Home | History | Annotate | Download | only in articles
      1 <!--{
      2 	"title": "About the go command"
      3 }-->
      4 
      5 <p>The Go distribution includes a command, named
      6 "<code><a href="/cmd/go/">go</a></code>", that
      7 automates the downloading, building, installation, and testing of Go packages
      8 and commands.  This document talks about why we wrote a new command, what it
      9 is, what it's not, and how to use it.</p>
     10 
     11 <h2>Motivation</h2>
     12 
     13 <p>You might have seen early Go talks in which Rob Pike jokes that the idea
     14 for Go arose while waiting for a large Google server to compile.  That
     15 really was the motivation for Go: to build a language that worked well
     16 for building the large software that Google writes and runs. It was
     17 clear from the start that such a language must provide a way to
     18 express dependencies between code libraries clearly, hence the package
     19 grouping and the explicit import blocks.  It was also clear from the
     20 start that you might want arbitrary syntax for describing the code
     21 being imported; this is why import paths are string literals.</p>
     22 
     23 <p>An explicit goal for Go from the beginning was to be able to build Go
     24 code using only the information found in the source itself, not
     25 needing to write a makefile or one of the many modern replacements for
     26 makefiles.  If Go needed a configuration file to explain how to build
     27 your program, then Go would have failed.</p>
     28 
     29 <p>At first, there was no Go compiler, and the initial development
     30 focused on building one and then building libraries for it. For
     31 expedience, we postponed the automation of building Go code by using
     32 make and writing makefiles.  When compiling a single package involved
     33 multiple invocations of the Go compiler, we even used a program to
     34 write the makefiles for us.  You can find it if you dig through the
     35 repository history.</p>
     36 
     37 <p>The purpose of the new go command is our return to this ideal, that Go
     38 programs should compile without configuration or additional effort on
     39 the part of the developer beyond writing the necessary import
     40 statements.</p>
     41 
     42 <h2>Configuration versus convention</h2>
     43 
     44 <p>The way to achieve the simplicity of a configuration-free system is to
     45 establish conventions. The system works only to the extent that those conventions
     46 are followed. When we first launched Go, many people published packages that
     47 had to be installed in certain places, under certain names, using certain build
     48 tools, in order to be used. That's understandable: that's the way it works in
     49 most other languages. Over the last few years we consistently reminded people
     50 about the <code>goinstall</code> command
     51 (now replaced by <a href="/cmd/go/#hdr-Download_and_install_packages_and_dependencies"><code>go get</code></a>)
     52 and its conventions: first, that the import path is derived in a known way from
     53 the URL of the source code; second, that the place to store the sources in
     54 the local file system is derived in a known way from the import path; third,
     55 that each directory in a source tree corresponds to a single package; and
     56 fourth, that the package is built using only information in the source code.
     57 Today, the vast majority of packages follow these conventions.
     58 The Go ecosystem is simpler and more powerful as a result.</p>
     59 
     60 <p>We received many requests to allow a makefile in a package directory to
     61 provide just a little extra configuration beyond what's in the source code.
     62 But that would have introduced new rules. Because we did not accede to such
     63 requests, we were able to write the go command and eliminate our use of make
     64 or any other build system.</p>
     65 
     66 <p>It is important to understand that the go command is not a general
     67 build tool. It cannot be configured and it does not attempt to build
     68 anything but Go packages.  These are important simplifying
     69 assumptions: they simplify not only the implementation but also, more
     70 important, the use of the tool itself.</p>
     71 
     72 <h2>Go's conventions</h2>
     73 
     74 <p>The <code>go</code> command requires that code adheres to a few key,
     75 well-established conventions.</p>
     76 
     77 <p>First, the import path is derived in an known way from the URL of the
     78 source code.  For Bitbucket, GitHub, Google Code, and Launchpad, the
     79 root directory of the repository is identified by the repository's
     80 main URL, without the <code>http://</code> prefix.  Subdirectories are named by
     81 adding to that path.
     82 For example, the Go example programs are obtained by running</p>
     83 
     84 <pre>
     85 git clone https://github.com/golang/example
     86 </pre>
     87 
     88 <p>and thus the import path for the root directory of that repository is
     89 "<code>github.com/golang/example</code>".
     90 The <a href="https://godoc.org/github.com/golang/example/stringutil">stringutil</a>
     91 package is stored in a subdirectory, so its import path is
     92 "<code>github.com/golang/example/stringutil</code>".</p>
     93 
     94 <p>These paths are on the long side, but in exchange we get an
     95 automatically managed name space for import paths and the ability for
     96 a tool like the go command to look at an unfamiliar import path and
     97 deduce where to obtain the source code.</p>
     98 
     99 <p>Second, the place to store sources in the local file system is derived
    100 in a known way from the import path, specifically
    101 <code>$GOPATH/src/&lt;import-path&gt;</code>.
    102 If unset, <code>$GOPATH</code> defaults to a subdirectory
    103 named <code>go</code> in the user's home directory.
    104 If <code>$GOPATH</code> is set to a list of paths, the go command tries
    105 <code>&lt;dir&gt;/src/&lt;import-path&gt;</code> for each of the directories in
    106 that list.
    107 </p>
    108 
    109 <p>Each of those trees contains, by convention, a top-level directory named
    110 "<code>bin</code>", for holding compiled executables, and a top-level directory
    111 named "<code>pkg</code>", for holding compiled packages that can be imported,
    112 and the "<code>src</code>" directory, for holding package source files.
    113 Imposing this structure lets us keep each of these directory trees
    114 self-contained: the compiled form and the sources are always near each
    115 other.</p>
    116 
    117 <p>These naming conventions also let us work in the reverse direction,
    118 from a directory name to its import path. This mapping is important
    119 for many of the go command's subcommands, as we'll see below.</p>
    120 
    121 <p>Third, each directory in a source tree corresponds to a single
    122 package. By restricting a directory to a single package, we don't have
    123 to create hybrid import paths that specify first the directory and
    124 then the package within that directory.  Also, most file management
    125 tools and UIs work on  directories as fundamental units.  Tying the
    126 fundamental Go unit&mdash;the package&mdash;to file system structure means
    127 that file system tools become Go package tools.  Copying, moving, or
    128 deleting a package corresponds to copying, moving, or deleting a
    129 directory.</p>
    130 
    131 <p>Fourth, each package is built using only the information present in
    132 the source files.  This makes it much more likely that the tool will
    133 be able to adapt to changing build environments and conditions. For
    134 example, if we allowed extra configuration such as compiler flags or
    135 command line recipes, then that configuration would need to be updated
    136 each time the build tools changed; it would also be inherently tied
    137 to the use of a specific tool chain.</p>
    138 
    139 <h2>Getting started with the go command</h2>
    140 
    141 <p>Finally, a quick tour of how to use the go command.
    142 As mentioned above, the default <code>$GOPATH</code> on Unix is <code>$HOME/go</code>.
    143 We'll store our programs there.
    144 To use a different location, you can set <code>$GOPATH</code>;
    145 see <a href="/doc/code.html">How to Write Go Code</a> for details.
    146 
    147 <p>We first add some source code.  Suppose we want to use
    148 the indexing library from the codesearch project along with a left-leaning
    149 red-black tree.  We can install both with the "<code>go get</code>"
    150 subcommand:</p>
    151 
    152 <pre>
    153 $ go get github.com/google/codesearch/index
    154 $ go get github.com/petar/GoLLRB/llrb
    155 $
    156 </pre>
    157 
    158 <p>Both of these projects are now downloaded and installed into <code>$HOME/go</code>,
    159 which contains the two directories
    160 <code>src/github.com/google/codesearch/index/</code> and
    161 <code>src/github.com/petar/GoLLRB/llrb/</code>, along with the compiled
    162 packages (in <code>pkg/</code>) for those libraries and their dependencies.</p>
    163 
    164 <p>Because we used version control systems (Mercurial and Git) to check
    165 out the sources, the source tree also contains the other files in the
    166 corresponding repositories, such as related packages. The "<code>go list</code>"
    167 subcommand lists the import paths corresponding to its arguments, and
    168 the pattern "<code>./...</code>" means start in the current directory
    169 ("<code>./</code>") and find all packages below that directory
    170 ("<code>...</code>"):</p>
    171 
    172 <pre>
    173 $ cd $HOME/go/src
    174 $ go list ./...
    175 github.com/google/codesearch/cmd/cgrep
    176 github.com/google/codesearch/cmd/cindex
    177 github.com/google/codesearch/cmd/csearch
    178 github.com/google/codesearch/index
    179 github.com/google/codesearch/regexp
    180 github.com/google/codesearch/sparse
    181 github.com/petar/GoLLRB/example
    182 github.com/petar/GoLLRB/llrb
    183 $
    184 </pre>
    185 
    186 <p>We can also test those packages:</p>
    187 
    188 <pre>
    189 $ go test ./...
    190 ?   	github.com/google/codesearch/cmd/cgrep	[no test files]
    191 ?   	github.com/google/codesearch/cmd/cindex	[no test files]
    192 ?   	github.com/google/codesearch/cmd/csearch	[no test files]
    193 ok  	github.com/google/codesearch/index	0.203s
    194 ok  	github.com/google/codesearch/regexp	0.017s
    195 ?   	github.com/google/codesearch/sparse	[no test files]
    196 ?       github.com/petar/GoLLRB/example          [no test files]
    197 ok      github.com/petar/GoLLRB/llrb             0.231s
    198 $
    199 </pre>
    200 
    201 <p>If a go subcommand is invoked with no paths listed, it operates on the
    202 current directory:</p>
    203 
    204 <pre>
    205 $ cd github.com/google/codesearch/regexp
    206 $ go list
    207 github.com/google/codesearch/regexp
    208 $ go test -v
    209 === RUN   TestNstateEnc
    210 --- PASS: TestNstateEnc (0.00s)
    211 === RUN   TestMatch
    212 --- PASS: TestMatch (0.00s)
    213 === RUN   TestGrep
    214 --- PASS: TestGrep (0.00s)
    215 PASS
    216 ok  	github.com/google/codesearch/regexp	0.018s
    217 $ go install
    218 $
    219 </pre>
    220 
    221 <p>That "<code>go install</code>" subcommand installs the latest copy of the
    222 package into the pkg directory. Because the go command can analyze the
    223 dependency graph, "<code>go install</code>" also installs any packages that
    224 this package imports but that are out of date, recursively.</p>
    225 
    226 <p>Notice that "<code>go install</code>" was able to determine the name of the
    227 import path for the package in the current directory, because of the convention
    228 for directory naming.  It would be a little more convenient if we could pick
    229 the name of the directory where we kept source code, and we probably wouldn't
    230 pick such a long name, but that ability would require additional configuration
    231 and complexity in the tool. Typing an extra directory name or two is a small
    232 price to pay for the increased simplicity and power.</p>
    233 
    234 <h2>Limitations</h2>
    235 
    236 <p>As mentioned above, the go command is not a general-purpose build
    237 tool.
    238 In particular, it does not have any facility for generating Go
    239 source files <em>during</em> a build, although it does provide
    240 <a href="/cmd/go/#hdr-Generate_Go_files_by_processing_source"><code>go</code>
    241 <code>generate</code></a>,
    242 which can automate the creation of Go files <em>before</em> the build.
    243 For more advanced build setups, you may need to write a
    244 makefile (or a configuration file for the build tool of your choice)
    245 to run whatever tool creates the Go files and then check those generated source files
    246 into your repository. This is more work for you, the package author,
    247 but it is significantly less work for your users, who can use
    248 "<code>go get</code>" without needing to obtain and build
    249 any additional tools.</p>
    250 
    251 <h2>More information</h2>
    252 
    253 <p>For more information, read <a href="/doc/code.html">How to Write Go Code</a>
    254 and see the <a href="/cmd/go/">go command documentation</a>.</p>
    255