Chapter 10. Packages and the Go Tool

In a program, the majority of the functions were written by others and made available for reuse through packages.

Go has over 100 standard packages that provide the foundations for most applications. The Go community has published many more, and you can find a searchable index of them at This chapter discusses how to use existing packages and create new ones.

Go also comes with the go tool, a sophisticated but simple-to-use command for managing workspaces of Go packages. Since the beginning of the book, we’ve been showing how to use the go tool to download, build, and run example programs. This chapter studies the tools' underlying concepts and its capabilities, which include printing documentation and querying metadata about the packages in the workspace.


The purpose of any package system is to make the design and maintenance of large programs practical by grouping related features together into units that can be easily understood and changed, independent of the other packages of the program.

This modularity allows packages to be:

Each package defines a distinct name space that encloses its identifiers. Each name is associated with a particular package, so we can choose short, clear names for the types, functions, etc. that we use most often, without creating conflicts with other parts of the program.

Packages also provide encapsulation by controlling which names are visible or exported outside the package. Restricting the visibility of package members has two advantages:

When we change a file, we must recompile the file's package and potentially all the packages that depend on it. Go compilation is notably faster than most other compiled languages, even when building from scratch. There are three main reasons for the compiler's speed.

  1. All imports must be explicitly listed at the beginning of each source file, so the compiler does not have to read and process an entire file to determine its dependencies.
  2. The dependencies of a package form a directed acyclic graph, and because there are no cycles, packages can be compiled separately and perhaps in parallel.
  3. The object file for a compiled Go package records export information not just for the package itself, but also for its dependencies. When compiling a package, the compiler must read one object file for each import but need not look beyond these files.

Import Paths

Each package is identified by a unique string called its import path. Import paths are the strings that appear in import declarations.

import (

As mentioned in Section 2.6.1, the Go language specification doesn't define the meaning of these strings or how to determine a packages' import path, but leaves these issues to the tools. This chapter discusses how the go tool interprets them, which is what the majority of Go programmers use for building, testing, and so on. However, other tools do exist. For example, Go programmers using Google's internal multi-language build system follow different rules for naming and locating packages, specifying tests, and so on, that more closely match the conventions of that system.

For packages you intend to share or publish, import paths should be globally unique. To avoid conflicts, the import paths of all packages other than those from the standard library should start with the Internet domain name of the organization that owns or hosts the package; this also makes it possible to find packages.

The Package Declaration

A package declaration is required at the start of every Go source file. Its main purpose is to determine the default identifier for that package (called the package name) when it is imported by another package.

For example, every file of the math/rand package starts with package rand, so when you import this package, you can access its members as rand.Int, rand.Float64, and so on.

package main

import (

func main() {

The "Last Segment" Convention *

Conventionally, the package name is the last segment of the import path, and as a result, two packages may have the same name even though their import paths differ. For example, the packages whose import paths are math/rand and crypto/rand both have the name rand. How to use both in the same program will be discussed in a moment.

There are three major exceptions to the "last segment" convention.

  1. A package defining a command (an executable Go program) always has the name main, regardless of the packages' import path. This is a signal to go build (Section 10.7.3) that it must invoke the linker to make an executable file.
  2. Some files in the directory may have the suffix _test on their package name if the file name ends with _test.go. Such a directory may define two packages: the usual one, plus another one called an external test package. The _test suffix signals to go test that it must build both packages, and it indicates which files belong to each package. External test packages are used to avoid cycles in the import graph arising from dependencies of the test; they are covered in more detail in Section 11.2.4.
  3. Some tools for dependency management append version number suffixes to package import paths, such as "". The package name excludes the suffix, so in this case it would be just yaml.

Import Declarations

A Go source file may contain zero or more import declarations immediately after the package declaration and before the first non-import declaration. Each import declaration may specify the import path of a single package, or multiple packages in a parenthesized list. The two forms below are equivalent but the second form is more common.

import "fmt"
import "os"
import (

Imported packages may be grouped by introducing blank lines; such groupings usually indicate different domains. The order is not significant, but by convention the lines of each group are sorted alphabetically. (Both gofmt and goimports will group and sort for you.)

import (


If you need to import two packages whose names are the same, like math/rand and crypto/rand, into a third package, the import declaration must specify an alternative name for at least one of them to avoid a conflict. This is called a renaming import.

import (
    mrand "math/rand" // alternative name mrand avoids conflict

The alternative name affects only the importing file, not other files (even ones in the same package).

A renaming import may be useful even when there is no conflict:

Each import declaration establishes a dependency from the current package to the imported package. The go build tool reports an error if these dependencies form a cycle.

Blank Imports

Within a file, it is an error to import a package but not refer to the name it defines. However, on occasion we must import a package merely for the side effects of doing so: evaluation of the initializer expressions of its package-level variables and execution of its init functions (Section 2.6.2). To suppress the "unused import" error, we must use a renaming import in which the alternative name is _, the blank identifier, which can never be referenced.

import _ "image/png" // register PNG decoder

This is known as a blank import. It is most often used to implement a compile-time mechanism whereby the main program can enable optional features by blank-importing additional packages.

As an example, the standard library's image package exports a Decode function that reads bytes from an io.Reader, figures out which image format was used to encode the data, invokes the appropriate decoder, then returns the resulting image.Image. Using image.Decode, it's easy to build a simple image converter that reads an image in one format and writes it out in another:

// The jpeg command reads a PNG image from the standard input
// and writes it as a JPEG image to the standard output.
package main

import (
    _ "image/png" // register PNG decoder

func main() {
    if err := toJPEG(os.Stdin, os.Stdout); err != nil {
        fmt.Fprintf(os.Stderr, "jpeg: %v\n", err)

func toJPEG(in io.Reader, out io.Writer) error {
    img, kind, err := image.Decode(in)
    if err != nil {
        return err
    fmt.Fprintln(os.Stderr, "Input format =", kind)
    return jpeg.Encode(out, img, &jpeg.Options{Quality: 95})

If we feed the output of (Section 3.3) to the converter program, it detects the PNG input format and writes a JPEG image.

$ go build
$ go build
$ ./mandelbrot | ./jpeg >mandelbrot.jpg

If without the blank import of image/png, the program compiles and links as usual but can no longer recognize or decode input in PNG format:

$ go build
$ ./mandelbrot | ./jpeg >mandelbrot.jpg
jpeg: image: unknown format

The cause is: the standard library provides decoders for GIF, PNG, and JPEG, and users may provide others, but to keep executables small, decoders are not included in an application unless explicitly requested.

The image.Decode function consults a table of supported formats. Each entry in the table specifies four things:

  1. The name of the format
  2. A string that is a prefix of all images encoded this way, used to detect the encoding
  3. A function Decode that decodes an encoded image
  4. Another function DecodeConfig that decodes only the image metadata, such as its size and color space

An entry is added to the table by calling image.RegisterFormat, typically from within the package initializer of the supporting package for each format, like this one in image/png:

package png // image/png

func Decode(r io.Reader) (image.Image, error)
func DecodeConfig(r io.Reader) (image.Config, error)

func init() {
    const pngHeader = "\x89PNG\r\n\x1a\n"
    image.RegisterFormat("png", pngHeader, Decode, DecodeConfig)

The effect is that an application need only blank-import the package for the format it needs to make the image.Decode function able to decode it.

The database/sql package uses a similar mechanism allowing users install just the database drivers they need. For example:

import (
    _ ""              // enable support for Postgres
    _ "" // enable support for MySQL

db, err = sql.Open("postgres", dbname) // OK
db, err = sql.Open("mysql", dbname)    // OK
db, err = sql.Open("sqlite3", dbname)  // returns error: unknown driver "sqlite3"

Packages and Naming

This section offers some advice on how to follow Go's distinctive conventions for naming packages and their members.

The package name should be short, but not so short as to be cryptic. The most frequently used packages in the standard library are named bufio, bytes, flag, fmt, http, io, json, os, sort, sync, and time.

Be descriptive and unambiguous where possible. For example, don't name a utility package util when a name such as imageutil or ioutil is specific yet still concise. Avoid choosing package names that are commonly used for related local variables, or you may compel the package's clients to use renaming imports, as with the path package.

Package names usually take the singular form, with some exceptions:

Avoid package names that already have other connotations. For example, we originally used the name temp for the temperature conversion package in Section 2.5, but that didn't last long. It was a terrible idea because "temp" is an almost universal synonym for "temporary". We went through a brief period with the name temperature, but that was too long and didn't say what the package did. In the end, it became tempconv, which is shorter and parallel with strconv.

Since each reference to a member of another package uses a qualified identifier such as fmt.Println, the burden of describing the package member is shared equally by the package name and the member name. We need not mention the concept of formatting in Println because the package name fmt does that already. When designing a package, consider how the two parts of a qualified identifier work together, not the member name alone. For example:

We can identify some common naming patterns. The strings package provides a number of independent functions for manipulating strings:

package strings

func Index(needle, haystack string) int

type Replacer struct{ /* ... */ }
func NewReplacer(oldnew ...string) *Replacer

type Reader struct{ /* ... */ }
func NewReader(s string) *Reader

The word string does not appear in any of their names. Clients refer to them as strings.Index, strings.Replacer, and so on.

Other packages that we might describe as single-type packages, such as html/template and math/rand, expose one principal data type plus its methods, and often a New function to create instances.

package rand // "math/rand"

type Rand struct{ /* ... */ }
func New(source Source) *Rand

This can lead to repetition, as in template.Template or rand.Rand, which is why the names of these kinds of packages are often especially short.

At the other extreme, there are packages like net/http that have a lot of names without a lot of structure, because they perform a complicated task. Despite having over twenty types and many more functions, the package's most important members have the simplest names: Get, Post, Handle, Error, Client, Server.

The Go Tool

The rest of this chapter concerns the go tool, which is used for downloading, querying, formatting, building, testing, and installing packages of Go code.

The go tool combines the features of a diverse set of tools into one command set:

Its command-line interface uses the "Swiss army knife" style, with over a dozen subcommands, some of which we have already seen, like get, run, build, and fmt. You can run go help to see the index of its built-in documentation. The most commonly used commands are:

$ go
    build     compile packages and dependencies
    clean     remove object files
    doc       show documentation for package or symbol
    env       print Go environment information
    fmt       run gofmt on package sources
    get       download and install packages and dependencies
    install   compile and install packages and dependencies
    list      list packages
    run       compile and run Go program
    test      test packages
    version   print Go version
    vet       run go tool vet on packages
Use "go help [command]" for more information about a command.

To keep the need for configuration to a minimum, the go tool relies heavily on conventions. For example:

Workspace Organization

The only configuration most users ever need is the GOPATH environment variable, which specifies the root of the workspace. When switching to a different workspace, users update the value of GOPATH. For example:

$ export GOPATH=$HOME/gobook
$ go get

After you download all the programs for this book using the command above, your workspace will contain a hierarchy like this one:


GOPATH has three subdirectories:

A second environment variable, GOROOT, specifies the root directory of the Go distribution, which provides all the packages of the standard library. The directory structure beneath GOROOT resembles that of GOPATH. For example, the source files of the fmt package reside in the $GOROOT/src/fmt directory. Users never need to set GOROOT since, by default, the go tool will use the location where it was installed.

The go env command prints the effective values of the environment variables relevant to the toolchain, including the default values for the missing ones. GOOS specifies the target operating system (for example, android, linux, darwin, or windows) and GOARCH specifies the target processor architecture, such as amd64, 386, or arm. Although GOPATH is the only variable you must set, the others occasionally appear in our explanations.

$ go env

Downloading Packages

When using the go tool, a package's import path indicates not only where to find it in the local workspace, but where to find it on the Internet so that go get can retrieve and update it.

The go get command can download a single package or an entire subtree or repository using the ... notation, as in the previous section. The tool also computes and downloads all the dependencies of the initial packages, which is why the package appeared in the workspace in the previous example.

Once go get has downloaded the packages, it builds them and then installs the libraries and commands.

In the following example:

$ go get
$ $GOPATH/bin/golint
  package comment should be of the form "Package popcount ..."

The go get command has support for popular code-hosting sites like GitHub, Bitbucket, and Launchpad and can make the appropriate requests to their version-control systems. For less well-known sites, you may have to indicate which version-control protocol to use in the import path, such as Git or Mercurial. Run go help importpath for the details.

The directories that go get creates are true clients of the remote repository, not just copies of the files, so you can use version-control commands to see a diff of local edits you've made or to update to a different revision. For example, the directory is a Git client:

$ cd $GOPATH/src/
$ git remote -v
origin (fetch)
origin (push)

Notice that the apparent domain name in the package's import path,, differs from the actual domain name of the Git server, This is a feature of the go tool that lets packages use a custom domain name in their import path while being hosted by a generic service such as or HTML pages beneath include the metadata shown below, which redirects the go tool to the Git repository at the actual hosting site:

$ go build
$ ./fetch | grep go-import
<meta name="go-import"
      content=" git">

If you specify the -u flag, go get will ensure that all packages it visits, including dependencies, are updated to their latest version before being built and installed. Without that flag, packages that already exist locally will not be updated.

Vendor Directories *

The go get -u command generally retrieves the latest version of each package, which is convenient when you're getting started but may be inappropriate for deployed projects, where precise control of dependencies is critical for release hygiene. The usual solution to this problem is to vendor the code, that is, to make a persistent local copy of all the necessary dependencies, and to update this copy carefully and deliberately. Prior to Go 1.5, this required changing those packages' import paths, so our copy of would become More recent versions of the go tool support vendoring directly. See Vendor Directories in the output of the go help gopath command.

Building Packages

The go build command compiles each argument package.

Since each directory contains one package, each executable program, or command in Unix terminology, requires its own directory. These directories are sometimes children of a directory named cmd, such as the command which serves Go package documentation through a web interface (Section 10.7.4).

Packages may be specified by their import paths, or by a relative directory name, which must start with a . or .. segment even if this would not ordinarily be required. If no argument is provided, the current directory is assumed. Thus the following commands build the same package, though each writes the executable to the directory in which go build is run:

$ cd $GOPATH/src/
$ go build


$ cd anywhere
$ go build


$ cd $GOPATH
$ go build ./src/

but not:

$ cd $GOPATH
$ go build src/
Error: cannot find package "src/".

Packages may also be specified as a list of file names, though this tends to be used only for small programs and one-off experiments. If the package name is main, the executable name comes from the basename of the first .go file.

$ cat quoteargs.go
package main
import (
func main() {
    fmt.Printf("%q\n", os.Args[1:])
$ go build quoteargs.go
$ ./quoteargs one "two three" four\ five
["one" "two three" "four five"]

For throwaway programs like this one, we can run the executable as soon as we've built it. The go run command combines these two steps:

$ go run quoteargs.go one "two three" four\ five
["one" "two three" "four five"]

By default, the go build command builds the requested package and all its dependencies, then throws away all the compiled code except the final executable, if any. Both the dependency analysis and the compilation are surprisingly fast, but as projects grow to dozens of packages and hundreds of thousands of lines of code, the time to recompile dependencies can become noticeable, potentially several seconds, even when those dependencies haven't changed at all.

The go install command is very similar to go build, except that it saves the compiled code for each package and command instead of throwing it away. Compiled packages are saved beneath the $GOPATH/pkg directory corresponding to the src directory in which the source resides, and command executables are saved in the $GOPATH/bin directory. Thereafter, go build and go install do not run the compiler for those packages and commands if they have not changed, making subsequent builds much faster. For convenience, go build -i installs the packages that are dependencies of the build target.

Since compiled packages vary by platform and architecture, go install saves them beneath a subdirectory whose name incorporates the values of the GOOS and GOARCH environment variables. For example, on a Mac the package is compiled and installed in the file under $GOPATH/pkg/darwin_amd64.

It is straightforward to cross-compile a Go program, that is, to build an executable intended for a different operating system or CPU. This can be done by setting the GOOS or GOARCH variables during the build. The cross program below prints the operating system and architecture for which it was built:

func main() {
    fmt.Println(runtime.GOOS, runtime.GOARCH)

The following commands produce 64-bit and 32-bit executables respectively:

$ go build
$ ./cross
darwin amd64
$ GOARCH=386 go build
$ ./cross
darwin 386

Some packages may need to compile different versions of the code for certain platforms or processors, for example, to deal with low-level portability issues or to provide optimized versions of important routines. If a file name includes an operating system or processor architecture name like net_linux.go or asm_amd64.s, then the go tool will compile the file only when building for that target.

Special comments called build tags give more fine-grained control. For example, if a file contains the following comment before the package declaration (and its doc comment):

// +build linux darwin

go build will compile it only when building for Linux or Mac OS X.

The following comment says never to compile the file:

// +build ignore

For more details, see the Build Constraints section of the go/build package's documentation:

$ go doc go/build

Documenting Packages

Go style strongly encourages good documentation of package APIs. Each declaration of an exported package member and the package declaration itself should be immediately preceded by a comment explaining its purpose and usage.

Go doc comments are always complete sentences, and the first sentence is usually a summary that starts with the name being declared. Function parameters and other identifiers are mentioned without quotation or markup. For example, the following is the doc comment for fmt.Fprintf:

// Fprintf formats according to a format specifier and writes to w.
// It returns the number of bytes written and any write error encountered.
func Fprintf(w io.Writer, format string, a ...interface{}) (int, error)

The details of Fprintf's formatting are explained in a doc comment associated with the fmt package itself. A comment immediately preceding a package declaration is considered the doc comment for the package as a whole. There must be only one, though it may appear in any file. Longer package comments may warrant a file of their own. For example, fmt's package comment is over 300 lines. This file is usually called doc.go (e.g. fmt/doc.go).

Go's conventions favor brevity and simplicity in documentation as in all things, since documentation also requires maintenance just like code. Many declarations can be explained in one well-worded sentence, and if the behavior is truly obvious, no comment is needed.

Two tools can help you find better examples as you browse the standard library.

go doc *

The go doc tool prints the declaration and doc comment of the entity specified on the command line, which may be a package:

$ go doc time
package time // import "time"

Package time provides functionality for measuring and displaying time.

const Nanosecond Duration = 1 ...
func After(d Duration) <-chan Time
func Sleep(d Duration)
func Since(t Time) Duration
func Now() Time
type Duration int64
type Time struct { ... }

or a package member:

$ go doc time.Since
func Since(t Time) Duration

    Since returns the time elapsed since t.
    It is shorthand for time.Now().Sub(t).

or a method:

$ go doc time.Duration.Seconds
func (d Duration) Seconds() float64

    Seconds returns the duration as a floating-point number of seconds.

The tool does not need complete import paths or correct identifier case. For example, this command prints the documentation of (*json.Decoder).Decode from the encoding/json package:

$ go doc json.decode
func (dec *Decoder) Decode(v interface{}) error

    Decode reads the next JSON-encoded value from its input and stores
    it in the value pointed to by v.
godoc *

The second tool, confusingly named godoc, serves cross-linked HTML pages that provide the same information as go doc and much more. The godoc server at covers the standard library. The godoc server at has a searchable index of thousands of open-source packages.

You can also run an instance of godoc in your workspace if you want to browse your own packages. Visit http://localhost:8000/pkg in your browser while running this command:

$ godoc -http :8000

Its -analysis=type and -analysis=pointer flags augment the documentation and the source code with the results of advanced static analysis.

Internal Packages

The package is the most important mechanism for encapsulation in Go programs.

Sometimes a middle ground would be helpful, a way to define identifiers that are visible to a small set of trusted packages, but not to everyone. For example:

To address these needs, the go build tool treats a package specially if its import path contains a path segment named internal. Such packages are called internal packages. An internal package may be imported only by another package that is inside the tree rooted at the parent of the internal directory. For example, given the packages below, net/http/internal/chunked can be imported from net/http/httputil or net/http, but not from net/url. However, net/url may import net/http/httputil.


Querying Packages

The go list tool reports information about available packages. In its simplest form, go list tests whether a package is present in the workspace and prints its import path if so:

$ go list

An argument to go list may contain the "..." wildcard, which matches any substring of a package's import path. We can use it to enumerate all the packages within a Go workspace:

$ go list ...

Or within a specific subtree:

$ go list

Or related to a particular topic:

$ go list ...xml...

The go list command obtains the complete metadata for each package. The -json flag causes go list to print the entire record of each package in JSON format:

$ go list -json hash
    "Dir": "/home/gopher/go/src/hash",
    "ImportPath": "hash",
    "Name": "hash",
    "Doc": "Package hash provides interfaces for hash functions.",
    "Target": "/home/gopher/go/pkg/darwin_amd64/hash.a",
    "Goroot": true,
    "Standard": true,
    "Root": "/home/gopher/go",
    "GoFiles": [
    "Imports": [
    "Deps": [

The -f flag customizes the output format using the template language of package text/template (Section 4.6). The following command prints the transitive dependencies of the strconv package, separated by spaces:

$ go list -f '{{join .Deps " "}}' strconv
errors math runtime unicode/utf8 unsafe

The following command prints the direct imports of each package in the compress subtree of the standard library:

$ go list -f '{{.ImportPath}} -> {{join .Imports " "}}' compress/...
compress/bzip2 -> bufio io sort
compress/flate -> bufio fmt io math sort strconv
compress/gzip -> bufio compress/flate errors fmt hash hash/crc32 io time
compress/lzw -> bufio errors fmt io
compress/zlib -> bufio compress/flate errors fmt hash hash/adler32 io

The go list command is useful for both one-off interactive queries and for build and test automation scripts. It will be used again in Section 11.2.4. For more information, including the set of available fields and their meaning, see the output of go help list.