This question already has answers here:
Can I type assert a slice of interface values?
(2 answers)
Closed 3 years ago.
I have a type that implements the stringer interface
// RowID stores the ID of a single row in a table
type RowID []string
// String implements Stringer interface for RowID
func (r RowID) String() string {
return fmt.Sprintf("[%s]", strings.Join(r, ", "))
}
And I have a function that I want to pass a slice of this type (or any other type that implements the Stringer interface) to.
// PrintChanges ...
func PrintChanges(ids []fmt.Stringer) {
for _, id := range ids {
fmt.Println(id)
}
}
However, The go compiler gives me an error:
cannot use rowIDs (type []RowID) as type []fmt.Stringer in argument to PrintChanges
I can pass a RowID to a func that accepts a single fmt.Stringer
func PrintChange(id fmt.Stringer) {
fmt.Println(id)
}
...
PrintChange(RowID{"1", "24"})
But for some reason I am not able to pass a slice of RowID to a func that accepts a slice of fmt.Stringer. What am I missing?
Go Playground
Keep it simple
It is considered okay by professional Go programmers to repeat functions like this for every type, or to have a for loop over every slice you want to print. This is because Go aims to be as easy to read as possible, i.e. a person who reads a chunk of code for the first time should not be asking questions like "which function overload will this function call go to" (common pitfall in C++, Go does not have function overloads). So you can just write in main():
Playground: https://ideone.com/IL3rGR
for _, id := range rowIDs { fmt.Println(id) }
Simple and concise.
Note that fmt.Println(id) does not call your String() function
This is because the fmt library uses the reflect library and hardcodes behavior for the string type, which you are trying to replace. RowID instances are also string instances, the library always prefers string over its type aliases. I would say it is a bug in the library:
Library source: https://golang.org/src/fmt/print.go#L649
// Some types can be done without reflection.
switch f := arg.(type) {
...
case string:
p.fmtString(f, verb)
If you really want to
You can use a function that takes an interface{} and makes a runtime reflect type cast to a slice of Stringers. Note that this means you will not see type mismatches during compilation, only in runtime:
Playground: https://ideone.com/vlrBP9
func castToStringerSlice(iface interface{}) ([]fmt.Stringer, bool /* ok */) {
if reflect.TypeOf(iface).Kind() != reflect.Slice {
return nil, false
}
v := reflect.ValueOf(iface)
stringers := make([]fmt.Stringer, v.Len())
for i := 0; i < v.Len(); i++ {
stringers[i] = v.Index(i)
}
return stringers, true
}
func PrintChanges(iface_ids interface{}) {
ids, ok := castToStringerSlice(iface_ids)
if !ok {
log.Fatal(errors.New("the argument to PrintChanges must be a slice of Stringers"))
}
for _, id := range ids {
fmt.Println(id)
}
}
Resources:
Go documentation: Why does Go not support overloading of methods and operators?
Go documentation: Why does Go not have generic types?
Stack Overflow: Express function that takes any slice
Stack Overflow: Range over interface{} which stores a slice
Go documentation: Package reflect
Related
This question already has an answer here:
Golang static identifier resolution
(1 answer)
Closed 4 months ago.
How can I determine the type of a variable through static analysis?
Suppose I have the following code:
func doSomething(x interface{}) {}
func main() {
p := Person()
doSomething(p)
}
And I want to analyze doSomething(person), is it possible to get the type of Person through static analysis?
What if there were multiple levels of assignment?
p1 := Person()
p2 := p1
doSomething(p2)
or
parent := Parent()
p := Parent.Child() // type Person
doSomething(p)
The use case is that I have a generic function that is commonly used throughout the (very large) codebase, and would like to introduce a new type safe version of this function. To do this, I hope to automatically determine the "type" of the function and refactor it accordingly:
// old
DB.InsertRow(person)
// new
Person.InsertRow(person)
Finding the type of an expression through static analysis is non-trivial, and sometimes not possible, for details see Golang static identifier resolution.
The use case is that I have a generic function that is commonly used throughout the (very large) codebase, and would like to introduce a new type safe version of this function. To do this, I hope to automatically determine the "type" of the function and refactor it accordingly:
// old
DB.InsertRow(person)
// new
Person.InsertRow(person)
Just for refactoring purposes, I don't think it is worth the hassle to implement it.
What you may do is change the signature of DB.InsertRow() temporarily to accept only a specific type such as int or your custom type you're sure is not used anywhere (e.g. type tempFoo struct{}).
To what end? Doing so, the compiler will do the hard work for you. You will see error messages showing exactly the types your codebase is trying to pass to DB.InsertRow(), so I'd say mission accomplished.
For example this code compiles:
func doSomething(x interface{}) {}
func main() {
doSomething(image.Pt(1, 2))
doSomething("abc")
doSomething(image.Rect) // image.Rect is a function which we don't call,
// so we're passing a value of a function type here
}
If we change doSomething():
func doSomething(x int) {}
We get the types we're seeking for from the compiler:
./prog.go:10:14: cannot use image.Pt(1, 2) (value of type image.Point) as type int in argument to doSomething
./prog.go:11:14: cannot use "abc" (untyped string constant) as int value in argument to doSomething
./prog.go:12:14: cannot use image.Rect (value of type func(x0 int, y0 int, x1 int, y1 int) image.Rectangle) as type int in argument to doSomething
Using the advice from Golang static identifier resolution to use golang.org/x/tools/go/types, I found that this was pretty straight forward to do with the golang.org/x/tools/go/analysis package, which has the types info available alongside the parsed ast.
This was my solution:
package rewriter
import (
"go/ast"
"golang.org/x/tools/go/analysis"
"golang.org/x/tools/go/analysis/passes/inspect"
"golang.org/x/tools/go/ast/inspector"
)
func run(pass *analysis.Pass) (interface{}, error) {
inspect := pass.ResultOf[inspect.Analyzer].(*inspector.Inspector)
nodeFilter := []ast.Node{
(*ast.CallExpr)(nil),
}
inspect.Nodes(nodeFilter, func(node ast.Node, push bool) bool {
callExpr, ok := node.(*ast.CallExpr)
if !ok {
return true
}
funcExpr, ok := callExpr.Fun.(*ast.SelectorExpr)
if !ok {
return true
}
// check method name
if funcExpr.Sel.Name != "doSomething" {
return true
}
for _, arg := range callExpr.Args {
// lookup type of the arg
argType := pass.TypesInfo.Types[arg].Type
if argType.String() == "*rewriter.Person" {
// do whatever you want here
}
}
return false
})
return nil, nil
}
One can augment this to look at the receiver of the method and add refactoring logic as needed (using analysis.Diagnostic).
I want to use Scan() in package sql, but the number of columns, and hence the number of arguments, will change at runtime. This is the signature of Scan():
func (rs *Rows) Scan(dest ...interface{}) error
According to the documentation, *interface{} is one of the types accepted by Scan(). So I want to create a slice of []*interface{} and that expand as arguments.
This is what I thought would work:
func query(database *sql.DB) {
rows, _ := database.Query("select * from testTable")
for rows.Next() {
data := make([]*interface{}, 2)
err := rows.Scan(data...) // Compilation error
fmt.Printf("%v%v\n", *data[0], *data[1])
if err != nil {
fmt.Println(err.Error())
}
}
}
Compilation fails with cannot use data (type []*interface {}) as type []interface {} in argument to rows.Scan. I thought that data... would expand to &data[0], &data[1], but apparently not. I don't understand the error message. *interface{} is compatible with interface{}, so why can't I expand the slice of pointers to interface types?
This works:
func query(database *sql.DB) {
rows, _ := database.Query("select * from testTable")
for rows.Next() {
data := make([]*interface{}, 2)
err := rows.Scan(&data[0], &data[1]) // Only changed this line
fmt.Printf("%v%v\n", *data[0], *data[1]) // Outputs "[48][116 101 120 116]"
if err != nil {
fmt.Println(err.Error())
}
}
}
I can't use this however, because the number of columns is unknown at compile time. How can I write this code so that I can pass a variable number of *interface{} to rows.Scan()?
First, you must not use []*interface{} slice of pointers to interface rather than []interface{} where the interfaces are pointers. []*interface{} is different from []interface{}. Just create a slice of interfaces where each element is a pointer to a concrete type.
Here is a snippet how you would do this.
var x int
var s string
data := []interface{}{&x, &s}
rows.Scan(data...)
Note on the use of the ... spread operator.
Here are some related questions that will explain a bit more:
golang: slice of struct != slice of interface it implements?
Cannot convert []string to []interface {}
If you really want to pass a []*interface{} (perhaps you don't know the concrete types of the output) you must first wrap each *interface{} in a interface{}:
values := make([]interface{}, columnsCount)
for i := range values {
values[i] = new(interface{})
}
Individual values passed into a ...interface{} parameter are automatically wrapped in a interface{}, but just like []int... won't satisfy ...interface{}, neither will []*interface{}....
What cast / assertion need I do in Go in order to pass to a function expecting a generic function like func(interface{}) interface{}, a more specific function like func(int) int instead?
For example, in code like this, fooA can be passed to MakeExclamer, but not fooB:
func MakeExclamer(foo func (interface{}) interface{}, n int) func () {
return func() {
fmt.Printf("%v!!!", foo(n))
}
}
func fooA(x interface{}) interface{} {
return x.(int)*2
}
func fooB(x int) int {
return x * 10
}
func main() {
exclamerA := MakeExclamer(fooA, 12)
exclamerA()
exclamerB := MakeExclamer(fooB, 66)
// >> cannot use fooB (type func(int) int) as type func(interface {}) interface {} in argument to MakeExclamer
exclamerB()
}
(Go Playground link: https://play.golang.org/p/xGzfco0IAG)
I'm not interested much in alternative code structure patterns, since this is how I want it to work: a specific function should be passed to a general function transformer (accepting function of type Any -> Any) that will return another general function (Any -> Any). This may not be idiomatic in Go, but it is the pattern that I want my code to follow.
To use type assertions, every possible type must be enumerated in MakeExclamer:
func MakeExclamer(fn interface{}, arg interface{}) func() {
switch fn := fn.(type) {
case func(int) int:
return func() {
fmt.Printf("%v!!!\n", fn(arg.(int)))
}
case func(interface{}) interface{}:
return func() {
fmt.Printf("%v!!!\n", fn(arg))
}
default:
panic("not supported")
}
}
To accept a function of any type, the fn argument is declared as type interface{}. The code uses a type switch to handle the different function types.
playground example
Reflection can be used to write a more general function.
func MakeExclamer(fn interface{}, arg interface{}) func() {
fnr := reflect.ValueOf(fn)
argr := reflect.ValueOf(arg)
return func() {
resultr := fnr.Call([]reflect.Value{argr})
fmt.Printf("%v!!!\n", resultr[0].Interface())
}
}
playground example
First things first : When it comes to typing in Go, everything is theoretically possible. That's because even though the compiler does a lot of checks at compile-time, it is possible to change the runtime... at runtime. So-called runtime hacks, where you dynamically manipulate runtime structs that you're NOT supposed to handle.
Now, you have an interesting question, whose answer doesn't include the need to use the 'unsafe' package. However, the way I found of generalizing a function involves heavy reflection.
How to call a function (via reflection) ?
The documentation for the reflect package can be found here.
So, like all elements in Golang, functions have a Type. Without going through all fields, functions do take an array of arguments and produce an array of results. It is possible to investigate the Type of arguments and results through the In(int) and Out(int) method.
func investigate(fn interface{}) {
fnType := reflect.TypeOf(fn)
for idx := 0; idx < fnType.NumIn(); idx ++ {
fmt.Printf("Input arg %d has type %v\n", idx, fnType.In(idx))
}
for idx := 0; idx < fnType.NumOut(); idx ++ {
fmt.Printf("Output arg %d has type %v\n", idx, fnType.Out(idx))
}
}
We won't use this code. However, two important things are to be noted at this point :
The generic type under which a function can be passed around without caring about its type is interface{}. Something like "func(interface{}) interface{}" is not a generalization of a function, it is already a concrete type. Hence, "func(interface{}) interface{}" is not a generalization of "func(int) int", those are two different function types entirely. This is why you can't use type assertions/cast to convert from one function type to another.
A function can be represented as something that takes an input array and produces and output array.
Now, in order to call a function, you have to get not its Type, but its Value. Once you get its value, you can call it using an array of arguments, which must all be Values.
The prototype is:
func (v Value) Call(in []Value) []Value
Using this method, it is possible to call any function.
The code
So, the only thing you need is to convert whichever arguments array you have to an array of Values, then you will be able to call your function.
Here is your code:
package main
import (
"fmt"
"reflect"
)
func MakeExclamer(foo interface{}, n int) func() {
exclamer := generalize(foo, n)
return func() {
fmt.Printf("%v!!!\n", exclamer())
}
}
func fooA(x interface{}) interface{} {
return x.(int) * 2
}
func fooB(x int) int {
return x * 10
}
func generalize(implem interface{}, args ...interface{}) func() interface{} {
valIn := make([]reflect.Value, len(args), len(args))
fnVal := reflect.ValueOf(implem)
for idx, elt := range args {
valIn[idx] = reflect.ValueOf(elt)
}
ret := func() interface{} {
res := fnVal.Call(valIn)
// We assume the function produces exactly one result
return res[0].Interface()
}
return ret
}
func main() {
exclamerA := MakeExclamer(fooA, 12)
exclamerA()
exclamerB := MakeExclamer(fooB, 66)
exclamerB()
}
Playground
The important bit is the generalize function which makes the translation between your arguments and an array of Values, then returns a new function whith all parameters already filled.
Do not hesitate if you need any precision !
In my actual code, I'm parsing an XML document using encoding/xml, and I basically have a bunch of nested structures of the following form — all of which may occur multiple times, except the top-level statements element:
statements
statement
opcode
args
pre
post
I'm fairly new to Go, and I'm clearly misunderstanding how interface{} (the empty interface) works:
.\stmtgen.go:58: cannot use print_name (type func(Statement)) as type func(interface {}) in argument to performAction
.\stmtgen.go:58: cannot use slist (type []Statement) as type []interface {} in argument to performAction
Relevant example code:
package main
import "fmt"
// Actually a structure in my code, but this suffices for demonstration.
type Opcode int
// A Statement has a Name and multiple Opcodes may use this Name.
type Statement struct {
Name string
Opcodes []Opcode
}
// Print the statement name.
func print_name(stmt Statement) {
fmt.Println(stmt.Name)
}
// Perform an action on each item of a collection.
func performAction(action func(interface{}), v []interface{}) {
for i := range v {
action(v[i])
}
}
func main() {
slist := make([]Statement, 3)
slist[0] = Statement{"Statement 1"}
slist[1] = Statement{"Statement 2"}
slist[2] = Statement{"Statement 3"}
//ERROR HERE
performAction(print_name, slist)
}
Must I create functions to print the values for every single type?
An empty interface{} can contain any value and passed around as the type interface{}. when you need the value from it, you can perform a type assertion like this:
var anyValue interface{}
anyValue = "hello"
strValue := anyValue.(string)
If anyValue is not of the type being asserted then it will cause a panic
the type assertion can also be used to return a bool if the interface is of that type with a multiple return
strValue, ok := anyValue.(string)
if ok {
//anyValue contains a string!
}
If you dont know the type of the interface, you can use a switch to determine it's type like this:
switch val := anyValue.(type) {
case string:
// anyValue contains a string
// and val is a string
break
case int:
// anyValue contains an int
// and val is and int
break
default:
//unhandled interface type
}
Hopefully this makes the empty interface{} type clearer.
interfaces{...} which have methods declared in them are different, they can not have members (like structs can), only methods, and their underlying type must implement all the methods declared in the interface. You could have an interface actionPerformer (interface names should have the suffix "er" as they are doing something)
type actionPerformer interface {
action(interface{})
}
A type that implements all the methods in an interface can be cast to that interface type, then if you call one of those methods on the interface, it will run the method on the underlying type.
For example, if the Statement struct implements the action(interface{}) method, the Statement struct can be cast to an actionPerformer type and if the action(interface{}) function is called on the actionPerformer, the action function on the Statement struct is run. So you could have multiple types that all have the action(interface{}) method and they can all be cast to an actionPerformer which you can call the action function on.
func (code Opcode) action(arg interface{}) {
fmt.Println(arg.(int) + int(code))
}
func (stmt Statement) action(arg interface{}) {
fmt.Println(arg.(string), stmt.Name)
}
stmt := Statement{"Statement 1", nil}
stmtActionPerformer := actionPerformer(stmt)
opcode := Opcode(5)
opcodeActionPerformer := actionPerformer(opcode)
stmtActionPerformer.action("hello") // will print "hello "+whatever the statements name is
opcodeActionPerformer.action(2) //will print be 7
Type assertions can still be used on these types of interface e.g.
stmt := stmtActionPerformer.(Statement)
fmt.Println(stmt.Name)
This is a contrived example, but with this in mind, you might want to write your code using interfaces like this.
Remember casting between interfaces is costly, so should be done sparingly, however they are a powerful tool when used correctly.
For your example, a simple printNames function would be much more efficient than all that interface casting (note that in golang, names should be in the CamelCase format, not using underscores)
func printNames(stmts []Statement) {
for _, stmt := range stmts {
fmt.Println(stmt.Name)
}
}
It might also be useful to have a type StatementList and add methods to it:
type StatementList []Statement
func (list StatementList) printNames() {
for _, stmt := range list {
fmt.Println(stmt.Name)
}
}
Getting the hang of this stuff make golang a lot more fun, hope this helps :)
You have to declare the parameters of performAction exactly same like the arguments type.
func performAction(action func(Statement), v []Statement) {
for i := range v {
action(v[i])
}
}
Or you could use interface{} on all parameters instead. Then cast it according to the needs.
func performAction(action interface{}, v interface{}) {
for _, each := range v.([]Statement) {
action.(func(Statement))(each)
}
}
data with type []Statement cannot be assigned to []interface{}
also for type func(Statement) cannot be assigned to func(interface{})
Use interface{}, then cast it to the original type.
this works for me:
package main
import (
"fmt"
)
// Actually a structure in my code, but this suffices for demonstration.
type Opcode int
// A Statement has a Name and multiple Opcodes may use this Name.
type Statement struct {
Name string
Opcodes []Opcode
}
// Print the statement name.
func print_name(stmt interface{}) {
if s, ok := stmt.(Statement); !ok {
panic("typ err")
} else {
fmt.Println(s.Name)
}
}
// Perform an action on each item of a collection.
func performAction(action func(interface{}), v []interface{}) {
for i := range v {
action(v[i])
}
}
func main() {
slist := make([]interface{}, 3)
slist[0] = Statement{"Statement 1", nil}
slist[1] = Statement{"Statement 2", nil}
slist[2] = Statement{"Statement 3", nil}
performAction(print_name, slist)
/*output:
Statement 1
Statement 2
Statement 3
*/
}
I have custom types Int64Array, Channel and ChannelList like:
type Int64Array []int64
func (ia *Int64Array) Scan(src interface{}) error {
rawArray := string(src.([]byte))
if rawArray == "{}" {
*ia = []int64{}
} else {
matches := pgArrayPat.FindStringSubmatch(rawArray)
if len(matches) > 1 {
for _, item := range strings.Split(matches[1], ",") {
i, _ := strconv.ParseInt(item, 10, 64)
*ia = append(*ia, i)
}
}
}
return nil
}
func (ia Int64Array) Value() (driver.Value, error) {
var items []string
for _, item := range ia {
items = append(items, strconv.FormatInt(int64(item), 10))
}
return fmt.Sprintf("{%s}", strings.Join(items, ",")), nil
}
type Channel int64
type ChannelList []Channel
How can I embed Int64Array to ChannelList such that I can call Scan and Value methods on it? I tried the following:
type ChannelList []Channel {
Int64Array
}
but I'm getting syntax error. What's important is to make sure ChannelList items are of type Channel, if this isn't possible via embedding I might just create stand-alone functions to be called by both ChannelList and Int64Array.
An anonymous (or embedded field) is found in a struct (see struct type), not in a type alias (or "type declaration").
You cannot embed a type declaration within another type declaration.
Plus, as illustrated by the answers to "Go: using a pointer to array", you shouldn't be using pointers to slice, use directly the slice themselves (passed by value).
Wessie kindly points out in the comments that (ia *Int64Array) Scan() uses pointer to a slice in order to mutate the underlying array referenced by said slice.
I would prefer returning another slice instead of mutating the existing one.
That being said, the Golang Code Review does mention:
If the receiver is a struct, array or slice and any of its elements is a pointer to something that might be mutating, prefer a pointer receiver, as it will make the intention more clear to the reader.