I've always found the package.New() syntax in go rather awkward to work with.
The suggestion is that if a package holds only a single type, using package.New() to create an instance; if multiple types exist, using package.NewBlah().
http://golang.org/doc/effective_go.html#package-names
However, this approach falls down if you if you have an existing package with a New() api, adding a new external type to the package breaks the api, because you must now rename this NewFoo(). Now you have to go and change anything that uses New(), which is deeply irritating.
...and I'm just discontent with the aesthetic of writing this:
import "other"
import "bar"
import "foo"
o := other.New() // <-- Weird, what type am I getting? No idea.
x := bar.New()
y := foo.NewFoo() // <-- Awkward, makes constructor naming look inconsistent
z := foo.NewBar()
So, recently I've been using this pattern instead:
x := foo.Foo{}.New() // <-- Immediately obvious I'm getting a Foo
y := foo.Bar{}.New() // <-- Only an additional 3 characters on NewBar{}
o := other.Foo{}.New() // <-- Consistent across all packages, no breakage on update
Where the module is defined something like this:
package foo
type Foo struct {
x int
}
func (s Foo) New() *Foo {
// Normal init stuff here
return &s // <-- Edit: notice the single instance is returned
}
type Bar struct {
}
func (Bar) New() *Bar {
return &Bar{} // <-- Edit: Bad, results in double alloc. Not like this.
}
Godoc seems to work fine with it, and it seems more obvious and consistent to me, without additional verbosity.
So, question: Is there any tangible downside to this?
Yes, it has a downside. This approach may generate unnecessary garbage - depending on how good the optimization of a specific Go compiler implementation is.
It's not terribly idiomatic and may if done badly create excess garbage as you note. Essentially you are just creating an Init method for your object. I don't use a lot of constructors myself tending to prefer having valid zero values for my objects and only using a constructor if that doesn't hold true.
In your case I think I'd just stop calling the method new and instead call it Init or Setup to better reflect what it's doing. That would avoid giving people the wrong idea about what it's doing.
Edit:
I should have been more detailed here. Calling the method Init or Setup and then using it on a Zero Value would better reflect what is going on to the consumer. eg
f := &foo{}
f.Init()
This avoids the excess garbage and gives you an initializer method as you describe.
Related
Go Doc is indenting/making a group without me knowingly telling it to.
Here is a screenshot from my browser showing the problem. The four Parse functions should not be indented:
What is causing this behaviour?
I've tried searching for indents/subsections/grouping in Go Docs but I've not found anything beside feature requests. I'm sure the answer to my question is in the documentation somewhere but I can't find it as I don't have the correct vocabulary.
I'm relatively new to Go and Go Doc, so I'm assuming the answer is something simple which I've overlooked.
Here is an extract of my code. Please let me know if I need to share more code.
status.go
package synop
// Status codes returned when parsing Blocks.
type Status int64
const (
VALID Status = iota
NO_DATA
// code omitted
)
// Other functions omitted
cloudwind_block.go
package synop
import (
"strings"
)
/*
Extract cloud cover from the cloud-wind block, Cxxxx.
Cloud, C, is the first digit of the block. Cloud over is given in [okta]:
*/
func ParseCloud(block string) (okta int, s Status) {
slice := [2]int{0, 1}
return parseIfValid(block, slice, str2int)
}
/*
Extract wind direction from from the cloud-wind block, xDDxxx.
Direction, DD, are the second and third digits.
*/
func ParseDir(block string) (dir string, s Status) {
slice := [2]int{1, 3}
return parseIfValid(block, slice, getDir)
}
// Other functions omitted
I have another file, blocks.go, which has almost the same structure as status.go and it does not cause this behaviour. I also don't know if the problem is caused by the preceding type Status or something in the cloudwind_block.go file.
I'm using // for single-line documentation and /* */ for multi line. I've tried making this consistent on the off chance and, as expected, it had no effect.
The reason for the grouping and indentation is that those functions are considered "constructors" of the type under which they are grouped/indented.
https://go.dev/doc/comment#func (if you scroll down a bit, you'll see this):
This example also shows that top-level functions returning a type T or pointer *T, perhaps with an additional error result, are shown alongside the type T and its methods, under the assumption that they are T’s constructors.
// reflect/value.go
func ValueOf(i interface{}) Value {
if i == nil {
return Value{}
}
// TODO: Maybe allow contents of a Value to live on the stack.
// For now we make the contents always escape to the heap. It
// makes life easier in a few places (see chanrecv/mapassign
// comment below).
escapes(i)
The code above is the source code of Value.go in golang, and the comment above the escapes(i) shows that each time we call the ValueOf function, the i will escape to the heap, that's why? Namely, how to explain the It makes life easier in a few places?
I am still learning go, so I can't describe more, that's why a community wiki answer. But here's what excerpted note says (note above the chanrecv function):
Note: some of the noescape annotations below are technically a lie,
but safe in the context of this package. Functions like chansend
and mapassign don't escape the referent, but may escape anything
the referent points to (they do shallow copies of the referent).
It is safe in this package because the referent may only point
to something a Value may point to, and that is always in the heap
(due to the escapes() call in ValueOf).
Also see:
// Dummy annotation marking that the value x escapes,
// for use in cases where the reflect code is so clever that
// the compiler cannot follow.
func escapes(x interface{}) {
if dummy.b {
dummy.x = x
}
}
var dummy struct {
b bool
x interface{}
}
I hope, this will be helpful.
var response Response
switch wrapper.Domain {
case "":
response = new(TypeA)
case "TypeB":
response = new(TypeB)
case "TypeC":
response = new(TypeC)
case "TypeD":
response = new(TypeD)
}
_ = decoder.Decode(response)
As shown in the code snippet, I got enough information from the Domain filed of wrapper to determine the type of response, and for each type, the following operations are performed:
create a new instance of that type using new
use the decoder to decode the byte slice to the instance created in step 1
I am wondering if there is a way to make the first step more generic and get rid of the switch statement.
A bit about your code
As per discussion in comments, I would like to share some experience.
I do not see nothing bad in your solution, but there are few options to improve it, depends what you want to do.
Your code looks like classic Factory. The Factory is a pattern, that create object of a single family, based on some input parameters.
In Golang this is commonly used in simpler way as a Factory Method, sometimes called Factory function.
Example:
type Vehicle interface {};
type Car struct {}
func NewCar() Vehicle {
return &Car{}
}
But you can easily expand it to do something like you:
package main
import (
"fmt"
"strings"
)
type Vehicle interface {}
type Car struct {}
type Bike struct {}
type Motorbike struct {}
// NewDrivingLicenseCar returns a car for a user, to perform
// the driving license exam.
func NewDrivingLicenseCar(drivingLicense string) (Vehicle, error) {
switch strings.ToLower(drivingLicense) {
case "car":
return &Car{}, nil
case "motorbike":
return &Motorbike{}, nil
case "bike":
return &Bike{}, nil
default:
return nil, fmt.Errorf("Sorry, We are not allowed to make exam for your type of car: \"%s\"", drivingLicense)
}
}
func main() {
fmt.Println(NewDrivingLicenseCar("Car"))
fmt.Println(NewDrivingLicenseCar("Tank"))
}
Above code produces output:
&{} <nil>
<nil> Sorry, We are not allowed to make exam for your type of car: "Tank"
So probably you can improve your code by:
Closing into a single function, that takes a string and produces the Response object
Adding some validation and the error handling
Giving it some reasonable name.
There are few related patterns to the Factory, which can replace this pattern:
Chain of responsibility
Dispatcher
Visitor
Dependency injection
Reflection?
There is also comment from #icza about Reflection. I agree with him, this is used commonly, and We cannot avoid the reflection in our code, because sometimes things are so dynamic.
But in your scenario it is bad solution because:
You lose compile-time type checking
You have to modify code when you are adding new type, so why not to add new line in this Factory function?
You make your code slower(see references), it adds 50%-100% lose of performance.
You make your code so unreadable and complex
You have to add a much more error handling to cover not trivial errors from reflection.
Of course, you can add a lot of tests to cover a huge number of scenarios. You can support TypeA, TypeB, TypeC in your code and you can cover it with tests, but in production code sometime you can pass TypeXYZ and you will get runtime error if you do not catch it.
Conclusion
There is nothing bad with your switch/case scenario, probably this is the most readable and the easiest way to do what you want to do.
Reference
Factory method: https://www.sohamkamani.com/golang/2018-06-20-golang-factory-patterns/
Classic book about patterns in programming: Design Patterns: Elements of Reusable Object-Oriented Software, Erich Gamma and his band of four, ISBN: 978-0201633610
Reflection benchmarks: https://gist.github.com/crast/61779d00db7bfaa894c70d7693cee505
I get cannot use map[string]MyType literal (type map[string]MyType) as type map[string]IterableWithID in argument to MapToList with the code below, how do I pass in a concrete map type to method that expects a interface type?
https://play.golang.org/p/G7VzMwrRRw
Go's interface convention doesn't quite work the same way as in, say, Java (and the designers apparently didn't like the idea of getters and setters very much :-/ ). So you've got two core problems:
A map[string]Foo is not the same as a map[string]Bar, even if Bar implements Foo, so you have to break it out a bit (use make() beforehand, then assign in a single assignment).
Interface methods are called by value with no pointers, so you really need to do foo = foo.Method(bar) in your callers or get really pointer-happy to implement something like this.
What you can do to more-or-less simulate what you want:
type IterableWithID interface {
SetID(id string) IterableWithID // use as foo = foo.SetID(bar)
}
func (t MyType) SetID(id string) IterableWithID {
t.ID = id
return t
}
...and to deal with the typing problem
t := make(map[string]IterableWithID)
t["foo"] = MyType{}
MapToList(t) // This is a map[string]IterableWithID, so compiler's happy.
...and finally...
value = value.SetID(key) // We set back the copy of the value we mutated
The final value= deals with the fact that the method gets a fresh copy of the value object, so the original would be untouched by your method (the change would simply vanish).
Updated code on the Go Playground
...but it's not particularly idiomatic Go--they really want you to just reference struct members rather than use Java-style mutators in interfaces (though TBH I'm not so keen on that little detail--mutators are supes handy to do validation).
You can't do what you want to do because the two map types are different. It doesn't matter that the element type of one is a type that implements the interface which is the element type of the other. The map type that you pass into the function has to be map[string]IterableWithID. You could create a map of that type, assign values of type MyType to the map, and pass that to the function.
See https://play.golang.org/p/NfsTlunHkW
Also, you probably don't want to be returning a pointer to a slice in MapToList. Just return the slice itself. A slice contains a reference to the underlying array.
If you look at the image package here http://golang.org/src/pkg/image/image.go you can see that the implementation of Opaque() for every image does the same thing, differing only in the pixel-specific logic.
Is there a reason for this? Would any general solution be less efficient? Is it just an oversight? Is there some limitation (I cannot see one) to the type system that would make a polymorphic [was: generic] approach difficult?
[edit] The kind of solution I was thinking of (which does not need generics in the Java sense) would be like:
type ColorPredicate func(c image.Color) bool;
func AllPixels (p *image.Image, q ColorPredicate) bool {
var r = p.Bounds()
if r.Empty() {
return true
}
for y := r.Min.Y; y < r.Max.Y; y++ {
for x := r.Min.X; x < r.Max.X; x++ {
if ! q(p.At(x,y)) {
return false
}
}
}
return true
}
but I am having trouble getting that to compile (still very new to Go - it will compile with an image, but not with an image pointer!).
Is that too hard to optimise? (you would need to have function inlining, but then wouldn't any type checking be pulled out of the loop?). Also, I now realise I shouldn't have used the word "generic" earlier - I meant it only in a generic (ha) way.
There is a limitation to the type system which prevents a general solution (or at least makes it very inefficient).
For example, the bodies of RGBA.Opaque and NRGBA.Opaque are identical, so you'd think that they could be factored out into a third function with a signature something like this:
func opaque(pix []Color, stride int, rect Rectangle) bool
You'd like to call that function this way:
func (p *RGBA) Opaque() bool {
return opaque([]Color(p.Pix), p.Stride, p.Rect)
}
But you can't. p.Pix can't be converted to []Color because those types have different in-memory representations and the spec forbids it. We could allocate a new slice, convert each individual element of p.Pix, and pass that, but that would be very inefficient.
Observe that RGBAColor and NRGBAColor have the exact same structure. Maybe we could factor out the function for just those two types, since the in-memory representation of the pixel slices is exactly the same:
func opaque(pix []RGBAColor, stride int, rect Rectangle) bool
func (p *NRGBA) Opaque() bool {
return opaque([]RGBAColor(p.Pix), p.Stride, p.Rect)
}
Alas, again this isn't allowed. This seems to be more of a spec/language issue than a technical one. I'm sure this has come up on the mailing list before, but I can't find a good discussion of it.
This seems like an area where generics would come in handy, but there's no solution for generics in Go yet.
Why does Go not have generic
types?
Generics may well be added at some
point. We don't feel an urgency for
them, although we understand some
programmers do.
Generics are convenient but they come
at a cost in complexity in the type
system and run-time. We haven't yet
found a design that gives value
proportionate to the complexity,
although we continue to think about
it. Meanwhile, Go's built-in maps and
slices, plus the ability to use the
empty interface to construct
containers (with explicit unboxing)
mean in many cases it is possible to
write code that does what generics
would enable, if less smoothly.
This remains an open issue.