Skip to content

G
go
Commit 718de6f4 authored by Andrew Gerrand's avatar Andrew Gerrand
Browse files
Options

doc: expand code.html to discuss the go tool in more depth 

R=golang-dev, balasanjay, r
CC=golang-dev
https://golang.org/cl/5640045
parent c3fbc9a5
Show whitespace changes
Inline Side-by-side
Showing with 354 additions and 195 deletions
doc/code.html
View file @ 718de6f4
......@@ -5,340 +5,499 @@
<h2 id="Introduction">Introduction</h2>
<p>
This document explains how to write a new package
and how to test code.
It assumes you have installed Go using the
<a href="install.html">installation instructions</a>.
This document demonstrates the development of a simple Go package and
introduces the <a href="/cmd/go/">go command</a>, the standard way to fetch,
build, and install Go packages and commands.
</p>
<h2 id="GOPATH">Code organization</h2>
<h3><code>GOPATH</code> and workspaces</h3>
<p>
One of Go's design goals is to make writing software easier. To that end, the
<code>go</code> command doesn't use Makefiles or other configuration files to
guide program construction. Instead, it uses the source code to find
dependencies and determine build conditions. This means your source code and
build scripts are always in sync; they are one and the same.
</p>
<p>
The one thing you must do is set a <code>GOPATH</code> environment variable.
<code>GOPATH</code> tells the <code>go</code> command (and other related tools)
where to find and install the Go packages on your system.
</p>
<p>
<code>GOPATH</code> is a list of paths. It shares the syntax of your system's
<code>PATH</code> environment variable. A typical <code>GOPATH</code> on
a Unix system might look like this:
</p>
<pre>
GOPATH=/home/user/ext:/home/user/mygo
</pre>
<p>
Before embarking on a change to an existing
package or the creation of a new package,
be sure to send mail to the
<a href="http://groups.google.com/group/golang-nuts">mailing list</a>
to let people know what you are thinking of doing.
Doing so helps avoid duplication of effort and
enables discussions about design before any code
has been written.
(On a Windows system use semicolons as the path separator instead of colons.)
</p>
<h2 id="Community_resources">Community resources</h2>
<p>
Each path in the list (in this case <code>/home/user/ext</code> or
<code>/home/user/mygo</code>) specifies the location of a <i>workspace</i>.
A workspace contains Go source files and their associated package objects, and
command executables. It has a prescribed structure of three subdirectories:
</p>
<ul>
<li><code>src</code> contains Go source files,
<li><code>pkg</code> contains compiled package objects, and
<li><code>bin</code> contains executable commands.</code>
</ul>
<p>
Subdirectories of the <code>src</code> directory hold independent packages, and
all source files (<code>.go</code>, <code>.c</code>, <code>.h</code>, and
<code>.s</code>) in each subdirectory are elements of that subdirectory's
package.
</p>
<p>
For real-time help, there may be users or developers on
<code>#go-nuts</code> on the <a href="http://freenode.net/">Freenode</a> IRC server.
When building a program that imports the package "<code>widget</code>" the
<code>go</code> command looks for <code>src/pkg/widget</code> inside the Go root,
and then&mdash;if the package source isn't found there&mdash;it searches
for <code>src/widget</code> inside each workspace in order.
</p>
<p>
The official mailing list for discussion of the Go language is
<a href="http://groups.google.com/group/golang-nuts">Go Nuts</a>.
Multiple workspaces can offer some flexibility and convenience, but for now
we'll concern ourselves with only a single workspace.
</p>
<p>
Bugs can be reported using the <a href="http://code.google.com/p/go/issues/list">Go issue tracker</a>.
Let's work through a simple example. First, create a <code>$HOME/mygo</code>
directory and its <code>src</code> subdirectory:
</p>
<pre>
$ mkdir -p $HOME/mygo/src # create a place to put source code
</pre>
<p>
For those who wish to keep up with development,
there is another mailing list, <a href="http://groups.google.com/group/golang-checkins">golang-checkins</a>,
that receives a message summarizing each checkin to the Go repository.
Next, set it as the <code>GOPATH</code>. You should also add the
<code>bin</code> subdirectory to your <code>PATH</code> environment variable so
that you can run the commands therein without specifying their full path.
To do this, add the following lines to <code>$HOME/.profile</code> (or
equivalent):
</p>
<pre>
export GOPATH=$HOME/mygo
export PATH=$PATH:$HOME/mygo/bin
</pre>
<h2 id="New_package">Creating a new package</h2>
<h3>Choosing an import path</h3>
<h3>Import paths</h3>
<p>
The standard packages are given short names like <code>fmt</code> and
<code>net/http</code> for convenience.
For your own projects, choose a name space that is unlikely
to collide with future additions to the standard library or other
The standard packages are given short import paths such as <code>"fmt"</code>
and <code>"net/http"</code> for convenience.
For your own projects, it is important to choose a base import path that is
unlikely to collide with future additions to the standard library or other
external libraries.
</p>
<p>
The best way to choose an import path is to use the location of your version
control repository.
For instance, if your source repository is at <code>example.com</code>
or <code>code.google.com/p/example</code>, you should begin your package
paths with that URL, as in "<code>example.com/foo/bar</code>" or
"<code>code.google.com/p/example/foo/bar</code>".
This way the <a href="/cmd/go/"><code>go</code> tool</a> can automatically
check out and build the source code from its import path.
Using this convention, the <code>go</code> command can automatically check out and
build the source code by its import path alone.
</p>
<p>
If you don't intend your code to be installed in this way, you should at
If you don't intend to install your code in this way, you should at
least use a unique prefix like "<code>widgets/</code>", as in
"<code>widgets/foo/bar</code>". A good rule is to use a prefix such as your
company or project name since it is unlikely to be used by another group.
company or project name, since it is unlikely to be used by another group.
</p>
<p>
We'll use <code>example/</code> as our base import path:
</p>
<pre>
$ mkdir -p $GOPATH/src/example
</pre>
<h3>The <code>go</code> tool and <code>GOPATH</code></h3>
<h3>Package names</h3>
<p>
The <a href="/cmd/go/"><code>go</code> tool</a> is the standard means of
building and installing Go libraries and programs. It is a "zero configuration"
tool; it determines how to build Go packages from their source code alone.
The first statement in a Go source file should be
</p>
<pre>
package <i>name</i>
</pre>
<p>
To use the <code>go</code> tool effectively you must set the
<code>GOPATH</code> variable.
<code>GOPATH</code> specifies a list of paths that contain Go source code
and package binaries. Source code, package objects, and command binaries are
located inside the <code>GOPATH</code>s' <code>src</code>, <code>pkg</code>,
and <code>bin</code> subdirectories respectively.
where <code><i>name</i></code> is the package's default name for imports.
(All files in a package must use the same <code><i>name</i></code>.)
</p>
<p>
You should set <code>GOPATH</code> in your shell profile
(<code>$HOME/.bashrc</code>, <code>$HOME/.profile</code>, or equivalent).
Go's convention is that the package name is the last element of the
import path: the package imported as "<code>crypto/rot13</code>"
should be named <code>rot13</code>.
There is no requirement that package names be unique
across all packages linked into a single binary,
only that the import paths (their full file names) be unique.
</p>
<p>
Create a new package under <code>example</code> called <code>newmath</code>:
</p>
<pre>
$ cd $GOPATH/src/example
$ mkdir newmath
</pre>
<p>
This shell session demonstrates setting <code>GOPATH</code>, creating a trivial
<code>widgets/foo</code> package, and building and installing the package.
Then create a file named <code>$GOPATH/src/example/newmath/sqrt.go</code>
containing the following Go code:
</p>
<pre>
$ export GOPATH=$HOME/gocode
$ mkdir -p $GOPATH/src/widgets/foo
$ cat &gt; $GOPATH/src/widgets/foo/foo.go
package foo
const String = "Go rules!"
^D
$ go install widgets/foo
$ ls $GOPATH/pkg/*/widgets
foo.a
// Package newmath is a trivial example package.
package newmath
// Sqrt returns an approximation to the square root of x.
func Sqrt(x float64) float64 {
// This is a terrible implementation.
// Real code should import "math" and use math.Sqrt.
z := 0.0
for i := 0; i < 1000; i++ {
z -= (z*z - x) / (2 * x)
}
return z
}
</pre>
<p>(<code>^D</code> means to type Control-D.)</p>
<p>
This package is imported by the path name of the directory it's in, starting
after the <code>src</code> component:
</p>
<pre>
import "example/newmath"
</pre>
<p>
Type <code>go help gopath</code> on the command line for more information
about <code>GOPATH</code>.
See <a href="/doc/effective_go.html#names">Effective Go</a> to learn more about
Go's naming conventions.
</p>
<h3>Go source files</h3>
<h2>Building and installing</h3>
<p>
The first statement in a Go source file should be <code>package
<i>name</i></code>, where <code><i>name</i></code> is the package's default
name for imports.
(All files in a package must use the same <code><i>name</i></code>.)
Go's convention is that the package name is the last element of the
import path: the package imported as "<code>crypto/rot13</code>"
should be named <code>rot13</code>.
There is no requirement that package names be unique
across all packages linked into a single binary,
only that the import paths (their full file names) be unique.
The <code>go</code> command comprises several subcommands, the most central being
<code>install</code>. Running <code>go install <i>importpath</i></code> builds
and installs a package and its dependencies.
</p>
<p>
Go compiles all the source files in a package at once, so one file
can refer to constants, variables, types, and functions in another
file without special arrangement or declarations.
To "install a package" means to write the package object or executable command
to the <code>pkg</code> or <code>bin</code> subdirectory of the workspace in
which the source resides.
</p>
<h3>Building a package</h3>
<p>
Writing clean, idiomatic Go code is beyond the scope of this document.
<a href="effective_go.html">Effective Go</a> is an introduction to
that topic.
To build and install the <code>newmath</code> package, type
</p>
<h2 id="Building_programs">Building programs</h2>
<pre>
$ go install example/newmath
</pre>
<p>
The <a href="/cmd/go/"><code>go</code> tool</a> treats code belonging to
<code>package main</code> as an executable command, and installs the package
binary to the <code>GOPATH</code>'s <code>bin</code> subdirectory.
This command will produce no output if the package and its dependencies
are built and installed correctly.
</p>
<p>
Building executable commands is the same as building packages.
Use "<code>go install</code>":
As a convenience, the <code>go</code> command will assume the current directory
if no import path is specified on the command line. This sequence of commands
has the same affect as the one above:
</p>
<pre>
$ mkdir -p $GOPATH/src/widgets/bar
$ cat &gt; $GOPATH/src/widgets/bar/bar.go
$ cd $GOPATH/src/example/newmath
$ go install
</pre>
<p>
The resulting workspace directory tree (assuimg we're running Linux on a 64-bit
system) looks like this:
</p>
<pre>
pkg/
linux_amd64/
example/
newmath.a # package object
src/
example/
newmath/
sqrt.go # package source
</pre>
<h3>Building a command</h3>
<p>
The <code>go</code> command treats code belonging to <code>package main</code> as
an executable command and installs the package binary to the
<code>GOPATH</code>'s <code>bin</code> subdirectory.
</p>
<p>
Add a command named <code>hello</code> to the source tree.
First create the <code>example/hello</code> directory:
</p>
<pre>
$ cd $GOPATH/src/example
$ mkdir hello
</pre>
<p>
Then create the file <code>$GOPATH/src/example/hello/hello.go</code>
containing the following Go code.
</p>
<pre>
// Hello is a trivial example of a main package.
package main
import (
"example/newmath"
"fmt"
"widgets/foo"
)
func main() {
fmt.Println(foo.String)
fmt.Printf("Hello, world. Sqrt(2) = %v\n", newmath.Sqrt(2))
}
^D
$ go install widgets/bar
$ $GOPATH/bin/bar
Go rules!
</pre>
<p>
Run <code>go help build</code> and <code>go help install</code> for more
about building and installing Go binaries.
Next, run <code>go install</code>, which builds and installs the binary to
<code>$GOPATH/bin</code>:
</p>
<pre>
$ go install example/hello
</pre>
<p>
To run the program, invoke it by name as you would any other command:
</p>
<pre>
$ $GOPATH/bin/hello
Hello, world. Sqrt(2) = 1.414213562373095
</pre>
<p>
If you added <code>$HOME/mygo/bin</code> to your <code>PATH</code>, you may omit
the path to the executable:
</p>
<pre>
$ hello
Hello, world. Sqrt(2) = 1.414213562373095
</pre>
<p>
The workspace directory tree now looks like this:
</p>
<pre>
bin/
hello # command executable
pkg/
linux_amd64/
example/
newmath.a # package object
src/
example/
hello/
hello.go # command source
newmath/
sqrt.go # package source
</pre>
<p>
The <code>go</code> command also provides a <code>build</code> command, which is
like <code>install</code> except it builds all objects in a temporary directory
and does not install them under <code>pkg</code> or <code>bin</code>.
When building a command an executable named after the last element of the
import path is written to the current directory. When building a package,
<code>go build</code> serves merely to test that the package and its
dependencies can be built. (The resulting package object is thrown away.)
</p>
<h2 id="Testing">Testing</h2>
<p>
Go has a lightweight test framework composed of the <code>go</code> tool and
the <code>testing</code> package.
Go has a lightweight test framework composed of the <code>go test</code>
command and the <code>testing</code> package.
</p>
<p>
You write a test by creating a file with a name ending in <code>_test.go</code>
that contains functions named <code>TestXXX</code> with signature
<code>func (t *testing.T)</code>.
The test framework runs each such function;
if the function calls a failure function such as <code>t.Error</code> or
<code>t.Fail</code>, the test is considered to have failed.
Run <code>go help test</code> and see the
<a href="/pkg/testing/">testing package documentation</a> for more detail.
</p>
<p>
To run the test, run "<code>go test</code>":
Add a test to the <code>newmath</code> package by creating the file
<code>$GOPATH/src/example/newmath/sqrt_test.go</code> containing the following
Go code.
</p>
<pre>
$ cat &gt; $GOPATH/src/widgets/foo/foo_test.go
package foo
package newmath
import "testing"
func TestString(t *testing.T) {
const expect = "Go rules!"
if String != expect {
t.Errorf("String == %q, want %q", String, expect)
func TestSqrt(t *testing.T) {
const in, out = 9, 3
if x := Sqrt(in); x != out {
t.Errorf("Sqrt(%v) = %v, want %v", in, x, out)
}
}
^D
$ go test widgets/foo
ok widgets/foo 0.018s
</pre>
<p>
If your change affects performance, add a <code>Benchmark</code> function
(run <code>go help testfunc</code>) and run it using <code>go test
-test.bench=.*</code>.
Now run the test with <code>go test</code>:
</p>
<h2 id="pkg_example">An example package with tests</h2>
<pre>
$ go test example/newmath
ok example/newmath
</pre>
<p>
This example package, <code>numbers</code>, consists of the function
<code>Double</code>, which takes an <code>int</code> and returns that value
multiplied by 2. It consists of two files.
Run <code><a href="/cmd/go/#Test_packages">go help test</a></code> and see the
<a href="/pkg/testing/">testing package documentation</a> for more detail.
</p>
<h2 id="remote">Remote packages</h3>
<p>
First, the package implementation, <code>numbers.go</code>:
An import path can describe how to obtain the package source code using a
revision control system such as Git or Mercurial. The <code>go</code> command uses
this property to automatically fetch packages from remote repositories.
For instance, the examples described in this document are also kept in a
Mercurial repository hosted at Google Code,
<code><a href="http://code.google.com/p/go.example">code.google.com/p/go.example</a></code>.
If you include the repository URL in the package's import path,
<code>go get</code> will fetch, build, and install it automatically:
</p>
<pre>
package numbers
func Double(i int) int {
return i * 2
}
$ go get code.google.com/p/go.example/hello
$ $GOPATH/bin/hello
Hello, world. Sqrt(2) = 1.414213562373095
</pre>
<p>
Next, the tests, <code>numbers_test.go</code>:
If the specified package is not present in a workspace, <code>go get</code>
will place it inside the first workspace specified by <code>GOPATH</code>.
(If the package does already exist, <code>go get</code> skips the remote
fetch and behaves the same as <code>go install</code>.)
</p>
<pre>
package numbers
import (
"testing"
)
type doubleTest struct {
in, out int
}
var doubleTests = []doubleTest{
doubleTest{1, 2},
doubleTest{2, 4},
doubleTest{-5, -10},
}
<p>
After issuing the above <code>go get</code> command, the workspace directory
tree should now now look like this:
</p>
func TestDouble(t *testing.T) {
for _, dt := range doubleTests {
v := Double(dt.in)
if v != dt.out {
t.Errorf("Double(%d) = %d, want %d.", dt.in, v, dt.out)
}
}
}
<pre>
bin/
hello # command executable
pkg/
linux_amd64/
code.google.com/p/go.example/
newmath.a # package object
example/
newmath.a # package object
src/
code.google.com/p/go.example/
hello/
hello.go # command source
newmath/
sqrt.go # package source
sqrt_test.go # test source
example/
hello/
hello.go # command source
newmath/
sqrt.go # package source
sqrt_test.go # test source
</pre>
<p>
Running <code>go install</code> will build and install the package to
the <code>GOPATH</code>'s <code>pkg</code> directory
(it can then be imported by any other Go program).
The <code>hello</code> command hosted at Google Code depends on the
<code>newmath</code> package within the same repository. The imports in
<code>hello.go</code> file use the same import path convention, so the <code>go
get</code> command is able to locate and install the dependent package, too.
</p>
<pre>
import "code.google.com/p/go.example/newmath"
</pre>
<p>
Running <code>go test</code> will rebuild the package, including the
<code>numbers_test.go</code> file, and then run the <code>TestDouble</code>
function. The output "<code>ok</code>" indicates that all tests passed
successfully. Breaking the implementation by changing the multiplier from
<code>2</code> to <code>3</code> will allow you to see how failing tests are
reported.
This convention is the easiest way to make your Go packages available for
others to use.
The <a href="http://godashboard.appspot.com/package">Go Package Dashboard</a>
displays a list of packages recently installed with the <code>go</code> command.
</p>
<p>
Run <code>go help test</code>, <code>go help testfunc</code>,
and <code>go help testflag</code> and see the
<a href="/pkg/testing/">testing package documentation</a> for more detail.
For more information on using remote repositories with the <code>go</code> command, see
<code><a href="/cmd/go/#Remote_import_path_syntax">go help remote</a></code>.
</p>
<h2 id="arch_os_specific">Architecture- and operating system-specific code</h2>
<p>First, a disclaimer: very few Go packages should need to know about the
hardware and operating system they run on. In the vast majority of cases the
language and standard library handle most portability issues. This section is
a guide for experienced systems programmers who have a good reason to write
platform-specific code, such as assembly-language support for fast
trigonometric functions or code that implements a common interface above
different operating systems.</p>
<p>To compile such code, use the <code>$GOOS</code> and <code>$GOARCH</code>
<a href="/doc/install.html#environment">environment variables</a> in your
source file names.</p>
<p>For example, consider the package <code>foo</code> that consists of four
files:</p>
<pre>
foo.go
foo_386.go
foo_amd64.go
foo_arm.go
</pre>
<p>describes a package that builds on
different architectures by parameterizing the file name with
<code>$GOARCH</code>.</p>
<h2 id="more">Further reading</h2>
<p>The general code goes in <code>foo.go</code>, while architecture-specific
code goes in <code>foo_386.go</code>, <code>foo_amd64.go</code>, and
<code>foo_arm.go</code>.</p>
<p>If you follow these conventional parameterizations, tools such as the <a
href="/cmd/go/"><code>go</code> tool</a> will work seamlessly with your
package:</p>
<p>
See <a href="/doc/effective_go.html">Effective Go</a> for tips on writing
clear, idiomatic Go code.
</p>
<pre>
foo_$GOOS.go
foo_$GOARCH.go
foo_$GOOS_$GOARCH.go
</pre>
<p>
Take <a href="http://tour.golang.org/">A Tour of Go</a> to learn the language
proper.
</p>
<p>The same holds for <code>.s</code> (assembly) and <code>.c</code> files.</p>
<p>
Visit the <a href="/doc/#articles">documentation page</a> for a set of in-depth
articles about the Go language and its libraries and tools.
</p>
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7