postconstruct does not support dependency, postconstruct executes random about di HOT 10 CLOSED

Please, provide more detailed description. And also steps to reproduce the issue (code snippet).

from di.

Please, provide more detailed description. And also steps to reproduce the issue (code snippet).

1. code example:

package main

import (
	"fmt"
	"reflect"

	"github.com/goioc/di"
)

type A1 struct {
	Name string
}

func (a1 *A1) PostConstruct() error {
	//time.Sleep(2 * time.Second)
	a1.Name = "a1"
	return nil
}

type A2 struct {
	A1      *A1 `di.inject:""`
	Message string
}

func (a2 *A2) PostConstruct() error {
	a2.Message = "hello, " + a2.A1.Name
	fmt.Printf("a2.A1: %v, a2.Message: %v \n", a2.A1, a2.Message)
	return nil
}

func main() {
	for i := 0; i < 20; i++ {
		di.RegisterBean("A1", reflect.TypeOf((*A1)(nil)))
		di.RegisterBean("A2", reflect.TypeOf((*A2)(nil)))
		di.InitializeContainer()
		di.Close()
	}
}

2. test result:

a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{a1}, a2.Message: hello, a1 
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{a1}, a2.Message: hello, a1 
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{a1}, a2.Message: hello, a1 
a2.A1: &{a1}, a2.Message: hello, a1 
a2.A1: &{a1}, a2.Message: hello, a1 
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  
a2.A1: &{}, a2.Message: hello,  

from di.

OK, I see now, thanks. The reason is that there's no defined order for calling PostConstruct(), it's indeed random. When the PostConstruct() is called for A2 - it doesn't necessarily mean that it was already called for A1.

I'm not sure if it's something to be fixed, to be honest, because most likely your real scenario can be implemented in a better way, i.e. by defining whatever you like to inject as a bean.

But, out of curiosity, how do you see the solution for this? Would you like to have a feature that would allow ordered initialization with PostConstruct()?

from di.

OK, I see now, thanks. The reason is that there's no defined order for calling PostConstruct(), it's indeed random. When the PostConstruct() is called for A2 - it doesn't necessarily mean that it was already called for A1.

I'm not sure if it's something to be fixed, to be honest, because most likely your real scenario can be implemented in a better way, i.e. by defining whatever you like to inject as a bean.

But, out of curiosity, how do you see the solution for this? Would you like to have a feature that would allow ordered initialization with PostConstruct()?

i would better to ask you a better implement on this scenario,
then, allow ordered initialization with PostConstruct() is much better （i also test spring, it's postconstruct is ok）
thanks very much

from di.

How about this?

type A1 struct {
	Name *string `di.inject:"A1Name"`
}

type A2 struct {
	A1      *A1 `di.inject:""`
	Message string
}

func (a2 *A2) PostConstruct() error {
	a2.Message = "hello, " + *a2.A1.Name
	return nil
}

func (suite *TestSuite) Test() {
	for i := 0; i < 20; i++ {
		a1Name := "a1"
		RegisterBeanInstance("A1Name", &a1Name)
		RegisterBean("A1", reflect.TypeOf((*A1)(nil)))
		RegisterBean("A2", reflect.TypeOf((*A2)(nil)))
		InitializeContainer()
		a2 := GetInstance("A2").(*A2)
		fmt.Printf("a2.A1: %v, a2.Message: %v \n", a2.A1, a2.Message)
		Close()
	}
}

If you define the name for the A1 as a string-bean, then the problem is solved, because now it's an explicit dependency which is guaranteed to be already set by the time you call PostConstruct() for A2.

from di.

thanks very much,
also looking forward to the ordered initialization with PostConstruct feature ~

from di.

This issue is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 5 days.

from di.

thanks very much, also looking forward to the ordered initialization with PostConstruct feature ~

how about this feature @dtitov

from di.

Sorry, but currently I have no time for this. If you like - feel free to provide a pull request.

from di.

This issue is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 5 days.

from di.