golang不允许循环import package,如果检测到import cycle，会在编译时报错，通常import cycle是因为设计错误或包的规划问题。
以下面的例子为例，package a依赖package b，同事package b依赖package a

package a

import (
  "fmt"

  "github.com/mantishK/dep/b"
)

type A struct {
}

func (a A) PrintA() {
  fmt.Println(a)
}

func NewA() *A {
  a := new(A)
  return a
}

func RequireB() {
  o := b.NewB()
  o.PrintB()
}

package b：

package b

import (
  "fmt"

  "github.com/mantishK/dep/a"
)

type B struct {
}

func (b B) PrintB() {
  fmt.Println(b)
}

func NewB() *B {
  b := new(B)
  return b
}

func RequireA() {
  o := a.NewA()
  o.PrintA()
}

就会在编译时报错：

import cycle not allowed
package github.com/mantishK/dep/a
  imports github.com/mantishK/dep/b
  imports github.com/mantishK/dep/a

现在的问题就是：

A depends on B 
B depends on A

那么如何避免？
引入package i， 引入interface

package i

type Aprinter interface {
  PrintA()
}

让package b import package i

package b

import (
  "fmt"

  "github.com/mantishK/dep/i"
)


func RequireA(o i.Aprinter) {
  o.PrintA()
}

引入package c

package c

import (
  "github.com/mantishK/dep/a"
  "github.com/mantishK/dep/b"
)

func PrintC() {
  o := a.NewA()
  b.RequireA(o)
}

现在依赖关系如下：

A depends on B
B depends on I
C depends on A and B