go/doc/effective_go.html

<h2 id="introduction">Introduction</h2>

<p>
Go is a new programming language.
It eliminates some of the pitfalls in languages
like C++ and Java but introduces other ones.
Many remain the same.
This page gives tips for writing clear, idiomatic Go code
and points out common mistakes to avoid.
It augments the <a href="go_spec.html">language specification</a>
and the <a href="go_tutorial.html">tutorial</a>, both of which you
should be familiar with.
</p>

<h3 id="read">Read</h3>

<p>
The first step to improving as a writer is to read.
This step is as necessary for programming as it is for prose,
and it is skipped as often by programmers as by writers.
The <a href="/src/pkg/">Go package sources</a>
are intended to serve not
only as the core library but also as examples of how to
use the language.  Read them.
</p>

<h3 id="be-consistent">Be consistent</h3>

<p>
Consistency makes programs easy to read.
If a program says the same thing twice,
it should say it the same way both times,
so that the parallel structure is clear.
Conversely, if two different sections of
program look different, the reader will
expect them to be doing different things.
</p>

<p>
Consider <code>for</code> loops.
Traditionally, a loop over <code>n</code>
elements begins:
</p>

<pre>
for i := 0; i &lt; n; i++ {
</pre>

<p>
Much of the time, the loop could run in the opposite order
and still be correct:
</p>

<pre>
for i := n-1; i &gt;= 0; i-- {
</pre>

<p>
The convention in most languages (including Go)
is to count up unless doing so would be incorrect.
A loop that counts down implicitly says &ldquo;there's
a specific reason to count down here; this situation is special.&rdquo;
A reader who finds a program in which half the
loops count up and the other half count down
will spend time trying to understand why.
A programmer who inserts counting-down
loops for variety wastes the reader's time.
</p>

<p>
Loop direction is hardly the only
programming decision which a programmer
might want to use to be distinctive:
tabs or spaces, choice of variable names,
choice of method names, whether a type
has a constructor, what tests look like, and on and on.
As in the loop example, inconsistency
sows confusion, and wastes time.
Why is this variable called n here and cnt here?
Why is the <code>Log</code> constructor <code>CreateLog</code> when
the <code>Vector</code> constructor is <code>NewVector</code>?
Why is this data structure initialized using
a structure literal when this other one
is initialized using individual assignments?
Why is this idiom used here but not there?
And on and on.
These questions distract from the important one:
what does the code do?
Being consistent about little things
lets readers concentrate on big ones.
</p>

<p>
This document describes how to use Go effectively.
Equally important, it describes how to use Go idiomatically,
so that a Go programmer seeing your code for
the first time can focus on what it does 
and not why it is inconsistent with typical Go practices.
Consistency trumps every item listed below.
When editing code, read the surrounding context
and try to mimic it as much as possible, even if it
disagrees with the rules here.
It should not be possible to tell which lines
you wrote or edited.
</p>

<p>
Internal consistency is important in a single file,
but source files do not exist in isolation.
Code must be consistent with the surrounding source file as well,
or the same problems arise.
</p>

<h2 id="formatting">Formatting</h2>

<p>
Formatting issues are the most contentious
but the least consequential.
People adapt quite well to different formatting styles,
even if at first the styles &ldquo;look weird.&rdquo;
The most important consideration is that if everyone
uses the same formatting, then a reader picking up an
unfamiliar piece of code can focus on what
the code does instead of what it looks like.
Most of the local formatting style can and should be
picked up by reading existing Go programs (see above).
The following are the most common issues 
that Google programmers run into.
</p>

<h3 id="tabs">Use tabs</h3>

<p>
The local style is to use tabs, not spaces, for indentation.
</p>

<h3 id="white-space">Trim trailing white space</h3>

<p>
Files should not contain trailing white space at the end of lines.
The script <code>/home/rsc/bin/g4ws</code> removes trailing
whitespace from all open files in the current g4 client.
</p>

<h3 id="line-wrapping">Don't wrap lines mechanically</h3>

<p>
Go has no 80-character limit.  Don't bother with fancy line
wrapping just because a line is wider than a punched card.
If you must wrap a line, indent with a single tab.
</p>

<h3 id="parens">Omit parentheses</h3>

<p>Go does not require parentheses around the expression
following the <code>for</code>, <code>if</code>, <code>range</code>,
and <code>switch</code> keywords.
</p>

<h3 id="line-comments">Use line comments</h3>

<p>
Go provides C-style <code>/* */</code> block comments 
and C++-style <code>//</code> line comments.
The local style is to use line comments by default,
reserving block comments for top-level package comments
and commenting out large swaths of code.
</p>

<h3 id="doc-comments">Write doc comments</h3>

<p>
If a comment immediately precedes a top-level declaration,
the <a href="/">Go documentation server</a>
uses that comment as the documentation
for the constant, function, package, type or variable being declared.
To detach a comment from a declaration, insert a blank
line between them.
</p>

<p>
Every exported (capitalized) name in a program should
have a doc comment, as should the package declaration itself.
</p>

<p>
Doc comments consist of complete English sentences.
The first sentence should be a one-sentence summary that
starts with the name being declared:
</p>

<pre>
// Quote returns a double-quoted Go string literal
// representing s.  The returned string s uses Go escape
// sequences (\t, \n, \xFF, \u0100) for control characters
// and non-ASCII characters.
func Quote(s string) string {
</pre>

<p>
instead of:
</p>

<pre class="bad">
/* not Go style */
// Return a double-quoted Go string literal representing s....
func Quote(s string) string {
</pre>

<p>
The complete English sentence form admits
a wider variety of automated presentations.
</p>


<h3 id="ascii-art">Avoid ASCII Art</h3>

<p>
Go programs are meant to read equally well using 
fixed-width and variable-width fonts.
Don't use fancy formattings that depend on fixed-width fonts.
In particular, don't assume that a single space is the same
width as every other character.
If you need to make a columnated table, use tabs to separate
the columns and the pretty printer (in progress) will make
sure the columns are lined up properly in the output.
</p>

<p>
If you must use comments to separate
sections in a file, use a simple block comment:
</p>

<pre>
/*
 * Helper routines for simplifying the fetching of optional fields of basic type.
 * If the field is missing, they return the zero for the type.
 */
</pre>

<p>
instead of:
</p>

<pre class="bad">
/* not Go style */
//////////////////////////////////////////////////////////////////////
// Helper routines for simplifying the fetching of optional fields of basic type.
// If the field is missing, they return the zero for the type.
</pre>

<p>
Comments are text, not HTML, and not any kind of markup.
Refrain from ASCII embellishment like *this* or /this/.
As usual, read the Go sources for examples.
</p>

<h2 id="names">Names</h2>

<h3 id="mixed-caps">Use MixedCaps</h3>

<p>
Go uses the case of the first letter in a name to decide
whether the name is visible in other packages.
In Go, multiword names use MixedCaps or mixedCaps
rather than underscores.
</p>

<h3 id="package-names">Use short package names</h3>

<p>
Package names are lowercase single-word names:
there should be no need for underscore or mixedCaps.
The package name is conventionally the base name of
the source directory: the package in <code>src/pkg/container/vector</code>
is installed as <code>"container/vector"</code> but has name vector,
not <code>container_vector</code> and not <code>containerVector</code>.
The package name is only the default name used
when importing the package; it need not be a unique
identifier.
</p>

<h3 id="name-length">Avoid lengthy names</h3>

<p>
A name's length should not exceed its information content.
For a function-local variable
in scope only for a few lines, the name <code>i</code> conveys just
as much information as <code>index</code> or <code>idx</code> and is easier to read.
On the same note, <code>i</code> and <code>j</code> are better pair of names for
index variables than <code>i1</code> and <code>i2</code> (or, worse, <code>index1</code> and <code>index2</code>),
because they are easier to tell apart when reading
the program quickly.
</p>

<p>
Exported names must convey more information,
because they appear in a larger variety of contexts.
Even so, longer names are not always better,
and the package name can help convey information:
the buffered <code>Reader</code> is <code>bufio.Reader</code>, not <code>bufio.BufReader</code>.
Similarly, <code>once.Do</code> is as precise and evocative as
<code>once.DoOrWaitUntilDone</code>, and <code>once.Do(f)</code> reads
better than <code>once.DoOrWaitUntilDone(f)</code>.
Contrary to popular belief, encoding small essays into
function names does not make it possible
to use them without documentation.
</p>

<h3 id="interfacers">Use the -er convention for interface names</h3>

<p>
One-method interfaces are conventionally named by
the method name plus the -er suffix: <code>Reader</code>,
<code>Writer</code>, <code>Formatter</code>.  Using an interface name distinct
from the method name keeps an anonymous struct
field of type <code>Reader</code> from conflicting with its own 
<code>Read</code> method.
</p>

<h3 id="common-names">Use canonical names</h3>

<p>
A few method names—<code>Read</code>, <code>Write</code>, <code>Close</code>, <code>Flush</code>, <code>String</code>—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, equally important, the same signature.
</p>

<p>
Some function-local variables have canonical names too.
Just as <code>i</code> is idiomatic in Go for an
index variable, <code>n</code> is idiomatic for a count, <code>b</code> for a <code>[]byte</code>,
<code>s</code> for a <code>string</code>, <code>r</code> for a <code>Reader</code>,
<code>err</code> for an <code>os.Error</code>
and so on.
Don't mix shorthands: it is especially confusing to
have two different variables <code>i</code> and <code>idx</code>,
or <code>n</code> and <code>cnt</code>.
</p>

<h2 id="idioms">Idioms</h2>

TODO: Add links to code once godoc can handle it.

<h3 id="address-literals">Address literals to allocate and initialize</h3>

<p>
Taking the address of a struct or array literal evaluates to a
new instance each time it is evaluated.
Use these expressions to avoid the repetition of filling
out a data structure.
</p>


<h3 id="buffer-slice">Use parallel assignment to slice a buffer</h3>

<pre>
hdr, body, checksum := buf[0:20], buf[20:len(buf)], buf[len(buf)-4:len(buf)];
</pre>

<h2 id="control-flow">Control Flow</h2>

<h3 id="else">Omit needless else bodies</h3>

<p>
If an <code>if</code> body doesn't flow off the end of the
body—that is, the body ends in <code>break</code>, <code>continue</code>,
<code>goto</code>, or <code>return</code>—it is preferable to omit the <code>else</code>.
</p>

<p>
For example:
</p>

<pre>
f, err := os.Open(name, os.O_RDONLY, 0);
if err != nil {
	return err;
}
codeUsing(f);
f.Close();
moreCode();
</pre>

<p>
is preferable to:

<pre class="bad">
/* not Go style */
if f, err := os.Open(name, os.O_RDONLY, 0); err != nil {
	return err;
} else {
	codeUsing(f);
	f.Close();
}
moreCode();
</pre>

<p>
The first form
avoids unnecessary indentation
and makes it clear that <code>moreCode()</code>
only runs when <code>f.Close()</code> does.
</p>

<h3 id="switch">Switch</h3>

<p>
Go's <code>switch</code> is more powerful than C's.
When an <code>if</code>-<code>else if</code>-<code>else</code> chain has three or more bodies,
or an <code>if</code> condition has a long list of alternatives,
consider rewriting it using <code>switch</code>.
</p>

<a href="/src/pkg/bytes/bytes.go">go/src/pkg/bytes/bytes.go</a>:
<pre>
// Compare returns an integer comparing the two byte arrays lexicographically.
// The result will be 0 if a==b, -1 if a &lt; b, and +1 if a &gt; b
func Compare(a, b []byte) int {
    for i := 0; i &lt; len(a) &amp;&amp; i &lt; len(b); i++ {
        switch {
        case a[i] &gt; b[i]:
            return 1
        case a[i] &lt; b[i]:
            return -1
        }
    }
    switch {
    case len(a) &lt; len(b):
        return -1
    case len(a) &gt; len(b):
        return 1
    }
    return 0
}
</pre>

<a href="/src/pkg/http/url.go">go/src/pkg/http/url.go</a>:
<pre>
func unhex(c byte) byte {
    switch {
    case '0' &lt;= c &amp;&amp; c &lt;= '9':
        return c - '0'
    case 'a' &lt;= c &amp;&amp; c &lt;= 'f':
        return c - 'a' + 10
    case 'A' &lt;= c &amp;&amp; c &lt;= 'F':
        return c - 'A' + 10
    }
    return 0
}
</pre>

<a href="/src/pkg/http/url.go">go/src/pkg/http/url.go</a>:
<pre>
func shouldEscape(c byte) bool {
    switch c {
    case ' ', '?', '&', '=', '#', '+', '%':
        return true
    }
    return false
}
</pre>

<h2 id="functions">Functions</h2>

<h3 id="omit-wrappers">Omit needless wrappers</h3>

<p>
Functions are great for factoring out common functionality.
If a function is only called once,
ask whether the function is really necessary,
especially if it is just a short wrapper around another function.
This style runs rampant in C++ code: wrappers
call wrappers that call wrappers that call wrappers.
Doing this hinders people trying to understand the program,
not to mention computers trying to execute it.
</p>

<h3 id="multiple-returns">Return multiple values</h3>

<p>
If a function must return multiple values, it can
do so directly.
The C &ldquo;pass in a pointer to a return value&rdquo;
idiom is dead.
</p>


<h2 id="errors">Errors</h2>

<h3 id="handle-errors-first">Handle errors first</h3>

<p>
Errors tend to be simpler than non-error cases,
and it helps readability when the non-error flow
of control is always down the page.
Also, error cases tend to end in jumps,
so that there is <a href="#else">no need for an explicit else</a>.
</p>

<pre>
f, err := os.Open(name, os.O_RDONLY, 0);
if err != nil {
	return err;
}
codeUsing(f);
f.Close();
moreCode();
</pre>

is preferable to:

<pre class="bad">
/* not Go style */
f, err := os.Open(name, os.O_RDONLY, 0);
if err == nil {
	codeUsing(f);
	f.Close();
} else {
	return err;
}
moreCode();
</pre>

<h3 id="error-returns">Return <code>os.Error</code>, not <code>bool</code></h3>

<p>
Few functions have just one failure mode.
Instead of returning a boolean to signal success,
return an <code>os.Error</code> that describes the failure.
Even if there is only one failure mode now, 
there may be more later.
</p>

<h3 id="error-context">Return structured errors</h3>

Implementations of <code>os.Error</code>s should
describe the error but also include context.
For example, <code>os.Open</code> returns an <code>os.PathError</code>:

<a href="/src/pkg/os/file.go">/src/pkg/os/file.go</a>:
<pre>
XXX definition of PathError and .String
</pre>

<code>PathError</code>'s <code>String</code> formats
the error nicely and is the usual way the error gets used.
Callers that care about the precise error details can
use a type switch or a type guard to look for specific
errors and then extract details.

<pre>
XXX example here - MkdirAll
</pre>

<h2 id="types">Programmer-defined types</h2>

<h3 id="constructors">Use <code>NewTypeName</code> for constructors</h3>

<p>
The constructor for the type <code>pkg.MyType</code> should
be named <code>pkg.NewMyType</code> and should return <code>*pkg.MyType</code>.
The implementation of <code>NewTypeName</code> often uses the
<a href="#allocating-a-struct">struct allocation idiom</a>.
</p>

<a href="xxx">go/src/pkg/os/file.go</a>:
<pre>
func NewFile(fd int, name string) *File {
	if file &lt; 0 {
		return nil
	}
	return &amp;File{fd, name, nil, 0}
}
</pre>

<p>Packages that export only a single type sometimes
shorten <code>NewTypeName</code> to <code>New</code>;
for example, the vector constructor is
<code>vector.New</code>, not <code>vector.NewVector</code>.
</p>

<p>
A type that is intended to be allocated
as part of a larger struct may have an <code>Init</code> method
that must be called explicitly.
Conventionally, the <code>Init</code> method returns
the object being initialized, to make the constructor trivial:
</p>

<a href="xxx">go/src/pkg/container/vector/vector.go</a>:
<pre>
func New(len int) *Vector {
	return new(Vector).Init(len)
}
</pre>

<h3 id="zero-value">Make the zero value meaningful</h3>

<p>
In Go, newly allocated memory and newly declared variables are zeroed.
If a type is intended to be allocated without using a constructor
(for example, as part of a larger struct or declared as a local variable),
define the meaning of the zero value and arrange for that meaning
to be useful.
</p>

<p>
For example, <code>sync.Mutex</code> does not
have an explicit constructor or <code>Init</code> method.
Instead, the zero value for a <code>sync.Mutex</code>
is defined to be an unlocked mutex.
</p>

<h2 id="interfaces">Interfaces</h2>

<h3 id="accept-interface-values">Accept interface values</h3>

buffered i/o takes a Reader, not an os.File.  XXX

<h3 id="return-interface-values">Return interface values</h3>

<p>
If a type exists only to implement an interface
and has no exported methods beyond that interface,
there is no need to publish the type itself.
Instead, write a constructor that returns an interface value.
</p>

<p>
For example, both <code>crc32.NewIEEE()</code> and <code>adler32.New()</code>
return type <code>hash.Hash32</code>.
Substituting the CRC-32 algorithm for Adler-32 in a Go program
requires only changing the constructor call:
the rest of the code cannot distinguish the two algorithms.
</p>

<h3 id="">Use interface adapters to expand an implementation</h3>

XXX

<h3 id="">Use anonymous fields to incorporate an implementation</h3>

XXX

<h2>Data-Driven Programming</h2>

<p>
tables
</p>

<p>
XXX struct tags for marshalling.
template
eventually datafmt
</p>

<h2>Concurrency</h2>

<h3 id="share-memory">Share memory by communicating</h3>

<p>
Do not communicate by sharing memory;
instead, share memory by communicating.
</p>

<p>
XXX, more here.
</p>


<h2>Testing</h2>

<h3 id="no-abort">Run tests to completion</h3>

<p>
Tests should not stop early just because one case has misbehaved.
If at all possible, let tests continue, in order to characterize the
problem in more detail.
For example, it is more useful for a test to report that <code>isPrime</code>
gives the wrong answer for 2, 3, 5, and 7 (or for 2, 4, 8, and 16) than to report
that <code>isPrime</code> gives the wrong answer for 2 and therefore
no more tests were run.
</p>

<h3 id="good-errors">Print useful errors when tests fail</h3>

<p>
If a test fails, print a concise message explaining the context,
what happened, and what was expected.
Many testing environments encourage causing the
program to crash, but stack traces and core dumps
have low signal to noise ratios and require reconstructing
the situation from scratch.
The programmer who triggers the test failure may be someone
editing the code months later or even someone editing a different
package on which the code depends.
Time invested writing a good error message now pays off when
the test breaks later.
</p>

<h3 id="data-driven-tests">Use data-driven tests</h3>

<p>
Many tests reduce to running the same code multiple times,
with different input and expected output.
Instead of using cut and paste to write this code,
create a table of test cases and write a single test that
iterates over the table.
Once the table is written, you might find that it
serves well as input to multiple tests.  For example,
a single table of encoded/decoded pairs can be
used by both <code>TestEncoder</code> and <code>TestDecoder</code>.
</p>

<p>
This data-driven style dominates in the Go package tests.
<br>
<!-- search for for.*range here -->
</p>

<h3 id="reflect.DeepEqual">Use reflect.DeepEqual to compare complex values</h3>

<p>
The <code>reflect.DeepEqual</code> function tests
whether two complex data structures have equal values.
If a function returns a complex data structure, 
<code>reflect.DeepEqual</code> combined with table-driven testing
makes it easy to check that the return value is 
exactly as expected.
</p>

</div>
</body>
</html>