Getting started with Go pointers

user: anonymous

{
  "path": "/blog/getting-started-with-go-part-2-pointers",
  "attrs": {
    "title": "Getting started with Go pointers",
    "splash": {
      "image": "/images/ladybug.jpg"
    },
    "date": "2021-04-18",
    "layout": "BlogPostLayout",
    "excerpt": "Part 2 of my experience as a new user of Go."
  },
  "md": "# Getting started with Go pointers\n\nThis is part 2 of my experience as a new user of Go, focusing on the quirks and gotchas of pointers. For installation, testing, and packages, see [Getting started with Go](getting-started-with-go).\n\nIf you'd like to follow along, and try out out the code in this article, all you need is the Go [playground](https://play.golang.org/p/-UiUJFrloVT) to run the examples.\n\n## Pointers\n\nThe [shortscale](https://github.com/jldec/shortscale-go/blob/main/shortscale.go) package which I covered last time, uses a string [Builder](https://pkg.go.dev/strings#Builder). Here is the example from the Builder docs.\n\n```go\npackage main\n\nimport (\n\t\"fmt\"\n\t\"strings\"\n)\n\nfunc main() {\n\tvar b strings.Builder\n\tfor i := 3; i >= 1; i-- {\n\t\tfmt.Fprintf(&b, \"%d...\", i)\n\t}\n\tb.WriteString(\"ignition\")\n\tfmt.Println(b.String())\n}\n```\n\nNotice that `var b` is an instance of the Builder. When you run the code, it will output: _3...2...1...ignition_.\n\n## Pointer receiver methods and interfaces\n\nThe first argument to fmt.Fprintf is `&b`, a [pointer](https://tour.golang.org/moretypes/1) to b. This is necessary, because [fmt.Fprintf](https://pkg.go.dev/fmt#Fprintf) expects an [io.Writer](https://pkg.go.dev/io#Writer) interface.\n\n```go\ntype Writer interface {\n\tWrite(p []byte) (n int, err error)\n}\n```\n\nThe [Builder.Write](https://pkg.go.dev/strings#Builder.Write) method matches the io.Writer interface. Notice the pointer syntax in the method receiver after the `func` keyword.\n\n```go\nfunc (b *Builder) Write(p []byte) (int, error)\n```\n\nI was tempted to replace `Fprintf(&b, ...)` with `Fprintf(b, ...)`, to make it more consistent with the `b.WriteString()` and `b.String()` further down, but doing this causes the compiler to complain:\n\n_\"cannot use b (type strings.Builder) as type io.Writer in argument to fmt.Fprintf:\nstrings.Builder does not implement io.Writer (**Write method has pointer receiver**)\"_\n\n## Value vs. pointer function arguments\n\nWhat if, instead of depending on the Writer interface, we called our own `write()` function?\n\n```go\nfunc main() {\n\tvar b strings.Builder\n\tfor i := 3; i >= 1; i-- {\n\t\twrite(b, fmt.Sprintf(\"%d...\", i))\n\t}\n\tb.WriteString(\"ignition\")\n\tfmt.Println(b.String())\n}\n\nfunc write(b strings.Builder, s string) {\n\tb.WriteString(s)\n}\n```\n\nRunning the code above in the [example sandbox](https://pkg.go.dev/strings#Builder) outputs just the word _ignition_. \n\nThe 3 calls to `write(b)` do not modify the builder declared at the top.\n\nThis makes sense, because passing a struct to a function [copies](https://tour.golang.org/methods/4) the struct value.\n\nTo fix this, we have to use a pointer to pass the struct by reference, and we have to invoke the function with `write(&b, ...)`. This works, but it doesn't make the code any more consistent.\n\n```go\nfunc main() {\n\tvar b strings.Builder\n\tfor i := 3; i >= 1; i-- {\n\t\twrite(&b, fmt.Sprintf(\"%d...\", i))\n\t}\n\tb.WriteString(\"ignition\")\n\tfmt.Println(b.String())\n}\n\nfunc write(b *strings.Builder, s string) {\n\tb.WriteString(s)\n}\n```\n\n## Why do the method calls work?\n\nWhy are we allowed to use `b` instead of `&b` in front of [b.WriteString](https://pkg.go.dev/strings#Builder.WriteString) and [b.String](https://pkg.go.dev/strings#Builder.String)? This is explained in [the tour](https://tour.golang.org/methods/6) as well.\n\n_\"...even though v is a value and not a pointer, the method with the pointer receiver is called automatically. That is, as a convenience, Go interprets the statement v.Scale(5) as (&v).Scale(5) since the Scale method has a pointer receiver.\"_\n\n## Start with a pointer\n\nIf all this mixing of values and pointers feels inconsistent, why not start with a pointer from the beginning?\n\nThe following code will compile just fine, but can you tell what's wrong with it?\n\n```go\nfunc main() {\n\tvar b *strings.Builder\n\tfor i := 3; i >= 1; i-- {\n\t\tfmt.Fprintf(b, \"%d...\", i)\n\t}\n\tb.WriteString(\"ignition\")\n\tfmt.Println(b.String())\n}\n```\nThe declaration above results in a nil pointer panic at run time, because b is uninitialized.\n\n> _panic: runtime error: invalid memory address or nil pointer dereference  \n[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x4991c7]_\n\n## Create the Builder with new()\n\nHere is one way to initialize a pointer so that it references a new Builder.\n\n```go\nfunc main() {\n\tb := new(strings.Builder)\n\tfor i := 3; i >= 1; i-- {\n\t\tfmt.Fprintf(b, \"%d...\", i)\n\t}\n\tb.WriteString(\"ignition\")\n\tfmt.Println(b.String())\n}\n```\n\n`new(strings.Builder)` returns a pointer to a freshly allocated Builder, which we can use for both functions and pointer receiver methods. This is the pattern which I now use in [shortscale-go](https://github.com/jldec/shortscale-go/blob/2485be23ef48660d8913b2ac884030220dc82d74/shortscale.go#L17-L24).\n\n\nAn alternative, which does the same thing, is the more explicit [struct literal](https://tour.golang.org/moretypes/5) shown below.\n\n```go\nfunc main() {\n\tb := &strings.Builder{}\n\t...\n```\n\n> There's no avoiding pointers in Go.  \n> Learn the quirks and the gotchas today.  \n> ✨ Keep learning! ✨\n\n_To leave a comment  \nplease visit [dev.to/jldec](https://dev.to/jldec/getting-started-with-go-part-2-pointers-4a47)_\n\n",
  "html": "<h1>Getting started with Go pointers</h1>\n<p>This is part 2 of my experience as a new user of Go, focusing on the quirks and gotchas of pointers. For installation, testing, and packages, see <a href=\"getting-started-with-go\">Getting started with Go</a>.</p>\n<p>If you'd like to follow along, and try out out the code in this article, all you need is the Go <a href=\"https://play.golang.org/p/-UiUJFrloVT\">playground</a> to run the examples.</p>\n<h2>Pointers</h2>\n<p>The <a href=\"https://github.com/jldec/shortscale-go/blob/main/shortscale.go\">shortscale</a> package which I covered last time, uses a string <a href=\"https://pkg.go.dev/strings#Builder\">Builder</a>. Here is the example from the Builder docs.</p>\n<pre><code class=\"language-go\">package main\n\nimport (\n\t&quot;fmt&quot;\n\t&quot;strings&quot;\n)\n\nfunc main() {\n\tvar b strings.Builder\n\tfor i := 3; i &gt;= 1; i-- {\n\t\tfmt.Fprintf(&amp;b, &quot;%d...&quot;, i)\n\t}\n\tb.WriteString(&quot;ignition&quot;)\n\tfmt.Println(b.String())\n}\n</code></pre>\n<p>Notice that <code>var b</code> is an instance of the Builder. When you run the code, it will output: <em>3...2...1...ignition</em>.</p>\n<h2>Pointer receiver methods and interfaces</h2>\n<p>The first argument to fmt.Fprintf is <code>&amp;b</code>, a <a href=\"https://tour.golang.org/moretypes/1\">pointer</a> to b. This is necessary, because <a href=\"https://pkg.go.dev/fmt#Fprintf\">fmt.Fprintf</a> expects an <a href=\"https://pkg.go.dev/io#Writer\">io.Writer</a> interface.</p>\n<pre><code class=\"language-go\">type Writer interface {\n\tWrite(p []byte) (n int, err error)\n}\n</code></pre>\n<p>The <a href=\"https://pkg.go.dev/strings#Builder.Write\">Builder.Write</a> method matches the io.Writer interface. Notice the pointer syntax in the method receiver after the <code>func</code> keyword.</p>\n<pre><code class=\"language-go\">func (b *Builder) Write(p []byte) (int, error)\n</code></pre>\n<p>I was tempted to replace <code>Fprintf(&amp;b, ...)</code> with <code>Fprintf(b, ...)</code>, to make it more consistent with the <code>b.WriteString()</code> and <code>b.String()</code> further down, but doing this causes the compiler to complain:</p>\n<p><em>&quot;cannot use b (type strings.Builder) as type io.Writer in argument to fmt.Fprintf:\nstrings.Builder does not implement io.Writer (<strong>Write method has pointer receiver</strong>)&quot;</em></p>\n<h2>Value vs. pointer function arguments</h2>\n<p>What if, instead of depending on the Writer interface, we called our own <code>write()</code> function?</p>\n<pre><code class=\"language-go\">func main() {\n\tvar b strings.Builder\n\tfor i := 3; i &gt;= 1; i-- {\n\t\twrite(b, fmt.Sprintf(&quot;%d...&quot;, i))\n\t}\n\tb.WriteString(&quot;ignition&quot;)\n\tfmt.Println(b.String())\n}\n\nfunc write(b strings.Builder, s string) {\n\tb.WriteString(s)\n}\n</code></pre>\n<p>Running the code above in the <a href=\"https://pkg.go.dev/strings#Builder\">example sandbox</a> outputs just the word <em>ignition</em>.</p>\n<p>The 3 calls to <code>write(b)</code> do not modify the builder declared at the top.</p>\n<p>This makes sense, because passing a struct to a function <a href=\"https://tour.golang.org/methods/4\">copies</a> the struct value.</p>\n<p>To fix this, we have to use a pointer to pass the struct by reference, and we have to invoke the function with <code>write(&amp;b, ...)</code>. This works, but it doesn't make the code any more consistent.</p>\n<pre><code class=\"language-go\">func main() {\n\tvar b strings.Builder\n\tfor i := 3; i &gt;= 1; i-- {\n\t\twrite(&amp;b, fmt.Sprintf(&quot;%d...&quot;, i))\n\t}\n\tb.WriteString(&quot;ignition&quot;)\n\tfmt.Println(b.String())\n}\n\nfunc write(b *strings.Builder, s string) {\n\tb.WriteString(s)\n}\n</code></pre>\n<h2>Why do the method calls work?</h2>\n<p>Why are we allowed to use <code>b</code> instead of <code>&amp;b</code> in front of <a href=\"https://pkg.go.dev/strings#Builder.WriteString\">b.WriteString</a> and <a href=\"https://pkg.go.dev/strings#Builder.String\">b.String</a>? This is explained in <a href=\"https://tour.golang.org/methods/6\">the tour</a> as well.</p>\n<p><em>&quot;...even though v is a value and not a pointer, the method with the pointer receiver is called automatically. That is, as a convenience, Go interprets the statement v.Scale(5) as (&amp;v).Scale(5) since the Scale method has a pointer receiver.&quot;</em></p>\n<h2>Start with a pointer</h2>\n<p>If all this mixing of values and pointers feels inconsistent, why not start with a pointer from the beginning?</p>\n<p>The following code will compile just fine, but can you tell what's wrong with it?</p>\n<pre><code class=\"language-go\">func main() {\n\tvar b *strings.Builder\n\tfor i := 3; i &gt;= 1; i-- {\n\t\tfmt.Fprintf(b, &quot;%d...&quot;, i)\n\t}\n\tb.WriteString(&quot;ignition&quot;)\n\tfmt.Println(b.String())\n}\n</code></pre>\n<p>The declaration above results in a nil pointer panic at run time, because b is uninitialized.</p>\n<blockquote>\n<p><em>panic: runtime error: invalid memory address or nil pointer dereference<br>\n[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x4991c7]</em></p>\n</blockquote>\n<h2>Create the Builder with new()</h2>\n<p>Here is one way to initialize a pointer so that it references a new Builder.</p>\n<pre><code class=\"language-go\">func main() {\n\tb := new(strings.Builder)\n\tfor i := 3; i &gt;= 1; i-- {\n\t\tfmt.Fprintf(b, &quot;%d...&quot;, i)\n\t}\n\tb.WriteString(&quot;ignition&quot;)\n\tfmt.Println(b.String())\n}\n</code></pre>\n<p><code>new(strings.Builder)</code> returns a pointer to a freshly allocated Builder, which we can use for both functions and pointer receiver methods. This is the pattern which I now use in <a href=\"https://github.com/jldec/shortscale-go/blob/2485be23ef48660d8913b2ac884030220dc82d74/shortscale.go#L17-L24\">shortscale-go</a>.</p>\n<p>An alternative, which does the same thing, is the more explicit <a href=\"https://tour.golang.org/moretypes/5\">struct literal</a> shown below.</p>\n<pre><code class=\"language-go\">func main() {\n\tb := &amp;strings.Builder{}\n\t...\n</code></pre>\n<blockquote>\n<p>There's no avoiding pointers in Go.<br>\nLearn the quirks and the gotchas today.<br>\n✨ Keep learning! ✨</p>\n</blockquote>\n<p><em>To leave a comment<br>\nplease visit <a href=\"https://dev.to/jldec/getting-started-with-go-part-2-pointers-4a47\">dev.to/jldec</a></em></p>\n"
}