to create
files and read and write to and from them. We'll also
briefly look into networking.
3.2. Official Documentation
There is a substantial amount of documentation written about Go. The
Go Tutorial [go_tutorial], the Go Tour (with lots of exercises) and
the Effective Go [effective_go] are helpful resources. The website
http://golang.org/doc/ (http://golang.org/doc/) is a very good
starting point for reading up on Go. Reading these documents is
certainly not required, but it is recommended.
<blockquote>:
When searching on the internet use the term "golang" instead of plain
"go".
Go comes with its own documentation in the form of a program called
"godoc". If you are interested in the documentation for the built-
ins, simply do this:
% godoc builtin
To get the documentation of the "hash" package, just:
% godoc hash
To read the documentation of "fnv" contained in "hash", you'll need
to issue "godoc hash/fnv" as "fnv" is a subdirectory of "hash".
<CODE BEGINS>
PACKAGE DOCUMENTATION
package fnv
import "hash/fnv"
Package fnv implements FNV-1 and FNV-1a, non-cryptographic hash
...
<CODE ENDS>
4. Basics
<blockquote>:
I am interested in this and hope to do something.
In this chapter we will look at the basic building blocks of the Go
programming language.
4.1. Hello World
In the Go tutorial, you get started with Go in the typical manner:
printing "Hello World" (Ken Thompson and Dennis Ritchie started this
when they presented the C language in the 1970s). That's a great way
to start, so here it is, "Hello World" in Go.
<CODE BEGINS>
package main <1>
import "fmt" <2> // Implements formatted I/O.
/* Print something */ <3>
func main() { <4>
fmt.Printf("Hello, world.") <5>
}
<CODE ENDS>
Lets look at the program line by line. This first line is just
required _1_. All Go files start with "package <something>", and
"package main" is required for a standalone executable.
"import "fmt"" says we need "fmt" in addition to "main" _2_. A
package other than "main" is commonly called a library, a familiar
concept in many programming languages (see Section 6). The line ends
with a comment that begins with "//".
Next we another comment, but this one is enclosed in "/*" "*/" _3_.
When your Go program is executed, the first function called will be
"main.main()", which mimics the behavior from C. Here we declare
that function _4_.
Finally we call a function from the package "fmt" to print a string
to the screen. The string is enclosed with """ and may contain non-
ASCII characters _5_.
4.2. Compiling and Running Code
To build a Go program, use the "go" tool."helloworld" we just enter:
% go build helloworld.go
"helloworld".
% ./helloworld
Hello, world.
You can combine the above and just call "go run helloworld.go".
4.3. Variables, Types and Keywords
In the next few sections we will look at the variables, basic types,
keywords, and control structures of our new language.
Go is different from (most) other languages in that the type of a
variable is specified _after_ the variable name. So not: "int a",
but "a int". When you declare a variable it is assigned the
"natural" null value for the type. This means that after "var a
int", "a" has a value of 0. With "var s string", "s" is assigned the
zero string, which is """". Declaring and assigning in Go is a two
step process, but they may be combined. Compare the following pieces
of code which have the same effect.
<CODE BEGINS>
var a int a := 15
var b bool b := false
a = 15
b = false
<CODE ENDS>
On the left we use the "var" keyword to declare a variable and _then_
assign a value to it. The code on the right uses ":=" to do this in
one step (this form may only be used _inside_ functions). In that
case the variable type is _deduced_ from the value. A value of 15
indicates an "int". A value of "false" tells Go that the type should
be "bool". Multiple "var" declarations may also be grouped; "const"
(see Section 4.3.3) and "import" also allow this. Note the use of
parentheses instead of braces:
<CODE BEGINS>
var (
x int
b bool
)
<CODE ENDS>
Multiple variables of the same type can also be declared on a single
line: "var x, y int" makes "x" and "y" both "int" variables. You can
also make use of _parallel assignment_"a, b := 20, 16". This makes
"a" and "b" both integer variables and assigns 20 to "a" and 16 to
"b".
A special name for a variable is "_". "/dev/null" on Unix). In this
example we only assign the integer value of 35 to "b" and discard the
value 34: "_, b := 34, 35". Declared but otherwise _unused_
variables are a compiler error in Go.
4.3.1. Boolean Types
A boolean type represents the set of boolean truth values denoted by
the predeclared constants _true_ and _false_. The boolean type is
"bool".
4.3.2. Numerical Types
Go has most of the well-known types such as "int". The "int" type
has the appropriate length for your machine, meaning that on a 32-bit
machine it is 32 bits and on a 64-bit machine it is 64 bits. Note:
an "int" is either 32 or 64 bits, no other values are defined. Same
goes for "uint", the unsigned int.
If you want to be explicit about the length, you can have that too,
with "int32", or "uint32". The full list for (signed and unsigned)
integers is "int8", "int16", "int32", "int64" and "byte", "uint8",
"uint16", "uint32", "uint64", with "byte" being an alias for "uint8".
For floating point values there is "float32" and "float64" (there is
no "float" type). A 64 bit integer or floating point value is
_always_ 64 bit, also on 32 bit architectures.
Note that these types are all distinct and assigning variables which
mix these types is a compiler error, like in the following code:
<CODE BEGINS>
package main
func main() {
var a int
var b int32
b = a + a
b = b + 5
}
<CODE ENDS>
We declare two different integers, a and b where a is an "int" and b
is an "int32". We want to set b to the sum of a and a. This fails
and gives the error: "cannot use a + a (type int) as type int32 in
assignment". Adding the constant 5 to b _does_ succeed, because
constants are not typed.
4.3.3. Constants
Constants in Go are just that --- constant. They are created at
compile time, and can only be numbers, strings, or booleans; "const x
= 42" makes "x" a constant. You can use _iota_
<CODE BEGINS>
const (
a = iota
b
)
<CODE ENDS>
The first use of "iota" will yield 0, so "a" is equal to 0. Whenever
"iota" is used again on a new line its value is incremented with 1,
so "b" has a value of 1. Or, as shown here, you can even let Go
repeat the use of "iota". You may also explicitly type a constant:
"const b string = "0"". Now "b" is a "string" type constant.
4.3.4. Strings
Another important built-in type is "string". Assigning a string is
as simple as:
<CODE BEGINS>
s := "Hello World!"
<CODE ENDS>
Strings in Go are a sequence of UTF-8 characters enclosed in double
quotes ("). If you use the single quote (') you mean one character
(encoded in UTF-8) --- which is _not_ a "string" in Go.
Once assigned to a variable, the string cannot be changed: strings in
Go are immutable. If you are coming from C, note that the following
is not legal in Go:
<CODE BEGINS>
var s string = "hello"
s[0] = 'c'
<CODE ENDS>
To do this in Go you will need the following:
<CODE BEGINS>
s := "hello"
c := []rune(s) <1>
c[0] = 'c' <2>
s2 := string(c) <3>
fmt.Printf("%s\n", s2) <4>
<CODE ENDS>
Here we convert "s" to an array of runes _1_. We change the first
element of this array _2_. Then we create a _new_ string "s2" with
the alteration _3_. Finally, we print the string with "fmt.Printf"
_4_.
4.3.5. Runes
"Rune" is an alias for "int32". It is an UTF-8 encoded code point.
When is this type useful? _not_ in Go!). But to get the actual
characters you should use the "rune" type.
4.3.6. Complex Numbers
Go has native support for complex numbers. To use them you need a
variable of type "complex128" (64 bit real and imaginary parts) or
"complex64" (32 bit real and imaginary parts). Complex numbers are
written as "re + im""i", where "re" is the real part, "im" is the
imaginary part and "i" is the literal '"i"' ("\sqrt{-1}").
4.3.7. Errors
Any non-trivial program will have the need for error reporting sooner
or later. Because of this Go has a builtin type specially for
errors, called "error". "var e error" creates a variable "e" of type
"error" with the value "nil". This error type is an interface --
we'll look more at interfaces in Section 8. For now you can just
assume that "error" is a type just like all other types.
4.4. Operators and Built-in Functions
Go supports the normal set of numerical operators. See Table 1 for
lists the current ones and their relative precedence. They all
associate from left to right.
+------------+--------------------+
| Precedence | Operator(s) |
+============+====================+
| Highest | "* / % << >> & &^" |
+------------+--------------------+
| | `+ - |
+------------+--------------------+
| | "== != < <= > >=" |
+------------+--------------------+
| | "<-" |
+------------+--------------------+
| | "&&" |
+------------+--------------------+
| Lowest | || |
+------------+--------------------+
Table 1: Operator precedence.
"+ - * /" and "%" all do what you would expect, "& | ^" and "&^" are
bit operators for bitwise _and__or__xor_"&&" and "||" operators are
logical _and_ _or_ "!"
Although Go does not support operator overloading (or method
overloading for that matter), some of the built-in operators _are_
overloaded. For instance, "+" can be used for integers, floats,
complex numbers and strings (adding strings is concatenating them).
4.5. Go Keywords
Let's start looking at keywords, Table 2 lists all the keywords in
Go.
+------------+---------------+----------+-------------+----------+
+============+===============+==========+=============+==========+
| "break" | "default" | "func" | "interface" | "select" |
+------------+---------------+----------+-------------+----------+
| "case" | "defer" | "go" | "map" | "struct" |
+------------+---------------+----------+-------------+----------+
| "chan" | "else" | "goto" | "package" | "switch" |
+------------+---------------+----------+-------------+----------+
| "const" | "fallthrough" | "if" | "range" | "type" |
+------------+---------------+----------+-------------+----------+
| "continue" | "for" | "import" | "return" | "var" |
+------------+---------------+----------+-------------+----------+
Table 2: Keywords in Go.
We've seen some of these already. We used "var" and "const" in the
Section 4.3 section, and we briefly looked at "package" and "import"
in our "Hello World" program at the start of the chapter. Others
need more attention and have their own chapter or section:
* "func" is used to declare functions and methods.
* "return" is used to return from functions. We'll look at both
"func" and "return" in detail in Section 5.
* "go" is used for concurrency. We'll look at this in
Section 9.3.1.
* "select" used to choose from different types of communication,
We'll work with "select" in Section 9.3.1.
* "interface" is covered in Section 8.
* "struct" is used for abstract data types. We'll work with
"struct" in Section 7.
* "type" is also covered in Section 7.
4.6. Control Structures
There are only a few control structures in Go. To write loops we use
the "for" keyword, and there is a "switch" and of course an "if".
When working with channels "select" will be used (see Section 9.3.1).
Parentheses are are not required around the condition, and the body
must _always_ be brace-delimited.
4.6.1. If-Else
In Go an "if"
<CODE BEGINS>
if x > 0 {
return y
} else {
return x
}
<CODE ENDS>
"if" and "switch" accept an initialization statement, it's common to
see one used to set up a (local) variable.
<CODE BEGINS>
if err := SomeFunction(); err == nil {
// do something
} else {
return err
}
<CODE ENDS>
It is idomatic in Go to omit the "else" when the "if" statement's
body has a "break", "continue", "return" or, "goto", so the above
code would be better written as:
<CODE BEGINS>
if err := SomeFunction(); err != nil {
return err
}
// do something
<CODE ENDS>
The opening brace on the first line must be positioned on the same
line as the "if" statement. There is no arguing about this, because
this is what "gofmt" outputs.
4.6.2. Goto
Go has a "goto" "goto" you jump to a
<CODE BEGINS>
func myfunc() {
i := 0
Here:
fmt.Println(i)
i++
goto Here
}
<CODE ENDS>
The string "Here:" indicates a label. A label does not need to start
with a capital letter and is case sensitive.
4.6.3. For
The Go "for"
* "for init; condition; post { }" - a loop using the syntax borrowed
from C;
* "for condition { }" - a while loop, and;
* "for { }" - an endless loop.
Short declarations make it easy to declare the index variable right
in the loop.
<CODE BEGINS>
sum := 0
for i := 0; i < 10; i++ {
sum = sum + i
}
<CODE ENDS>
Note that the variable "i" ceases to exist after the loop.
4.6.4. Break and Continue
With "break" "break" breaks the current loop.
<CODE BEGINS>
for i := 0; i < 10; i++ {
if i > 5 {
break <1>
}
fmt.Println(i) <2>
}
<CODE ENDS>
Here we "break" the current loop _1_, and don't continue with the
"fmt.Println(i)" statement _2_. So we only print 0 to 5. With loops
within loop you can specify a label after "break" to identify _which_
loop to stop:
<CODE BEGINS>
J: for j := 0; j < 5; j++ { <1>
for i := 0; i < 10; i++ {
if i > 5 {
break J <2>
}
fmt.Println(i)
}
}
<CODE ENDS>
Here we define a label "J" _1_, preceding the "for"-loop there. When
we use "break J" _2_, we don't break the inner loop but the "J" loop.
With "continue" "break", "continue" also accepts a label.
4.6.5. Range
The keyword "range" Section 9.3.1). "range" is an iterator that, when
called, returns the next key-value pair from the "thing" it loops
over. Depending on what that is, "range" returns different things.
When looping over a slice or array, "range" returns the index in the
slice as the key and value belonging to that index. Consider this
code:
<CODE BEGINS>
list := []string{"a", "b", "c", "d", "e", "f"}
for k, v := range list {
// do something with k and v
}
<CODE ENDS>
First we create a slice of strings. Then we use "range" to loop over
them. With each iteration, "range" will return the index as an "int"
and the key as a "string". It will start with 0 and "a", so "k" will
be 0 through 5, and v will be "a" through "f".
You can also use "range" on strings directly. Then it will break out
the individual Unicode characters ^[In the UTF-8 world characters are
sometimes called _runes_ "char" is "rune". and their start position,
by parsing the UTF-8. The loop:
<CODE BEGINS>
for pos, char := range "Gő!" {
fmt.Printf("character '%c' starts at byte position %d\n", char, pos)
}
<CODE ENDS>
prints
character 'G' starts at byte position 0
character 'ő' starts at byte position 1
character '!' starts at byte position 3
Note that "ő" took 2 bytes, so '!' starts at byte 3.
4.6.6. Switch
Go's "switch" "switch" has no expression it switches on "true". It's
therefore possible -- and idiomatic -- to write an "if-else-if-else"
chain as a "switch".
<CODE BEGINS>
// Convert hexadecimal character to an int value
switch { <1>
case '0' <= c && c <= '9': <2>
return c - '0' <3>
case 'a' <= c && c <= 'f': <4>
return c - 'a' + 10
case 'A' <= c && c <= 'F': <5>
return c - 'A' + 10
}
return 0
<CODE ENDS>
A "switch" without a condition is the same as "switch true" _1_. We
list the different cases. Each "case" statement has a condition that
is either true of false. Here _2_ we check if "c" is a number. If
"c" is a number we return its value _3_. Check if "c" falls between
"a" and "f" _4_. For an "a" we return 10, for "b" we return 11, etc.
We also do the same _5_ thing for "A" to "F".
There is no automatic fall through, you can use "fallthrough"
<CODE BEGINS>
switch i {
case 0: fallthrough
case 1: <1>
f()
default:
g() <2>
<CODE ENDS>
"f()" can be called when "i == 0" _1_. With "default" "g()" is called
when "i" is not 0 or 1 _2_. We could rewrite the above example as:
<CODE BEGINS>
switch i {
case 0, 1: <1>
f()
default:
g()
<CODE ENDS>
You can list cases on one line _1_, separated by commas.
4.7. Built-in Functions
A few functions are predefined, meaning you _don't_ have to include
any package to get access to them. Table 3 lists them all.
+----------+----------+-----------+-----------+
+==========+==========+===========+===========+
| "close" | "new" | "panic" | "complex" |
+----------+----------+-----------+-----------+
| "delete" | "make" | "recover" | "real" |
+----------+----------+-----------+-----------+
| "len" | "append" | "print" | "imag" |
+----------+----------+-----------+-----------+
| "cap" | "copy" | "println" | |
+----------+----------+-----------+-----------+
Table 3: Pre-defined functions in Go.
These built-in functions are documented in the "builtin"
"close"
is used in channel communication. It closes a channel. We'll
learn more about this in Section 9.3.1.
"delete"
is used for deleting entries in maps.
"len" and "cap"
are used on a number of different types, "len" is used to return
the lengths of strings, slices, and arrays. In the next section
Section 4.8.1 we'll look at slices, arrays and the function "cap".
"new"
is used for allocating memory for user defined data types. See
Section 7.1.1.
"make"
is used for allocating memory for built-in types (maps, slices,
and channels). See Section 7.1.2.
"copy", "append"
"copy" is for copying slices. "append" is for concatenating
slices. See Section 4.8.2 in this chapter.
"panic", "recover"
are used for an _exception_ mechanism. See Section 5.6 for more.
"print", "println"
are low level printing functions that can be used without
reverting to the "fmt"
"complex", "real", "imag"
all deal with complex numbers.
4.8. Arrays, Slices, and Maps
To store multiple values in a list, you can use arrays, or their more
flexible cousin: slices. A dictionary or hash type is also
available. It is called a "map" in Go.
4.8.1. Arrays
An array is defined by: "[n]<type>", where "n" is the length of the
array and "<type>" is the stuff you want to store. To assign or
index an element in the array, you use square brackets:
<CODE BEGINS>
var arr [10]int
arr[0] = 42
arr[1] = 13
fmt.Printf("The first element is %d\n", arr[0])
<CODE ENDS>
Array types like "var arr [10]int" have a fixed size. The size is
_part_ of the type. They can't grow, because then they would have a
different type. Also arrays are values: Assigning one array to
another _copies_ all the elements. In particular, if you pass an
array to a function it will receive a copy of the array, not a
pointer to it.
"var a [3]int". To initialize it to something other than zero, use a
_composite literal_ "a := [3]int{1, 2, 3}". This can be shortened to
"a := [...]int{1, 2, 3}", where Go counts the elements automatically.
<aside>:
A composite literal allows you to assign a value directly to an
array, slice, or map. See Section 7.1.3 for more information.
When declaring arrays you _always_ have to type something in between
the square brackets, either a number or three dots ("..."), when
using a composite literal. When using multidimensional arrays, you
can use the following syntax: "a := [2][2]int{ {1,2}, {3,4} }". Now
that you know about arrays you will be delighted to learn that you
will almost never use them in Go, because there is something much
more flexible: slices.
4.8.2. Slices
A slice is similar to an array, but it can grow when new elements are
added. A slice always refers to an underlying array. What makes
slices different from arrays is that a slice is a pointer _to_ an
array; slices are reference types.
<aside>:
Reference types are created with "make". We detail this further in
Section 7.
That means that if you assign one slice to another, both refer to the
_same_ underlying array. For instance, if a function takes a slice
argument, changes it makes to the elements of the slice will be
visible to the caller, analogous to passing a pointer to the
underlying array. With: "slice := make([]int, 10)", you create a
slice which can hold ten elements. Note that the underlying array
isn't specified. A slice is always coupled to an array that has a
fixed size. For slices we define a capacity Section 4.8.2 shows the
creation of an array, then the creation of a slice. First we create
an array of "m" elements of the type "int": "var array[m]int" .
Next, we create a slice from this array: "slice := array[:n]" . And
now we have:
* "len(slice) == n"
* "cap(slice) == m"
* "len(array) == cap(array) == m"
Array versus slice
Given an array, or another slice, a new slice is created via
"a[n:m]". This creates a new slice which refers to the variable "a",
starts at index "n", and ends before index "m". It has length "n -
m".
<CODE BEGINS>
a := [...]int{1, 2, 3, 4, 5} <1>
s1 := a[2:4] <2>
s2 := a[1:5] <3>
s3 := a[:] <4>
s4 := a[:4] <5>
s5 := s2[:] <6>
s6 := a[2:4:5] <7>
<CODE ENDS>
First we define _1_ an array with five elements, from index 0 to 4.
From this we create _2_ a slice with the elements from index 2 to 3,
this slices contains: "3, 4". Then we we create another slice _3_
from "a": with the elements from index 1 to 4, this contains: "2, 3,
4, 5". With "a[:]" _4_ we create a slice with all the elements in
the array. This is a shorthand for: "a[0:len(a)]". And with "a[:4]"
_5_ we create a slice with the elements from index 0 to 3, this is
short for: "a[0:4]", and gives us a slices that contains: "1, 2, 3,
4". With "s2[:]" we create a slice from the slice "s2" _6_, note
that "s5" still refers to the array "a". Finally, we create a slice
with the elements from index 3 to 3 _and_ also set the cap to 4 _7_.
When working with slices you can overrun the bounds, consider this
code.
<CODE BEGINS>
package main
func main() {
var array [100]int <1>
slice := array[0:99] <2>
slice[98] = 1 <3>
slice[99] = 2 <4>
}
<CODE ENDS>
At _1_ we create an array with a 100 elements, indexed from 0 to 99.
Then at _2_ we create a slice that has index 0 to 98. We assign 1 to
the 99th element _3_ of the slice. This works as expected. But at
_4_ we dare to do the impossible, and and try to allocate something
beyond the length of the slice and we are greeted with a _runtime_
error: "Error: "throw: index out of range"."
If you want to extend a slice, there are a couple of built-in
functions that make life easier: "append" and "copy". The append
function appends zero or more values to a slice and returns the
result: a slice with the same type as the original. If the original
slice isn't big enough to fit the added values, append will allocate
a new slice that is big enough. So the slice returned by append may
refer to a different underlying array than the original slice does.
Here's an example:
<CODE BEGINS>
s0 := []int{0, 0}
s1 := append(s0, 2) <1>
s2 := append(s1, 3, 5, 7) <2>
s3 := append(s2, s0...) <3>
<CODE ENDS>
At _1_ we append a single element, making "s1" equal to "[]int{0, 0,
2}". At _2_ we append multiple elements, making "s2" equal to
"[]int{0, 0, 2, 3, 5, 7}". And at _3_ we append a slice, giving us
"s3" equal to "[]int{0, 0, 2, 3, 5, 7, 0, 0}". Note the three dots
used after "s0..."! This is needed make it clear explicit that
you're appending another slice, instead of a single value.
The copy function copies slice elements from a source to a
destination, and returns the number of elements it copied. This
number is the minimum of the length of the source and the length of
the destination. For example:
<CODE BEGINS>
var a = [...]int{0, 1, 2, 3, 4, 5, 6, 7}
var s = make([]int, 6)
n1 := copy(s, a[0:]) <1>
n2 := copy(s, s[2:]) <2>
<CODE ENDS>
After _1_, "n1" is 6, and "s" is "[]int{0, 1, 2, 3, 4, 5}". And
after _2_, "n2" is 4, and "s" is "[]int{2, 3, 4, 5, 4, 5}".
4.8.3. Maps
Many other languages have a type similar to maps built-in. For
instance, Perl has hashes, Python has its dictionaries, and C++ also
has maps (as part of the libraries). In Go we have the "map" "map"
can be thought of as an array indexed by strings (in its most simple
form).
<CODE BEGINS>
monthdays := map[string]int{
"Jan": 31, "Feb": 28, "Mar": 31,
"Apr": 30, "May": 31, "Jun": 30,
"Jul": 31, "Aug": 31, "Sep": 30,
"Oct": 31, "Nov": 30, "Dec": 31, <1>
}
<CODE ENDS>
The general syntax for defining a map is "map[<from type>]<to type>".
Here, we define a map that converts from a "string" (month
abbreviation) to an "int" (number of days in that month). Note that
the trailing comma at _1_ is _required_.
Use "make" when only declaring a map: "monthdays :=
make(map[string]int)". A map is a reference type.
For indexing ("searching") the map, we use square brackets. For
example, suppose we want to print the number of days in December:
"fmt.Printf("%d\n", monthdays["Dec"])"
If you are looping over an array, slice, string, or map a, "range"
<CODE BEGINS>
year := 0
for _, days := range monthdays <1>
year += days
}
fmt.Printf("Numbers of days in a year: %d\n", year)
<CODE ENDS>
At _1_ we use the underscore to ignore (assign to nothing) the key
returned by "range". We are only interested in the values from
"monthdays".
To add elements to the map, you would add new month with:
"monthdays["Undecim"] = 30". If you use a key that already exists,
the value will be silently overwritten: "monthdays["Feb"] = 29". To
test for existence "value, present := monthdays["Jan"]". If the key
"Jan" exists, "present" will be true. It's more Go like to name
"present" "ok", and use: "v, ok := monthdays["Jan"]". In Go we call
this the "comma ok" form.
You can remove elements "map": "delete(monthdays, "Mar")" . In
general the syntax "delete(m, x)" will delete the map entry retrieved
by the expression "m[x]".
4.9. Exercises
4.9.1. For-loop
1. Create a loop with the "for" construct. Make it loop 10 times
and print out the loop counter with the "fmt" package.
2. Rewrite the loop from 1 to use "goto". The keyword "for" may not
be used.
3. Rewrite the loop again so that it fills an array and then prints
that array to the screen.
4.9.2. Answer
1. There are many possibilities. One solution could be:
<CODE BEGINS>
package main
import "fmt"
func main() {
for i := 0; i < 10; i++ {
fmt.Println("%d", i)
}
}
<CODE ENDS>
Let's compile this and look at the output.
% go build for.go
% ./for
0
1
.
.
.
9
2. Rewriting the loop results in code that should look something
like this (only showing the "main"-function):
<CODE BEGINS>
func main() {
i := 0 <1>
Loop: <2>
if i < 10 {
fmt.Printf("%d\n", i)
i++
goto Loop <3>
}
}
<CODE ENDS>
At _1_ we define our loop variable. And at _2_ we define a label
and at _3_ we jump to this label.
3. The following is one possible solution:
<CODE BEGINS>
package main
import "fmt"
func main() {
var arr [10]int <1>
for i := 0; i < 10; i++ {
arr[i] = i <2>
}
fmt.Printf("%v", arr) <3>
}
<CODE ENDS>
Here _1_ we create an array with 10 elements. Which we then fill
_2_ one by one. And finally we print it _3_ with "%v" which lets
Go to print the value for us. You could even do this in one fell
swoop by using a composite literal:
<CODE BEGINS>
fmt.Printf("%v\n", [...]int{0,1,2,3,4,5,6,7,8,9})
<CODE ENDS>
4.9.3. Average
1. Write code to calculate the average of a "float64" slice. In a
later exercise you will make it into a function.
4.9.4. Answer
1. The following code calculates the average.
<CODE BEGINS>
sum := 0.0
switch len(xs) {
case 0: <1>
avg = 0
default: <2>
for _, v := range xs {
sum += v
}
avg = sum / float64(len(xs)) <3>
}
<CODE ENDS>
Here at _1_ we check if the length is zero and if so, we return 0.
Otherwise we calculate the average at _2_. We have to convert the
value return from "len" to a "float64" to make the division work at
_3_.
4.9.5. FizzBuzz
1. Solve this problem, called the Fizz-Buzz [cite_fizzbuzz] problem:
Write a program that prints the numbers from 1 to 100. But for
multiples of three print, "Fizz" instead of the number, and for
multiples of five, print "Buzz". For numbers which are multiples of
both three and five, print "FizzBuzz".
4.9.6. Answer
1. A possible solution to this problem is the following program.
<CODE BEGINS>
package main
import "fmt"
func main() {
const (
FIZZ = 3 <1>
BUZZ = 5
)
var p bool <2>
for i := 1; i < 100; i++ { <3>
p = false
if i%FIZZ == 0 { <4>
fmt.Printf("Fizz")
p = true
}
if i%BUZZ == 0 { <5>
fmt.Printf("Buzz")
p = true
}
if !p { <6>
fmt.Printf("%v", i)
}
fmt.Println()
}
}
<CODE ENDS>
Here _1_ we define two constants to make our code more readable, see
Section 4.3.3. At _2_ we define a boolean that keeps track if we
already printed something. At _3_ we start our for-loop, see
Section 4.6.3. If the value is divisible by FIZZ - that is, 3 - , we
print "Fizz" _4_. And at _5_ we check if the value is divisble by
BUZZ -- that is, 5 -- if so print "Buzz". Note that we have also
taken care of the FizzBuzz case. At _6_, if printed neither Fizz nor
Buzz printed, we print the value.
5. Functions
<blockquote>:
I'm always delighted by the light touch and stillness of early
programming languages. Not much text; a lot gets done. Old programs
read like quiet conversations between a well-spoken research worker
and a well- studied mechanical colleague, not as a debate with a
compiler. Who'd have guessed sophistication bought such noise?
Functions are the basic building blocks of Go programs; all
interesting stuff happens in them.
Here is an example of how you can declare a function:
<CODE BEGINS>
type mytype int
func (p mytype) funcname(q int) (r,s int) { return 0,0 }
<1> <2> <3> <4> <5> <6>
<CODE ENDS>
To declare a function, you use the "func" keyword _1_. You can
optionally bind _2_ to a specific type called receiver Section 8.
Next _3_ you write the name of your function. Here _4_ we define
that the variable "q" of type "int" is the input parameter.
Parameters are passed _pass-by-value_."r" and "s" _5_ are the _named
return parameters_ (((functions, named return parameters))) for this
function. Functions in Go can have multiple return values. This is
very useful to return a value _and_ error. This removes the need for
in-band error returns (such as -1 for "EOF") and modifying an
argument. If you want the return parameters not to be named you only
give the types: "(int, int)". If you have only one value to return
you may omit the parentheses. If your function is a subroutine and
does not have anything to return you may omit this entirely.
Finally, we have the body _6_ of the function. Note that "return" is
a statement so the braces around the parameter(s) are optional.
As said the return or result parameters of a Go function can be given
names and used as regular variables, just like the incoming
parameters. When named, they are initialized to the zero values for
their types when the function begins. If the function executes a
"return" statement with no arguments, the current values of the
result parameters are returned. Using these features enables you
(again) to do more with less code.
The names are not mandatory but they can make code shorter and
clearer: _they are documentation_. However don't overuse this
feature, especially in longer functions where it might not be
immediately apparent what is returned.
Functions can be declared in any order you wish. The compiler scans
the entire file before execution, so function prototyping is a thing
of the past in Go. Go does not allow nested functions, but you can
work around this with anonymous functions. See the
Section Section 5.2 in this chapter. Recursive functions work just
as in other languages:
<CODE BEGINS>
func rec(i int) {
if i == 10 { <1>
return
}
rec(i+1) <2>
fmt.Printf("%d ", i)
}
<CODE ENDS>
Here _2_ we call the same function again, "rec" returns when "i" has
the value 10, this is checked on the second line _1_. This function
prints: "9 8 7 6 5 4 3 2 1 0", when called as "rec(0)".
5.1. Scope
Variables declared outside any functions are _global_ _local_
In the following example we call "g()" from "f()":
<CODE BEGINS>
package main
var a int <1>
func main() {
a = 5
print(a)
f()
}
func f() {
a := 6 <2>
print(a)
g()
}
func g() {
print(a)
}
<CODE ENDS>
Here _1_, we declare "a" to be a global variable of type "int". Then
in the "main" function we give the _global_ "a" the value of 5, after
printing it we call the function "f". Then here _2_, "a := 6", we
create a _new, local_ variable also called "a". This new "a" gets
the value of 6, which we then print. Then we call "g", which uses
the _global_ "a" again and prints "a"'s value set in "main". Thus
the output will be: "565". A _local_ variable is _only_ valid when
we are executing the function in which it is defined. Note that the
":=" used in line 12 is sometimes hard to spot so it is generally
advised _not_ to use the same name for global and local variables.
5.2. Functions as values
_just_ values. They can be assigned to variables as follows:
<CODE BEGINS>
import "fmt"
func main() {
a := func() { <1>
fmt.Println("Hello")
} <2>
a() <3>
}
<CODE ENDS>
"a" is defined as an anonymous (nameless) function _1_. Note the lack
of parentheses "()" after "a". If there were, that would be to
_call_ some function with the name "a" before we have defined what
"a" is. Once "a" is defined, then we can _call_ it, _3_.
Functions--as--values may be used in other places, for example maps.
Here we convert from integers to functions:
<CODE BEGINS>
var xs = map[int]func() int{
1: func() int { return 10 },
2: func() int { return 20 },
3: func() int { return 30 },
}
<CODE ENDS>
Note that the final comma on second to last line is _mandatory_.
Or you can write a function that takes a function as its parameter,
for example a "Map" function that works on "int" slices. This is
left as an exercise for the reader; see the exercise Section 5.7.15.
5.3. Callbacks
Because functions are values they are easy to pass to functions, from
where they can be used as callbacks. First define a function that
does "something" with an integer value:
<CODE BEGINS>
func printit(x int) {
fmt.Printf("%v\n", x)
}
<CODE ENDS>
This function does not return a value and just prints its argument.
The _signature_ "func printit(int)", or without the function name:
"func(int)". To create a new function that uses this one as a
callback we need to use this signature:
<CODE BEGINS>
func callback(y int, f func(int)) {
f(y)
}
<CODE ENDS>
Here we create a new function that takes two parameters: "y int",
i.e. just an "int" and "f func(int)", i.e. a function that takes an
int and returns nothing. The parameter "f" is the variable holding
that function. It can be used as any other function, and we execute
the function on line 2 with the parameter "y": "f(y)"
5.4. Deferred Code
Suppose you have a function in which you open a file and perform
various writes and reads on it. In such a function there are often
spots where you want to return early. If you do that, you will need
to close the file descriptor you are working on. This often leads to
the following code:
<CODE BEGINS>
func ReadWrite() bool {
file.Open("file")
// Do your thing
if failureX {
file.Close() <1>
return false
}
if failureY {
file.Close() <1>
return false
}
file.Close() <1>
return true <2>
}
<CODE ENDS>
Note that we repeat a lot of code here; you can see the that
"file.Close()" is called at _1_. To overcome this, Go has the "defer"
"defer" you specify a function which is called just _before_ _2_ the
current function exits.
With "defer" we can rewrite the above code as follows. It makes the
function more readable and it puts the "Close" _right next_ to the
"Open".
<CODE BEGINS>
func ReadWrite() bool {
file.Open("filename")
defer file.Close() <1>
// Do your thing
if failureX {
return false <2>
}
if failureY {
return false <2>
}
return true <2>
}
<CODE ENDS>
At _1_ "file.Close()" is added to the defer list. "Close" is now
done automatically at _2_. This makes the function shorter and more
readable. It puts the "Close" right next to the "Open".
You can put multiple functions on the "defer list", like this example
from
<CODE BEGINS>
for i := 0; i < 5; i++ {
defer fmt.Printf("%d ", i)
}
<CODE ENDS>
Deferred functions are executed in LIFO order, so the above code
prints: "4 3 2 1 0".
With "defer" you can even change return values, provided that you are
using named result parameters and a function literal
<CODE BEGINS>
defer func() {/* ... */}()
<CODE ENDS>
Here we use a function without a name and specify the body of the
function inline, basically we're creating a nameless function on the
spot. The final braces are needed because "defer" needs a function
call, not a function value. If our anonymous function would take an
parameter it would be easier to see why we need the braces:
<CODE BEGINS>
defer func(x int) {/* ... */}(5)
<CODE ENDS>
In this (unnamed) function you can access any named return parameter:
<CODE BEGINS>
func f() (ret int)
defer func() { <1>
ret++
}()
return 0
}
<CODE ENDS>
Here _1_ we specify our function, the named return value "ret" is
initialized with zero. The nameless function in the defer increments
the value of "ret" with 1. The "return 0" on line 5 _will not be the
returned value_, because of "defer". The function "f" will return 1!
5.5. Variadic Parameter
Functions that take a variable number of parameters are known as
variadic functions.
<CODE BEGINS>
func myfunc(arg ...int) {}
<CODE ENDS>
The "arg ...int" instructs Go to see this as a function that takes a
variable number of arguments. Note that these arguments all have to
have the type "int". In the body of your function the variable "arg"
is a slice of ints:
<CODE BEGINS>
for _, n := range arg {
fmt.Printf("And the number is: %d\n", n)
}
<CODE ENDS>
We range over the arguments on the first line. We are not interested
in the index as returned by "range", hence the use of the underscore
there. In the body of the "range" we just print the parameters we
were given.
If you don't specify the type of the variadic argument it defaults to
the empty interface "interface{}" (see Chapter Section 8).
Suppose we have another variadic function called "myfunc2", the
following example shows how to pass variadic arguments to it:
<CODE BEGINS>
func myfunc(arg ...int) {
myfunc2(arg...)
myfunc2(arg[:2]...)
}
<CODE ENDS>
With "myfunc2(arg...)" we pass all the parameters to "myfunc2", but
because the variadic parameters is just a slice, we can use some
slice tricks as well.
5.6. Panic and recovering
Go does not have an exception mechanism: you cannot throw exceptions.
Instead it uses a panic-and-recover mechanism. It is worth
remembering that you should use this as a last resort, your code will
not look, or be, better if it is littered with panics. It's a
powerful tool: use it wisely. So, how do you use it? In the words
of the Go Authors [go_blog_panic]:
Panic
is a built-in function that stops the ordinary flow of
control and begins panicking. When the function "F" calls
"panic", execution of "F" stops, any deferred functions in
"F" are executed normally, and then "F" returns to its
caller. To the caller, "F" then behaves like a call to
"panic". The process continues up the stack until all
functions in the current goroutine have returned, at which
point the program crashes. Panics can be initiated by
invoking "panic" directly. They can also be caused by
_runtime errors_, such as out-of-bounds array accesses.
Recover
is a built-in function that regains control of a panicking
goroutine. Recover is _only_ useful inside _deferred_
functions. During normal execution, a call to "recover"
will return "nil" and have no other effect. If the current
goroutine is panicking, a call to "recover" will capture the
value given to "panic" and resume normal execution.
This function checks if the function it gets as argument will panic
when it is executed:
<CODE BEGINS>
func Panic(f func()) (b bool) { <1>
defer func() { <2>
if x := recover(); x != nil {
b = true
}
}()
f() <3>
return <4>
}
<CODE ENDS>
We define a new function "Panic" _1_ that takes a function as an
argument (see Section 5.2). It returns true if "f" panics when run,
else false. We then _2_ define a "defer" function that utilizes
"recover". If the current goroutine panics, this defer function will
notice that. If "recover()" returns non-"nil" we set "b" to true.
At _3_ Execute the function we received as the argument. And finally
_4_ we return the value of "b". Because "b" is a named return
parameter.
The following code fragment, shows how we can use this function:
<CODE BEGINS>
func panicy() {
var a []int
a[3] = 5
}
func main() {
fmt.Println(Panic(panicy))
}
<CODE ENDS>
On line 3 the "a[3] = 5" triggers a _runtime_ out of bounds error
which results in a panic. Thus this program will print "true". If
we change line 2: "var a []int" to "var a [4]int" the function
"panicy" does not panic anymore. Why?
5.7. Exercises
5.7.1. Average
1. Write a function that calculates the average of a "float64"
slice.
5.7.2. Answer
1. The following function calculates the average:
<CODE BEGINS>
package main
func average(xs []float64) (avg float64) { //<1>
sum := 0.0
switch len(xs) {
case 0: //<2>
avg = 0
default: //<3>
for _, v := range xs {
sum += v
}
avg = sum / float64(len(xs)) //<4>
}
return //<5>
}
<CODE ENDS>
At _1_ we use a named return parameter. If the length of "xs" is
zero _2_, we return 0. Otherwise _3_, we calculate the average. At
_4_ we convert the value to a "float64" to make the division work as
"len" returns an "int". Finally, at _5_ we reutrn our avarage.
5.7.3. Bubble sort
1. Write a function that performs a bubble sort on a slice of ints.
From [bubblesort]:
<blockquote>:
It works by repeatedly stepping through the list to be sorted,
comparing each pair of adjacent items and swapping them if they are
in the wrong order. The pass through the list is repeated until no
swaps are needed, which indicates that the list is sorted. The
algorithm gets its name from the way smaller elements "bubble" to the
top of the list.
It also gives an example in pseudo code:
procedure bubbleSort( A : list of sortable items )
do
swapped = false
for each i in 1 to length(A) - 1 inclusive do:
if A[i-1] > A[i] then
swap( A[i-1], A[i] )
swapped = true
end if
end for
while swapped
end procedure
5.7.4. Answer
1. Bubble sort isn't terribly efficient. For "n" elements it scales
"O(n^2)". But bubble sort is easy to implement:
<CODE BEGINS>
func main() {
n := []int{5, -1, 0, 12, 3, 5}
fmt.Printf("unsorted %v\n", n)
bubblesort(n)
fmt.Printf("sorted %v\n", n)
}
func bubblesort(n []int) {
for i := 0; i < len(n)-1; i++ {
for j := i + 1; j < len(n); j++ {
if n[j] < n[i] {
n[i], n[j] = n[j], n[i]
}
<CODE ENDS>
Because a slice is a reference type, the "bubblesort" function
works and does not need to return a sorted slice.
5.7.5. For-loop II
1. Take what you did in exercise to write the for loop and extend it
a bit. Put the body of the for loop - the "fmt.Printf" - in a
separate function.
5.7.6. Answer
1. <{{src/for-func.go}}
5.7.7. Fibonacci
1. The Fibonacci sequence starts as follows: "1, 1, 2, 3, 5, 8, 13,
\ldots" Or in mathematical terms: "x_1 = 1; x_2 = 1; x_n =
x_{n-1} + x_{n-2}\quad\forall n > 2".
Write a function that takes an "int" value and gives that many
terms of the Fibonacci sequence.
5.7.8. Answer
1. The following program calculates Fibonacci numbers:
<CODE BEGINS>
package main
import "fmt"
func fibonacci(value int) []int {
x := make([]int, value) <1>
x[0], x[1] = 1, 1 <2>
for n := 2; n < value; n++ {
x[n] = x[n-1] + x[n-2] <3>
}
return x <4>
}
func main() {
for _, term := range fibonacci(10) { <5>
fmt.Printf("%v ", term)
}
}
<CODE ENDS>
At _1_ we create an array to hold the integers up to the value given
in the function call. At _2_ we start the Fibonacci calculation.
Then _3_: "x_n = x_{n-1} + x_{n-2}". At _4_ we return the _entire_
array. And at _5_ we use the "range" keyword to "walk" the numbers
returned by the Fibonacci function. Here up to 10. Finally, we
print the numbers.
5.7.9. Var args
1. Write a function that takes a variable number of ints and print
each integer on a separate line.
5.7.10. Answer
1. For this we need the "{...}"-syntax to signal we define a
function that takes an arbitrary number of arguments.
<CODE BEGINS>
package main
import "fmt"
func main() {
printthem(1, 4, 5, 7, 4)
printthem(1, 2, 4)
}
func printthem(numbers ...int) {
for _, d := range numbers {
fmt.Printf("%d\n", d)
}
}
<CODE ENDS>
5.7.11. Functions that return functions
1. Write a function that returns a function that performs a "+2" on
integers. Name the function "plusTwo". You should then be able
do the following:
<CODE BEGINS>
p := plusTwo()
fmt.Printf("%v\n", p(2))
<CODE ENDS>
Which should print 4. See Section 5.3.
2. Generalize the function from above and create a "plusX(x)" which
returns functions that add "x" to an integer.
5.7.12. Answer
1. Define a new function that returns a function: "return func(x
int) int { return x + 2 }" Function literals at work, we define
the +2--function right there in the return statement.
<CODE BEGINS>
func main() {
p2 := plusTwo()
fmt.Printf("%v\n",p2(2))
}
func plusTwo() func(int) int { <1>
return func(x int) int { return x + 2 } <2>
}
<CODE ENDS>
2. Here we use a closure:
<CODE BEGINS>
func plusX(x int) func(int) int { <1>
return func(y int) int { return x + y } <2>
}
<CODE ENDS>
Here _1_, we again define a function that returns a function. We
use the _local_ variable "x" in the function literal at _2_.
5.7.13. Maximum
1. Write a function that finds the maximum value in an "int" slice
("[]int").
5.7.14. Answer
1. This function returns the largest int in the slice \var{l}:
<CODE BEGINS>
func max(l []int) (max int) { <1>
max = l[0]
for _, v := range l { <2>
if v > max { <3>
max = v
}
}
return <4>
}
<CODE ENDS>
At _1_ we use a named return parameter. At _2_ we loop over "l".
The index of the element is not important. At _3_, if we find a
new maximum, we remember it. And at _4_ we have a "lone" return;
the current value of "max" is now returned.
5.7.15. Map function
A "map()"-function is a function that takes a function and a list.
The function is applied to each member in the list and a new list
containing these calculated values is returned. Thus:
"\mathrm{map}(f(), (a_1,a_2,\ldots,a_{n-1},a_n)) = (f(a_1),
f(a_2),\ldots,f(a_{n-1}), f(a_n))"
1. Write a simple "map()"-function in Go. It is sufficient for this
function only to work for ints.
5.7.16. Answer
1. A possible answer:
<CODE BEGINS>
func Map(f func(int) int, l []int) []int {
j := make([]int, len(l))
for k, v := range l {
j[k] = f(v)
}
return j
}
func main() {
m := []int{1, 3, 4}
f := func(i int) int {
return i * i
}
fmt.Printf("%v", (Map(f, m)))
}
<CODE ENDS>
5.7.17. Stack
1. Create a simple stack which can hold a fixed number of ints. It
does not have to grow beyond this limit. Define "push" -- put
something on the stack -- and "pop" -- retrieve something from
the stack -- functions. The stack should be a LIFO (last in,
first out) stack.
A stack
2. Write a "String" method which converts the stack to a string
representation. The stack in the figure could be represented as:
"[0:m] [1:l] [2:k]" .
5.7.18. Answer
1. First we define a new type that represents a stack; we need an
array (to hold the keys) and an index, which points to the last
element. Our small stack can only hold 10 elements.
<CODE BEGINS>
type stack struct {
i int
data [10]int
}
<CODE ENDS>
Next we need the "push" and "pop" functions to actually use the
thing. First we show the _wrong_ solution!
In Go, data passed to functions is _passed-by-value_ meaning a copy
is created and given to the function. The first stab for the
function "push" could be:
<CODE BEGINS>
func (s stack) push(k int) {
if s.i+1 > 9 {
return
}
s.data[s.i] = k
s.i++
}
<CODE ENDS>
The function works on the "s" which is of the type "stack". To use
this we just call "s.push(50)", to push the integer 50 on the stack.
But the push function gets a copy of "s", so it is _not_ working on
the _real_ thing. Nothing gets pushed to our stack. For example the
following code:
<CODE BEGINS>
var s stack
s.push(25)
fmt.Printf("stack %v\n", s);
s.push(14)
fmt.Printf("stack %v\n", s);
<CODE ENDS>
prints:
stack [0:0]
stack [0:0]
To solve this we need to give the function "push" a pointer to the
stack. This means we need to change "push" from
<CODE BEGINS>
func (s stack) push(k int)
<CODE ENDS>
to
<CODE BEGINS>
func (s *stack) push(k int).
<CODE ENDS>
We should now use "new()" (see Section 7.1.1). in Section 7 to create
a _pointer_ to a newly allocated "stack", so line 1 from the example
above needs to be "s := new(stack)" .
And our two functions become:
<CODE BEGINS>
func (s *stack) push(k int) {
s.data[s.i] = k
s.i++
}
func (s *stack) pop() int {
s.i--
ret := s.data[s.i]
s.data[s.i] = 0
return ret
}
<CODE ENDS>
Which we then use as follows:
<CODE BEGINS>
func main() {
var s stack
s.push(25)
s.push(14)
fmt.Printf("stack %v\n", s)
}
<CODE ENDS>
2. "fmt.Printf("%v")" can print any value ("%v") that satisfies the
"Stringer" interface (see Section 8). For this to work we only
need to define a "String()" function for our type:
<CODE BEGINS>
func (s stack) String() string {
var str string
for i := 0; i <= s.i; i++ {
str = str + "[" +
strconv.Itoa(i) + ":" + strconv.Itoa(s.data[i]) + "]"
}
return str
}
<CODE ENDS>
6. Packages
<blockquote>:
"^(")
A package is a collection of functions and data.
You declare a package with the "package""package <name>" line. Let's
define a package "even" in the file "even.go".
<CODE BEGINS>
package even <1>
func Even(i int) bool { <2>
return i%2 == 0
}
func odd(i int) bool { <3>
return i%2 == 1
}
<CODE ENDS>
Here _1_ we start a new namespace: "even". The function "Even" _2_
starts with a capital letter. This means the function is _exported_,
and may be used outside our package (more on that later). The
function "odd" _3_ does not start with a capital letter, so it is a
_private_ function.
Now we just need to build the package. We create a directory under
"$GOPATH", and copy "even.go" there (see Section 4.2 in Section 4).
% mkdir $GOPATH/src/even
% cp even.go $GOPATH/src/even
% go build
% go install
Now we can use the package in our own program "myeven.go":
<CODE BEGINS>
package main
import ( <1>
"even" <2>
"fmt" <3>
)
func main() {
i := 5
fmt.Printf("Is %d even? %v\n", i, even.Even(i)) <4>
}
<CODE ENDS>
Import _1_ the following packages. The _local_ package "even" is
imported here _2_. This _3_ imports the official "fmt" package. And
now we use _4_ the function from the "even" package. The syntax for
accessing a function from a package is "<package>.FunctionName()".
And finally we can build our program.
% go build myeven.go
% ./myeven
Is 5 even? false
If we change our "myeven.go" at _4_ to use the unexported function
"even.odd": "fmt.Printf("Is %d even? %v\n", i, even.odd(i))" We get
an error when compiling, because we are trying to use a _private_
function:
myeven.go: cannot refer to unexported name even.odd
Note that the "starts with capital "\rightarrow" exported", "starts
with lower-case "\rightarrow" private" rule also extends to other
names (new types, global variables) defined in the package. Note
that the term "capital" is not limited to US-ASCII -- it extends to
all bicameral alphabets (Latin, Greek, Cyrillic, Armenian and
Coptic).
6.1. Identifiers
The Go standard library names some function with the old (Unix) names
while others are in CamelCase. The convention is to leave well-known
legacy not-quite-words alone rather than try to figure out where the
capital letters go: "Atoi", "Getwd", "Chmod". CamelCasing works best
when you have whole words to work with: "ReadFile", "NewWriter",
"MakeSlice". The convention in Go is to use CamelCase rather than
underscores to write multi-word names.
As we did above in our "myeven" program, accessing content from an
imported (with "import" "import "bytes"" the importing program can
talk about "bytes.Buffer". A package name should be good, short,
concise and evocative. The convention in Go is that package names
are lowercase, single word names.
The package name used in the "import" statement is the default name
used. But if the need arises (two different packages with the same
name for instance), you can override this default: "import bar
"bytes"" The function "Buffer" is now accessed as "bar.Buffer".
Another convention is that the package name is the base name of its
source directory; the package in "src/compress/gzip" is imported as
"compress/gzip" but has name "gzip", not "compress/gzip".
It is important to avoid stuttering when naming things. For
instance, the buffered reader type in the "bufio" "Reader", not
"BufReader", because users see it as "bufio.Reader", which is a
clear, concise name.
Similarly, the function to make new instances of "ring.Ring" (package
"container/ring"), would normally be called "NewRing", but since
"Ring" is the only type exported by the package, and since the
package is called "ring""New". Clients of the package see that as
"ring.New". Use the package structure to help you choose good names.
Another short example is "once.Do" (see package "sync");
"once.Do(setup)" reads well and would not be improved by writing
"once.DoOrWaitUntilDone(setup)". Long names don't automatically make
things more readable.
6.2. Documenting packages
When we created our "even" package, we skipped over an important
item: documentation. Each package should have a _package comment_, a
block comment preceding the "package" clause. In our case we should
extend the beginning of the package, with:
<CODE BEGINS>
// The even package implements a fast function for detecting if an integer
// is even or not.
package even
<CODE ENDS>
When running "go doc" this will show up at the top of the page. When
a package consists of multiple files the package comment should only
appear in one file. A common convention (in really big packages) is
to have a separate "doc.go" that only holds the package comment.
Here is a snippet from the official "regexp" package:
<CODE BEGINS>
/*
The regexp package implements a simple library for
regular expressions.
The syntax of the regular expressions accepted is:
regexp:
concatenation { '|' concatenation }
*/
package regexp
<CODE ENDS>
Each defined (and exported) function should have a small line of text
documenting the behavior of the function. Again to extend our "even"
package:
<CODE BEGINS>
// Even returns true of i is even. Otherwise false is returned.
func Even(i int) bool {
<CODE ENDS>
And even though "odd" is not exported, it's good form to document it
as well.
<CODE BEGINS>
// odd is the opposite of Even.
func odd(i int) bool {
<CODE ENDS>
6.3. Testing packages
In Go it is customary to write (unit) tests for your package.
Writing tests involves the "testing" package and the program "go
test"
The "go test" program runs all the test functions. Without any
defined tests for our "even" package, "go test" yields:
% go test
? even [no test files]
Let us fix this by defining a test in a test file. Test files reside
in the package directory and are named "*_test.go". Those test files
are just like other Go programs, but "go test" will only execute the
test functions. Each test function has the same signature and its
name should start with "Test": "func TestXxx(t *testing.T)" .
When writing test you will need to tell "go test" whether a test was
successful or not. A successful test function just returns. When
the test fails you can signal this with the following functions.
These are the most important ones (see "go doc testing" or "go help
testfunc" for more):
* "func (t *T) Fail()", "Fail" marks the test function as having
failed but continues execution.
* "func (t *T) FailNow()", "FailNow" marks the test function as
having failed and stops its execution. Any remaining tests in
this file are skipped, and execution continues with the next test.
* "func (t *T) Log(args ...interface{})", "Log" formats its
arguments using default formatting, analogous to "Print()", and
records the text in the error log.
* "func (t *T) Fatal(args ...interface{})", "Fatal" is equivalent to
"Log()" followed by "FailNow()".
Putting all this together we can write our test. First we pick a
name: "even_test.go". Then we add the following contents:
<CODE BEGINS>
package even <1>
import "testing" <2>
func TestEven(t *testing.T) { <3>
if !Even(2) {
t.Log("2 should be even!")
t.Fail()
}
}
<CODE ENDS>
A test file belongs to the current _1_ package. This is not only
convenient, but also allows tests of unexported functions and
structures. We then _2_ import the "testing" package. And finally
the test we want to execute. The code here _3_ should hold no
surprises: we check if the "Even" function works OK. And now, the
moment we have been waiting form executing the test.
% go test
ok even 0.001s
Our test ran and reported "ok". Success! If we redefine our test
function, we can see the result of a failed test:
<CODE BEGINS>
// Entering the twilight zone
func TestEven(t *testing.T) {
if Even(2) {
t.Log("2 should be odd!")
t.Fail()
}
}
<CODE ENDS>
We now get:
FAIL even 0.004s
--- FAIL: TestEven (0.00 seconds)
2 should be odd!
FAIL
And you can act accordingly (by fixing the test for instance).
Writing new packages should go hand in hand with writing (some)
documentation and test functions. It will make your code better and
it shows that you really put in the effort.
The Go test suite also allows you to incorporate example functions
which serve as documentation _and_ as tests. These functions need to
start with "Example".
<CODE BEGINS>
func ExampleEven() {
if Even(2) {
fmt.Printf("Is even\n")
}
// Output: <1>
// Is even
}
<CODE ENDS>
Those last two comments lines _1_ are part of the example, "go test"
uses those to check the _generated_ output with the text in the
comments. If there is a mismatch the test fails.
6.4. Useful packages
The standard libary of Go includes a huge number of packages. It is
very enlightening to browse the "$GOROOT/src/pkg" directory and look
at the packages. We cannot comment on each package, but the
following are worth a mention:
"fmt"
"fmt" implements formatted I/O with functions analogous to C's
"printf" and "scanf". The format verbs are derived from C's but
are simpler. Some verbs (%-sequences) that can be used:
* _%v_, the value in a default format. when printing structs, the
plus flag (%+v) adds field names.
* _%#v_, a Go-syntax representation of the value.
* _%T_, a Go-syntax representation of the type of the value.
"io"
"bufio"
"io.Reader" or "io.Writer" object, creating another object (Reader
or Writer) that also implements the interface but provides
buffering and some help for textual I/O.
"sort"
"sort" package provides primitives for sorting arrays and user-
defined collections.
"strconv"
"strconv" package implements conversions to and from string
representations of basic data types.
"os"
"os" package provides a platform-independent interface to
operating system functionality. The design is Unix-like.
"sync"
"sync" provides basic synchronization primitives such as mutual
exclusion locks.
"flag"
"flag" package implements command-line flag parsing.
"encoding/json"
"encoding/json" package implements encoding and decoding of JSON
objects as defined in RFC 4627 [RFC4627].
"html/template"
Templates are executed by applying them to a data structure.
Annotations in the template refer to elements of the data
structure (typically a field of a struct or a key in a map) to
control execution and derive values to be displayed. The template
walks the structure as it executes and the "cursor" @ represents
the value at the current location in the structure.
"net/http"
"net/http" package implements parsing of HTTP requests, replies,
and URLs and provides an extensible HTTP server and a basic HTTP
client.
"unsafe"
"unsafe" package contains operations that step around the type
safety of Go programs. Normally you don't need this package, but
it is worth mentioning that _unsafe_ Go programs are possible.
"reflect"
"reflect" package implements run-time reflection, allowing a
program to manipulate objects with arbitrary types. The typical
use is to take a value with static type "interface{}" and extract
its dynamic type information by calling "TypeOf", which returns an
object with interface type "Type". See Section 8,
Section Section 8.8.
"os/exec"
"os/exec" package runs external commands.
6.5. Exercises
6.5.1. Stack as package
1. See the Stack exercise. In this exercise we want to create a
separate package for that code. Create a proper package for your
stack implementation, "Push", "Pop" and the "Stack" type need to
be exported.
2. Write a simple unit test for this package. You should at least
test that a "Pop" works after a "Push".
6.5.2. Answer
1. There are a few details that should be changed to make a proper
package for our stack. First, the exported functions should
begin with a capital letter and so should "Stack". The package
file is named "stack-as-package.go" and contains:
<CODE BEGINS>
package stack
// Stack holds the items.
type Stack struct {
i int
data [10]int
}
// Push pushes an item on the stack.
func (s *Stack) Push(k int) {
s.data[s.i] = k
s.i++
}
// Pop pops an item from the stack.
func (s *Stack) Pop() (ret int) {
s.i--
ret = s.data[s.i]
return
}
<CODE ENDS>
2. To make the unit testing work properly you need to do some
preparations. We'll come to those in a minute. First the actual
unit test. Create a file with the name "pushpop_test.go", with
the following contents:
<CODE BEGINS>
package stack
import "testing"
func TestPushPop(t *testing.T) {
c := new(Stack)
c.Push(5)
if c.Pop() != 5 {
t.Log("Pop doesn't give 5")
t.Fail()
}
}
<CODE ENDS>
For "go test" to work we need to put our package files in a directory
under "$GOPATH/src":
% mkdir $GOPATH/src/stack
% cp pushpop_test.go $GOPATH/src/stack
% cp stack-as-package.go $GOPATH/src/stack
Yields:
% go test stack
ok stack 0.001s
6.5.3. Calculator
1. Create a reverse polish calculator. Use your stack package.
6.5.4. Answer
1. This is one answer:
<CODE BEGINS>
package main
import (
"bufio"
"fmt"
"os"
"strconv"
)
var reader *bufio.Reader = bufio.NewReader(os.Stdin)
var st = new(Stack)
type Stack struct {
i int
data [10]int
}
func (s *Stack) push(k int) {
if s.i+1 > 9 {
return
}
s.data[s.i] = k
s.i++
}
func (s *Stack) pop() (ret int) {
s.i--
if s.i < 0 {
s.i = 0
return
}
ret = s.data[s.i]
return
}
func main() {
for {
s, err := reader.ReadString('\n')
var token string
if err != nil {
return
}
for _, c := range s {
switch {
case c >= '0' && c <= '9':
token = token + string(c)
case c == ' ':
r, _ := strconv.Atoi(token)
st.push(r)
token = ""
case c == '+':
fmt.Printf("%d\n", st.pop()+st.pop())
case c == '*':
fmt.Printf("%d\n", st.pop()*st.pop())
case c == '-':
p := st.pop()
q := st.pop()
fmt.Printf("%d\n", q-p)
case c == 'q':
return
default:
//error
}
}
}
}
<CODE ENDS>
7. Beyond the basics
<blockquote>:
Go has pointers but not pointer arithmetic. You cannot use a pointer
variable to walk through the bytes of a string.
In this chapter we delve deeper in to the language.
Go has pointers. There is however no pointer arithmetic, so they act
more like references than pointers that you may know from C.
Pointers are useful. Remember that when you call a function in Go,
the variables are _pass-by-value_. So, for efficiency and the
possibility to modify a passed value _in_ functions we have pointers.
You declare a pointer by prefixing the type with an '"*"': "var p
*int". Now "p" is a pointer to an integer value. All newly declared
variables are assigned their zero value and pointers are no
different. A newly declared pointer, or just a pointer that points
to nothing, has a nil-value "nil". To make a pointer point to
something you can use the address-of operator "&"), which we
demonstrate here:
<CODE BEGINS>
var p *int
fmt.Printf("%v", p) <1>
var i int <2>
p = &i <3>
fmt.Printf("%v", p) <4>
<CODE ENDS>
This _1_ Prints "nil". Declare _2_ an integer variable "i". Make
"p" point _3_ to "i", i.e. take the address of "i". And this _4_
will print something like "0x7ff96b81c000a". De-referencing a
pointer is done by prefixing the pointer variable with "*".
As said, there is no pointer arithmetic, so if you write: "*p++", it
is interpreted as "(*p)++": first reference and then increment the
value.
7.1. Allocation
Go also has garbage collection, meaning that you don't have to worry
about memory deallocation.
To allocate memory Go has two primitives, "new" and "make". They do
different things and apply to different types, which can be
confusing, but the rules are simple. The following sections show how
to handle allocation in Go and hopefully clarifies the somewhat
artificial distinction between "new" "make"
7.1.1. Allocation with new
The built-in function "new" is essentially the same as its namesakes
in other languages: "new(T)" allocates zeroed storage for a new item
of type "T" and returns its address, a value of type "*T". Or in
other words, it returns a pointer to a newly allocated zero value of
type "T". This is important to remember.
The documentation for "bytes.Buffer" states that "the zero value for
Buffer is an empty buffer ready to use.". Similarly, "sync.Mutex"
does not have an explicit constructor or Init method. Instead, the
zero value for a "sync.Mutex" is defined to be an unlocked mutex.
7.1.2. Allocation with make
The built-in function "make(T, args)" serves a purpose different from
"new(T)". It creates slices, maps, and channels _only_, and it
returns an initialized (not zero!) value of type "T", and not a
pointer: "*T". The reason for the distinction is that these three
types are, under the covers, references to data structures that must
be initialized before use. A slice, for example, is a three-item
descriptor containing a pointer to the data (inside an array), the
length, and the capacity; until those items are initialized, the
slice is "nil". For slices, maps, and channels, "make" initializes
the internal data structure and prepares the value for use.
For instance, "make([]int, 10, 100)" allocates an array of 100 ints
and then creates a slice structure with length 10 and a capacity of
100 pointing at the first 10 elements of the array. In contrast,
"new([]int)" returns a pointer to a newly allocated, zeroed slice
structure, that is, a pointer to a "nil" slice value. These examples
illustrate the difference between "new" and "make".
<CODE BEGINS>
var p *[]int = new([]int) <1>
var v []int = make([]int, 100) <2>
var p *[]int = new([]int) <3>
*p = make([]int, 100, 100)
v := make([]int, 100) <4>
<CODE ENDS>
Allocates _1_ slice structure; rarely useful. "v" _2_ refers to a new
array of 100 ints. At _3_ we make it unnecessarily complex, _4_ is
more idiomatic.
Remember that "make" applies only to maps, slices and channels and
does not return a pointer. To obtain an explicit pointer allocate
with "new".
<aside>:
*new* allocates; *make* initializes.
The above two paragraphs can be summarized as:
* "new(T)" returns "*T" pointing to a zeroed "T"
* "make(T)" returns an initialized "T"
And of course "make" is only used for slices, maps and channels.
7.1.3. Constructors and composite literals
Sometimes the zero value isn't good enough and an initializing
constructor is necessary, as in this example taken from the package
"os".
<CODE BEGINS>
func NewFile(fd int, name string) *File {
if fd < 0 {
return nil
}
f := new(File)
f.fd = fd
f.name = name
f.dirinfo = nil
f.nepipe = 0
return f
}
<CODE ENDS>
There's a lot of boiler plate in there. We can simplify it using a
_composite literal_
<CODE BEGINS>
func NewFile(fd int, name string) *File {
if fd < 0 {
return nil
}
f := File{fd, name, nil, 0}
return &f <1>
}
<CODE ENDS>
It is OK to return the address of a local variable _1_ the storage
associated with the variable survives after the function returns.
In fact, taking the address of a composite literal allocates a fresh
instance each time it is evaluated, so we can combine these last two
lines.
<CODE BEGINS>
return &File{fd, name, nil, 0}
<CODE ENDS>
The items (called fields) of a composite literal are laid out in
order and must all be present. However, by labeling the elements
explicitly as field:value pairs, the initializers can appear in any
order, with the missing ones left as their respective zero values.
Thus we could say
<CODE BEGINS>
return &File{fd: fd, name: name}
<CODE ENDS>
As a limiting case, if a composite literal contains no fields at all,
it creates a zero value for the type. The expressions "new(File)"
and "&File{}" are equivalent. In fact the use of "new" is
discouraged.
Composite literals can also be created for arrays, slices, and maps,
with the field labels being indices or map keys as appropriate. In
these examples, the initializations work regardless of the values of
"Enone", and "Einval", as long as they are distinct:
<CODE BEGINS>
ar := [...]string{Enone: "no error", Einval: "invalid argument"}
sl := []string{Enone: "no error", Einval: "invalid argument"}
ma := map[int]string {Enone: "no error", Einval: "invalid argument"}
<CODE ENDS>
7.2. Defining your own types
Of course Go allows you to define new types, it does this with the
"type""type foo int"
This creates a new type "foo" which acts like an "int". Creating
more sophisticated types is done with the "struct" "string") and age
("int") in a single structure and make it a new type:
<CODE BEGINS>
package main
import "fmt"
type NameAge struct {
name string // Both non exported fields.
age int
}
func main() {
a := new(NameAge)
a.name = "Pete"
a.age = 42
fmt.Printf("%v\n", a)
}
<CODE ENDS>
Apropos, the output of "fmt.Printf("%v\n", a)" is "&{Pete 42}"
That is nice! Go knows how to print your structure. If you only
want to print one, or a few, fields of the structure you'll need to
use ".<field name>". For example to only print the name:
<CODE BEGINS>
fmt.Printf("%s", a.name)
<CODE ENDS>
7.2.1. More on structure fields
As said each item in a structure is called a field"struct {}". Or
one with four fields:
<CODE BEGINS>
struct {
x, y int
A *[]int
F func()
}
<CODE ENDS>
If you omit the name for a field, you create an anonymous field
(((field, anonymous))), for instance:
<CODE BEGINS>
struct {
T1 // Field name is T1.
*T2 // Field name is T2.
P.T3 // Field name is T3.
x, y int // Field names are x and y.
}
<CODE ENDS>
Note that field names that start with a capital letter are exported,
i.e. can be set or read from other packages. Field names that start
with a lowercase are private to the current package. The same goes
for functions defined in packages, see Section 6 for the details.
7.2.2. Methods
If you create functions that work on your newly defined type, you can
take two routes:
1. Create a function that takes the type as an argument.
<CODE BEGINS>
func doSomething(n1 *NameAge, n2 int) { /* */ }
<CODE ENDS>
2. Create a function that works on the type (see _receiver_ in
Section 5):
<CODE BEGINS>
func (n1 *NameAge) doSomething(n2 int) { /* */ }
<CODE ENDS>
This is a method call, which can be used as:
<CODE BEGINS>
var n *NameAge
n.doSomething(2)
<CODE ENDS>
Whether to use a function or method is entirely up to the programmer,
but if you want to satisfy an interface (see the next chapter) you
must use methods. If no such requirement exists it is a matter of
taste whether to use functions or methods.
But keep the following in mind, this is quoted from [go_spec]:
<blockquote>:
If "x" is addressable and "&x"'s method set contains "m", "x.m()" is
shorthand for "(&x).m()".
In the above case this means that the following is _not_ an error:
<CODE BEGINS>
var n NameAge // Not a pointer
n.doSomething(2)
<CODE ENDS>
Here Go will search the method list for "n" of type "NameAge", come
up empty and will then _also_ search the method list for the type
"*NameAge" and will translate this call to "(&n).doSomething(2)".
There is a subtle but major difference between the following type
declarations. Also see the Section "Type Declarations" [go_spec].
Suppose we have:
<CODE BEGINS>
// A Mutex is a data type with two methods, Lock and Unlock.
type Mutex struct { /* Mutex fields */ }
func (m *Mutex) Lock() { /* Lock impl. */ }
func (m *Mutex) Unlock() { /* Unlock impl. */ }
<CODE ENDS>
We now create two types in two different manners:
* "type NewMutex Mutex".
* "type PrintableMutex struct{Mutex}".
"NewMutex" is equal to "Mutex", but it _does not_ have _any_ of the
methods of "Mutex". In other words its method set is empty. But
"PrintableMutex" _has_ _inherited_ "Mutex". The Go term for this is
_embedding_ [go_spec]:
<blockquote>:
The method set of "*PrintableMutex" contains the methods "Lock" and
"Unlock" bound to its anonymous field "Mutex".
7.3. Conversions
Sometimes you want to convert a type to another type. This is
possible in Go, but there are some rules. For starters, converting
from one value to another is done by operators (that look like
functions: "byte()") and not all conversions are allowed.
+---------+-----------+-----------+-----------+-----------+--------+------------+
+=========+===========+===========+===========+===========+========+============+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
+---------+-----------+-----------+-----------+-----------+--------+------------+
Table 4: Valid conversions, "float64" works the same as
"float32".
* From a "string" to a slice of bytes or runes.
<CODE BEGINS>
mystring := "hello this is string"
byteslice := []byte(mystring)
<CODE ENDS>
Converts to a "byte" slice, each "byte" contains the integer value
of the corresponding byte in the string. Note that as strings in
Go are encoded in UTF-8 some characters in the string may end up
in 1, 2, 3 or 4 bytes.
<CODE BEGINS>
runeslice := []rune(mystring)
<CODE ENDS>
Converts to an "rune" slice, each "rune" contains a Unicode code
point. Every character from the string corresponds to one rune.
* From a slice of bytes or runes to a "string".
<CODE BEGINS>
b := []byte{'h','e','l','l','o'} // Composite literal.
s := string(b)
i := []rune{257,1024,65}
r := string(i)
<CODE ENDS>
For numeric values the following conversions are defined:
* Convert to an integer with a specific (bit) length: "uint8(int)"
* From floating point to an integer value: "int(float32)". This
discards the fraction part from the floating point value.
* And the other way around: "float32(int)".
7.3.1. User defined types and conversions
How can you convert between the types you have defined yourself? We
create two types here "Foo" and "Bar", where "Bar" is an alias for
"Foo":
<CODE BEGINS>
type foo struct { int } // Anonymous struct field.
type bar foo // bar is an alias for foo.
<CODE ENDS>
Then we:
<CODE BEGINS>
var b bar = bar{1} // Declare `b` to be a `bar`.
var f foo = b // Assign `b` to `f`.
<CODE ENDS>
Which fails on the last line with: "cannot use b (type bar) as type
foo in assignment"
This can be fixed with a conversion: "var f foo = foo(b)"
Note that converting structures that are not identical in their
fields is more difficult. Also note that converting "b" to a plain
"int" also fails; an integer is not the same as a structure
containing an integer.
7.4. Exercises
7.4.1. Map function with interfaces
1. Use the answer from the earlier map exercise but now make it
generic using interfaces. Make it at least work for ints and
strings.
7.4.2. Answer
1.
<CODE BEGINS>
package main
import "fmt"
// Define the empty interface as a type.
type e interface{}
func mult2(f e) e {
switch f.(type) {
case int:
return f.(int) * 2
case string:
return f.(string) + f.(string) + f.(string) + f.(string)
}
return f
}
func Map(n []e, f func(e) e) []e {
m := make([]e, len(n))
for k, v := range n {
m[k] = f(v)
}
return m
}
func main() {
m := []e{1, 2, 3, 4}
s := []e{"a", "b", "c", "d"}
mf := Map(m, mult2)
sf := Map(s, mult2)
fmt.Printf("%v\n", mf)
fmt.Printf("%v\n", sf)
}
<CODE ENDS>
7.4.3. Pointers
1. Suppose we have defined the following structure:
<CODE BEGINS>
type Person struct {
name string
age int
}
<CODE ENDS>
What is the difference between the following two lines?
<CODE BEGINS>
var p1 Person
p2 := new(Person)
<CODE ENDS>
2. What is the difference between the following two allocations?
<CODE BEGINS>
func Set(t *T) {
x = t
}
<CODE ENDS>
and
<CODE BEGINS>
func Set(t T) {
x= &t
}
<CODE ENDS>
7.4.4. Answer
1. The expression, "var p1 Person" allocates a "Person"-_value_ to
"p1". The type of "p1" is "Person". The second line: "p2 :=
new(Person)" allocates memory and assigns a _pointer_ to "p2".
The type of "p2" is "*Person".
2. In the first function, "x" points to the same thing that "t"
does, which is the same thing that the actual argument points to.
So in the second function, we have an "extra" variable containing
a copy of the interesting value. In the second function, "x"
points to a new (heap-allocated) variable "t" which contains a
copy of whatever the actual argument value is.
7.4.5. Linked List
1. Make use of the package "container/list" to create a (doubly)
linked list. Push the values 1, 2 and 4 to the list and then
print it.
2. Create your own linked list implementation. And perform the same
actions as above.
7.4.6. Answer
1. The following is the implementation of a program using doubly
linked lists from "container/list".
<CODE BEGINS>
package main
import (
"container/list"
"fmt"
)
func main() {
l := list.New()
l.PushBack(1)
l.PushBack(2)
l.PushBack(4)
for e := l.Front(); e != nil; e = e.Next() {
fmt.Printf("%v\n", e.Value)
}
}
<CODE ENDS>
2. The following is a program implementing a simple doubly linked
list supporting "int" values.
<CODE BEGINS>
package main
import (
"errors" <1>
"fmt"
)
type Value int <2>
type Node struct { <3>
Value
prev, next *Node
}
type List struct {
head, tail *Node
}
func (l *List) Front() *Node { <4>
return l.head
}
func (n *Node) Next() *Node {
return n.next
}
func (l *List) Push(v Value) *List {
n := &Node{Value: v} <5>
if l.head == nil { <6>
l.head = n
} else {
l.tail.next = n <7>
n.prev = l.tail <8>
}
l.tail = n <9>
return l
}
var errEmpty = errors.New("List is empty")
func (l *List) Pop() (v Value, err error) {
if l.tail == nil { <10>
err = errEmpty
} else {
v = l.tail.Value <11>
l.tail = l.tail.prev <12>
if l.tail == nil {
l.head = nil <13>
}
}
return v, err
}
func main() {
l := new(List)
l.Push(1)
l.Push(2)
l.Push(4)
for n := l.Front(); n != nil; n = n.Next() {
fmt.Printf("%v\n", n.Value)
}
fmt.Println()
for v, err := l.Pop(); err == nil; v, err = l.Pop() {
fmt.Printf("%v\n", v)
}
}
<CODE ENDS>
Import <_1_> the packages we will need. At <_2_> we declare a type
for the value our list will contain, this is not strictly neccesary.
And at <_3_> we declare a type for the each node in our list. At
<_4_> we define the "Front" method for our list. When pushing,
create a new Node <_5_> with the provided value. If the list is
empty <_6_>, put the new node at the head. Otherwise <_7_> put it at
the tail and make sure <_8_> the new node points back to the
previously existing one. At <_9_> we re-adjust tail to the newly
inserted node.
In the Pop _10_ method, we return an error if the list is empty. If
it is not empty _11_ we save the last value. And then _12_ discard
the last node from the list. Finally at _13_ we make sure the list
is consistent if it becomes empty.
7.4.7. Cat
1. Write a program which mimics the Unix program "cat".
2. Make it support the "-n" flag, where each line is numbered.
3. The solution to the above question given in contains a bug. Can
you spot and fix it?
7.4.8. Answer
1. The following is implemention of "cat" which also supports a -n
flag to number each line.
<CODE BEGINS>
package main
import (
"bufio"
"flag"
"fmt"
"io" <1>
"os"
)
var numberFlag = flag.Bool("n", false, "number each line") // <<2>>
func cat(r *bufio.Reader) { <3>
i := 1
for {
buf, e := r.ReadBytes('\n') <4>
if e == io.EOF { <5>
break
}
if *numberFlag { <6>
fmt.Fprintf(os.Stdout, "%5d %s", i, buf)
i++
} else { <7>
fmt.Fprintf(os.Stdout, "%s", buf)
}
}
return
}
func main() {
flag.Parse()
if flag.NArg() == 0 {
cat(bufio.NewReader(os.Stdin))
}
for i := 0; i < flag.NArg(); i++ {
f, e := os.Open(flag.Arg(i))
if e != nil {
fmt.Fprintf(os.Stderr, "%s: error reading from %s: %s\n",
os.Args[0], flag.Arg(i), e.Error())
continue
}
cat(bufio.NewReader(f))
}
}
<CODE ENDS>
At _1_ we include all the packages we need. Here _2_ we define a new
flag "n", which defaults to off. Note that we get the help (-h) for
free. Start the function _3_ that actually reads the file's contents
and displays it; Read one line at the time at _4_. And stop _5_ if we
hit the end. If we should number each line, print the line number
and then the line itself _6_. Otherwise _7_ we could just print the
line.
2. The bug show itself when the last line of the input does not
contain a newline. Or worse, when the input contains one line
without a closing newline nothing is shown at all. A better
solution is the following program.
<CODE BEGINS>
package main
import (
"bufio"
"flag"
"fmt"
"io"
"os"
)
var numberFlag = flag.Bool("n", false, "number each line")
func cat(r *bufio.Reader) {
i := 1
for {
buf, e := r.ReadBytes('\n')
if e == io.EOF && string(buf) == "" {
break
}
if *numberFlag {
fmt.Fprintf(os.Stdout, "%5d %s", i, buf)
i++
} else {
fmt.Fprintf(os.Stdout, "%s", buf)
}
}
return
}
func main() {
flag.Parse()
if flag.NArg() == 0 {
cat(bufio.NewReader(os.Stdin))
}
for i := 0; i < flag.NArg(); i++ {
f, e := os.Open(flag.Arg(i))
if e != nil {
fmt.Fprintf(os.Stderr, "%s: error reading from %s: %s\n",
os.Args[0], flag.Arg(i), e.Error())
continue
}
cat(bufio.NewReader(f))
}
}
<CODE ENDS>
7.4.9. Method calls
1. Suppose we have the following program. Note the package
"container/vector" was once part of Go, but was removed when the
"append" built-in was introduced. However, for this question
this isn't important. The package implemented a stack-like
structure, with push and pop methods.
<CODE BEGINS>
package main
import "container/vector"
func main() {
k1 := vector.IntVector{}
k2 := &vector.IntVector{}
k3 := new(vector.IntVector)
k1.Push(2)
k2.Push(3)
k3.Push(4)
}
<CODE ENDS>
What are the types of "k1", "k2" and "k3"?
2. Now, this program compiles and runs OK. All the "Push"
operations work even though the variables are of a different
type. The documentation for "Push" says:
<blockquote>: "func (p *IntVector) Push(x int)" Push appends x to
the end of the vector.
So the receiver has to be of type "*IntVector", why does the code
above (the Push statements) work correctly then?
7.4.10. Answer
1. The type of "k1" is "vector.IntVector". Why? We use a composite
literal (the "{}"), so we get a value of that type back. The
variable "k2" is of "*vector.IntVector", because we take the
address ("&") of the composite literal. And finally "k3" has
also the type "*vector.IntVector", because "new" returns a
pointer to the type.
2. The answer is given in [go_spec] in the section "Calls", where
among other things it says:
<blockquote>:
A method call "x.m()" is valid if the method set of (the type of) "x"
contains "m" and the argument list can be assigned to the parameter
list of "m". If "x" is addressable and "&x"'s method set contains
"m", "x.m()" is shorthand for "(&x).m()".
In other words because "k1" is addressable and "*vector.IntVector"
_does_ have the "Push" method, the call "k1.Push(2)" is translated by
Go into "(&k1).Push(2)" which makes the type system happy again (and
you too -- now you know this).
8. Interfaces
<blockquote>:
I have this phobia about having my body penetrated surgically. You
know what I mean?
In Go, the word _interface__set of methods defined_ for "S" with one
field, and defines two methods for "S".
<CODE BEGINS>
type S struct { i int }
func (p *S) Get() int { return p.i }
func (p *S) Put(v int) { p.i = v }
<CODE ENDS>
Figure 1: Defining a struct and methods on it.
You can also define an "I" with two methods:
<CODE BEGINS>
type I interface {
Get() int
Put(int)
}
<CODE ENDS>
"S" is a valid _implementation_ for interface "I", because it defines
the two methods which "I" requires. Note that this is true even
though there is no explicit declaration that "S" implements "I".
A Go program can use this fact via yet another meaning of interface,
which is an interface value:
<CODE BEGINS>
func f(p I) { <1>
fmt.Println(p.Get()) <2>
p.Put(1) <3>
}
<CODE ENDS>
At _1_ we declare a function that takes an interface type as the
argument. Because "p" implements "I", it _must_ have the "Get()"
method, which we call at _2_. And the same holds true for the "Put()"
method at _3_. Because "S" implements "I", we can call the function
"f" passing in a pointer to a value of type "S": "var s S; f(&s)"
The reason we need to take the address of "s", rather than a value of
type "S", is because we defined the methods on "s" to operate on
pointers, see the definition in the code above. This is not a
requirement -- we could have defined the methods to take values --
but then the "Put" method would not work as expected.
The fact that you do not need to declare whether or not a type
implements an interface means that Go implements a form of duck
typing [duck_typing]. This is not pure duck typing, because when
possible the Go compiler will statically check whether the type
implements the interface. However, Go does have a purely dynamic
aspect, in that you can convert from one interface type to another.
In the general case, that conversion is checked at run time. If the
conversion is invalid -- if the type of the value stored in the
existing interface value does not satisfy the interface to which it
is being converted -- the program will fail with a run time error.
Interfaces in Go are similar to ideas in several other programming
languages: pure abstract virtual base classes in C++, typeclasses in
Haskell or duck typing in Python. However there is no other language
which combines interface values, static type checking, dynamic run
time conversion, and no requirement for explicitly declaring that a
type satisfies an interface. The result in Go is powerful, flexible,
efficient, and easy to write.
8.1. Which is what?
Let's define another type "R" that also implements the interface "I":
<CODE BEGINS>
type R struct { i int }
func (p *R) Get() int { return p.i }
func (p *R) Put(v int) { p.i = v }
<CODE ENDS>
The function "f" can now accept variables of type "R" and "S".
Suppose you need to know the actual type in the function "f". In Go
you can figure that out by using a type switch
<CODE BEGINS>
func f(p I) {
switch t := p.(type) { <1>
case *S: <2>
case *R: <2>
default: <3>
}
}
<CODE ENDS>
At _1_ we use the type switch, note that the ".(type)" syntax is
_only_ valid within a "switch" statement. We store the value in the
variable "t". The subsequent cases _2_ each check for a different
_actual_ type. And we can even have a "default" _3_ clause. It is
worth pointing out that both "case R" and "case s" aren't possible,
because "p" needs to be a pointer in order to satisfy "i".
A type switch isn't the only way to discover the type at _run-time_.
<CODE BEGINS>
if t, ok := something.(I); ok { <1>
// ...
}
<CODE ENDS>
You can also use a "comma, ok" form _1_ to see if an interface type
implements a specific interface. If "ok" is true, "t" will hold the
type of "something". When you are sure a variable implements an
interface you can use: "t := something.(I)" .
8.2. Empty interface
Since every type satisfies the empty interface: "interface{}" we can
create a generic function which has an empty interface as its
argument:
<CODE BEGINS>
func g(something interface{}) int {
return something.(I).Get()
}
<CODE ENDS>
The "return something.(I).Get()" is the tricky bit in this function.
The value "something" has type "interface{}", meaning no guarantee of
any methods at all: it could contain any type. The ".(I)" is a type
assertion "something" to an interface of type "I". If we have that
type we can invoke the "Get()" function. So if we create a new
variable of the type "*S", we can just call "g()", because "*S" also
implements the empty interface.
<CODE BEGINS>
s = new(S)
fmt.Println(g(s));
<CODE ENDS>
The call to "g" will work fine and will print 0. If we however
invoke "g()" with a value that does not implement "I" we have a
problem:
<CODE BEGINS>
var i int
fmt.Println(g(i))
<CODE ENDS>
This compiles, but when we run this we get slammed with: "panic:
interface conversion: int is not main.I: missing method Get".
Which is completely true, the built-in type "int" does not have a
"Get()" method.
8.3. Methods
Methods are functions that have a receiver (see Section 5). You can
define methods on any type (except on non-local types, this includes
built-in types: the type "int" can not have methods). You can
however make a new integer type with its own methods. For example:
<CODE BEGINS>
type Foo int
func (self Foo) Emit() {
fmt.Printf("%v", self)
}
type Emitter interface {
Emit()
}
<CODE ENDS>
Doing this on non-local (types defined in other packages) types
yields an error "cannot define new methods on non-local type int".
8.4. Methods on interface types
An interface defines a set of methods. A method contains the actual
code. In other words, an interface is the definition and the methods
are the implementation. So a receiver can not be an interface type,
doing so results in a "invalid receiver type ..." compiler error.
The authoritative word from the language spec [go_spec]:
<blockquote>:
The receiver type must be of the form "T" or "*T" where "T" is a type
name. "T" is called the receiver base type or just base type. The
base type must not be a pointer or interface type and must be
declared in the same package as the method.
<aside>:
Creating a pointer to an interface value is a useless action in Go.
It is in fact illegal to create a pointer to an interface value. The
release notes for an earlier Go release that made them illegal leave
no room for doubt:
<blockquote>:
The language change is that uses of pointers to interface values no
longer automatically de-reference the pointer. A pointer to an
interface value is more often a beginner's bug than correct code.
8.5. Interface names
By convention, one-method interfaces are named by the method name
plus the _-er_ suffix: Read_er_, Writ_er_, Formatt_er_ etc.
There are a number of such names and it's productive to honor them
and the function names they capture. "Read", "Write", "Close",
"Flush", "String" and so on have canonical signatures and meanings.
To avoid confusion, don't give your method one of those names unless
it has the same signature and meaning. Conversely, if your type
implements a method with the same meaning as a method on a well-known
type, give it the same name and signature; call your string-converter
method "String" not "ToString".
8.6. A sorting example
Recall the Bubblesort exercise, where we sorted an array of integers:
<CODE BEGINS>
func bubblesort(n []int) {
for i := 0; i < len(n)-1; i++ {
for j := i + 1; j < len(n); j++ {
if n[j] < n[i] {
n[i], n[j] = n[j], n[i]
}
}
}
}
<CODE ENDS>
A version that sorts strings is identical except for the signature of
the function: "func bubblesortString(n []string) { /* ... */ }" .
Using this approach would lead to two functions, one for each type.
By using interfaces we can make this more
<CODE BEGINS>
func sort(i []interface{}) { <1>
switch i.(type) { <2>
case string: <3>
// ...
case int:
// ...
}
return /* ... */ <4>
}
<CODE ENDS>
Our function will receive a slice of empty interfaces at _1_. We then
_2_ use a type switch to find out what the actual type of the input
is. And then _3_ then sort accordingly. And, when done, return _4_
the sorted slice.
But when we call this function with "sort([]int{1, 4, 5})", it fails
with: "cannot use i (type []int) as type []interface { } in function
argument"
This is because Go can not easily convert to a _slice_ of interfaces.
Just converting to an interface is easy, but to a slice is much more
costly. The full mailing list discussion on this subject can be
found at [go_nuts_interfaces]. To keep a long story short: Go does
not (implicitly) convert slices for you.
So what is the Go way of creating such a "generic" function? Instead
of doing the type inference ourselves with a type switch, we let Go
do it implicitly: The following steps are required:
* Define an interface type (called "Sorter" here) with a number of
methods needed for sorting. We will at least need a function to
get the length of the slice, a function to compare two values and
a swap function.
<CODE BEGINS>
type Sorter interface {
Len() int // len() as a method.
Less(i, j int) bool // p[j] < p[i] as a method.
Swap(i, j int) // p[i], p[j] = p[j], p[i] as a method.
}
<CODE ENDS>
* Define new types for the slices we want to sort. Note that we
declare slice types:
<CODE BEGINS>
type Xi []int
type Xs []string
<CODE ENDS>
* Implementation of the methods of the "Sorter" interface. For
integers:
<CODE BEGINS>
func (p Xi) Len() int {return len(p)}
func (p Xi) Less(i int, j int) bool {return p[j] < p[i]}
func (p Xi) Swap(i int, j int) {p[i], p[j] = p[j], p[i]}
<CODE ENDS>
And for strings:
<CODE BEGINS>
func (p Xs) Len() int {return len(p)}
func (p Xs) Less(i int, j int) bool {return p[j] < p[i]}
func (p Xs) Swap(i int, j int) {p[i], p[j] = p[j], p[i]}
<CODE ENDS>
* Write a _generic_ Sort function that works on the "Sorter"
interface.
<CODE BEGINS>
func Sort(x Sorter) { <1>
for i := 0; i < x.Len() - 1; i++ { <2>
for j := i + 1; j < x.Len(); j++ {
if x.Less(i, j) {
x.Swap(i, j)
}
}
}
}
<CODE ENDS>
At _1_ "x" is now of the "Sorter" type and using the defined
methods for this interface we implement Bubblesort at _2_.
Now we can use our _generic_ "Sort" function as follows:
<CODE BEGINS>
ints := Xi{44, 67, 3, 17, 89, 10, 73, 9, 14, 8}
strings := Xs{"nut", "ape", "elephant", "zoo", "go"}
Sort(ints)
fmt.Printf("%v\n", ints)
Sort(strings)
fmt.Printf("%v\n", strings)
<CODE ENDS>
8.7. Listing interfaces in interfaces
Take a look at the following example of an interface definition, this
one is from the package "container/heap":
<CODE BEGINS>
type Interface interface {
sort.Interface
Push(x interface{})
Pop() interface{}
}
<CODE ENDS>
Here another interface is listed inside the definition of
"heap.Interface", this may look odd, but is perfectly valid, remember
that on the surface an interface is nothing more than a listing of
methods. "sort.Interface" is also such a listing, so it is perfectly
legal to include it in the interface.
8.8. Introspection and reflection
In the following example we want to look at the "tag" (here named
"namestr") defined in the type definition of "Person". To do this we
need the "reflect"_type_ definition. So we use the "reflect" package
to figure out the type of the variable and _then_ access the tag.
<CODE BEGINS>
type Person struct {
name string "namestr"
age int
}
func ShowTag(i interface{}) { <1>
switch t := reflect.TypeOf(i); t.Kind() {
case reflect.Ptr: <2>
tag := t.Elem().Field(0).Tag
// <<3>> <<4>> <<5>>
<CODE ENDS>
Figure 2: Introspection using reflection.
We are calling "ShowTag" at _1_ with a "*Person", so at _2_ we're
expecting a "reflect.Ptr". We are dealing with a "Type" _3_ and
according to the documentation :
<blockquote>:
Elem returns a type's element type. It panics if the type's Kind is
not Array, Chan, Map, Ptr, or Slice.
So on "t" we use "Elem()" to get the value the pointer points to. We
have now dereferenced the pointer and are "inside" our structure. We
then _4_ use "Field(0)" to access the zeroth field.
The struct "StructField" has a "Tag" member which returns the tag-
name as a string. So on the "0^{th}" field we can unleash ".Tag" _5_
to access this name: "Field(0).Tag". This gives us "namestr".
To make the difference between types and values more clear, take a
look at the following code:
<CODE BEGINS>
func show(i interface{}) {
switch t := i.(type) {
case *Person:
t := reflect.TypeOf(i) <1>
v := reflect.ValueOf(i) <2>
tag := t.Elem().Field(0).Tag <3>
name := v.Elem().Field(0).String() <4>
}
}
<CODE ENDS>
Figure 3: Reflection and the type and value.
At _1_ we create "t" the type data of "i", and "v" gets the actual
values at _2_. Here at _3_ we want to get to the "tag". So we need
"Elem()" to redirect the pointer, access the first field and get the
tag. Note that we operate on "t" a "reflect.Type". Now _4_ we want
to get access to the _value_ of one of the members and we employ
"Elem()" on "v" to do the redirection. we have "arrived" at the
structure. Then we go to the first field "Field(0)" and invoke the
"String()" method on it.
Peeling away the layers using reflection.
Setting a value works similarly as getting a value, but only works on
_exported_ members. Again some code:
<CODE BEGINS>
type Person struct {
name string
age int
}
func Set(i interface{}) {
switch i.(type) {
case *Person:
r := reflect.ValueOf(i)
r.Elem(0).Field(0).SetString("Albert Einstein")
}
}
<CODE ENDS>
Figure 4: Reflect with _private_ member.
<CODE BEGINS>
type Person struct {
Name string
age int
}
func Set(i interface{}) {
switch i.(type) {
case *Person:
r := reflect.ValueOf(i)
r.Elem().Field(0).SetString("Albert Einstein")
}
}
<CODE ENDS>
Figure 5: Reflect with _public_ member.
The first program compiles and runs, but when you run it, you are
greeted with a stack trace and a _run time_ error: "panic:
reflect.Value.SetString using value obtained using unexported field".
The second program works OK and sets the member "Name" to "Albert
Einstein". Of course this only works when you call "Set()" with a
pointer argument.
8.9. Exercises
### Interfaces and max()
In the maximum exercise we created a max function that works on a
slice of integers. The question now is to create a program that
shows the maximum number and that works for both integers and floats.
Try to make your program as generic as possible, although that is
quite difficult in this case.
8.9.1. Answer
The following program calculates a maximum. It is as generic as you
can get with Go.
<CODE BEGINS>
package main
import "fmt"
func Less(l, r interface{}) bool { <1>
switch l.(type) {
case int:
if _, ok := r.(int); ok {