TypeScript手册给出了答案：

界面的几乎所有功能都是可用的类型。关键区别在于不能重新打开类型以添加新的财产vs始终可扩展的接口。

2019年更新

---

目前的答案和官方文件已经过时。对于那些新接触TypeScript的人来说，如果没有例子，使用的术语就不清楚了。以下是最新差异列表。

1.对象/功能

两者都可以用来描述对象或函数签名的形状。但语法不同。

界面

interface Point {
  x: number;
  y: number;
}

interface SetPoint {
  (x: number, y: number): void;
}

类型别名

type Point = {
  x: number;
  y: number;
};

type SetPoint = (x: number, y: number) => void;

2.其他类型

与接口不同，类型别名也可以用于其他类型，如基元、联合和元组。

// primitive
type Name = string;

// object
type PartialPointX = { x: number; };
type PartialPointY = { y: number; };

// union
type PartialPoint = PartialPointX | PartialPointY;

// tuple
type Data = [number, string];

3.延伸

两者都可以扩展，但语法也不同。此外，请注意，接口和类型别名不是互斥的。接口可以扩展类型别名，反之亦然。

接口扩展接口

interface PartialPointX { x: number; }
interface Point extends PartialPointX { y: number; }

类型别名扩展类型别名

type PartialPointX = { x: number; };
type Point = PartialPointX & { y: number; };

接口扩展类型别名

type PartialPointX = { x: number; };
interface Point extends PartialPointX { y: number; }

类型别名扩展接口

interface PartialPointX { x: number; }
type Point = PartialPointX & { y: number; };

4.工具

类可以以相同的方式实现接口或类型别名。但是请注意，类和接口被视为静态蓝图。因此，它们不能实现/扩展命名联合类型的类型别名。

interface Point {
  x: number;
  y: number;
}

class SomePoint implements Point {
  x = 1;
  y = 2;
}

type Point2 = {
  x: number;
  y: number;
};

class SomePoint2 implements Point2 {
  x = 1;
  y = 2;
}

type PartialPoint = { x: number; } | { y: number; };

// FIXME: can not implement a union type
class SomePartialPoint implements PartialPoint {
  x = 1;
  y = 2;
}

5.申报合并

与类型别名不同，接口可以定义多次，并将被视为单个接口（合并所有声明的成员）。

// These two declarations become:
// interface Point { x: number; y: number; }
interface Point { x: number; }
interface Point { y: number; }

const point: Point = { x: 1, y: 2 };

https://www.typescriptlang.org/docs/handbook/advanced-types.html

一个区别是，接口创建了一个新名称，该名称在任何地方都可以使用。键入别名不会创建新名称-例如，错误消息不会使用别名。

演示递归地重写Object-Literal类型和接口而不是类成员/财产/函数的能力。

此外，当Record<any，string|number>由于接口等原因而无法工作时，如何区分和键入检查差异以及解决上述问题的方法也可以解决。这将允许对猫鼬类型进行以下简化：https://github.com/wesleyolis/mongooseRelationalTypesmongooseRelationalTypes、DeepPopulate、populate

此外，还有一系列其他方法来实现高级类型泛型和类型推理，以及围绕它的速度怪癖，这些都是一些小技巧，可以通过多次试验和错误来获得正确的结果。

打字游戏场：单击此处查看现场游戏场地中的所有示例

    class TestC {
        constructor(public a: number, public b: string, private c: string) {
    
        }
    }
    
    class TestD implements Record<any, any> {
    
        constructor(public a: number, public b: string, private c: string) {
    
        }
    
        test() : number {
            return 1;
        }
    }
    
    type InterfaceA = {
         a: string,
        b: number,
        c: Date
        e: TestC,
        f: TestD,
        p: [number],
        neastedA: {
            d: string,
            e: number
            h: Date,
            j: TestC
            neastedB: {
                d: string,
                e: number
                h: Date,
                j: TestC
            }
        }
    }


    type TCheckClassResult = InterfaceA extends Record<any, unknown> ? 'Y': 'N' // Y

    const d = new Date();
    type TCheckClassResultClass = typeof d extends Record<any, unknown> ? 'Y': 'N'      // N

    const metaData = Symbol('metaData');
    type MetaDataSymbol = typeof metaData;

    // Allows us to not recuse into class type interfaces or traditional interfaces, in which properties and functions become optional.
    type MakeErrorStructure<T extends Record<any, any>> = {
        [K in keyof T] ?: (T[K] extends Record<any, unknown> ?         MakeErrorStructure<T[K]>: T[K] & Record<MetaDataSymbol, 'customField'>)
    }

    type MakeOptional<T extends Record<any, any>> = {
        [K in keyof T] ?: T[K] extends Record<any, unknown> ? MakeOptional<T[K]> : T[K]
    }

    type RRR = MakeOptional<InterfaceA>
    const res  = {} as RRR;

    const num = res.e!.a; // type == number
    const num2 = res.f!.test(); // type == number

使特定形状的递归形状或键递归

    type MakeRecusive<Keys extends string, T> = {
        [K in Keys]: T & MakeRecusive<K, T>
      } & T
  
    type MakeRecusiveObectKeys<TKeys extends string, T> = {
        [K in keyof T]: K extends TKeys ? T[K] & MakeRecusive<K, T[K]>: T[K]
    }

如何为记录类型应用类型约束，该约束可以验证诸如鉴别器之类的接口：

    type IRecordITypes = string | symbol | number;
    
    // Used for checking interface, because Record<'key', Value> excludeds interfaces
    type IRecord<TKey extends IRecordITypes, TValue> = {
        [K in TKey as `${K & string}`] : TValue
    } 


    // relaxies the valiation, older versions can't validate.
    // type IRecord<TKey extends IRecordITypes, TValue> = {
    //     [index: TKey] : TValue
    // } 

    
    type IRecordAnyValue<T extends Record<any,any>, TValue> = {
        [K in keyof T] : TValue
    }    

    interface AA {
        A : number,
        B : string
    }

    interface BB {
        A: number,
        D: Date
    }

    // This approach can also be used, for indefinitely recursive validation like a deep populate, which can't determine what validate beforehand.
    interface CheckRecConstraints<T extends IRecordAnyValue<T, number | string>> {
    }

    type ResA = CheckRecConstraints<AA> // valid

    type ResB = CheckRecConstraints<BB> // invalid

Alternative for checking keys:

    type IRecordKeyValue<T extends Record<any,any>, TKey extends IRecordITypes, TValue> = 
    {
        [K in keyof T] : (TKey & K) extends never ? never : TValue
    } 
    
    // This approach can also be used, for indefinitely recursive validation like a deep populate, which can't determine what validate beforehand.
    interface CheckRecConstraints<T extends IRecordKeyValue<T, number | string, number | string>> {
        A : T
    }

    type UUU = IRecordKeyValue<AA, string, string | number>

    type ResA = CheckRecConstraints<AA> // valid

    type ResB = CheckRecConstraints<BB> // invalid

然而，使用鉴别器的示例，对于速度，我宁愿使用字面上定义每个要记录的键，然后传递来生成混合值，因为使用更少的内存，而且比这种方法更快。

    type EventShapes<TKind extends string> = IRecord<TKind, IRecordITypes> | (IRecord<TKind, IRecordITypes> & EventShapeArgs)

    type NonClassInstance = Record<any, unknown>
    type CheckIfClassInstance<TCheck, TY, TN> = TCheck extends NonClassInstance ? 'N' : 'Y'

    type EventEmitterConfig<TKind extends string = string, TEvents extends EventShapes<TKind> = EventShapes<TKind>, TNever = never> = {
        kind: TKind
        events: TEvents
        noEvent: TNever
    }

    type UnionDiscriminatorType<TKind extends string, T extends Record<TKind, any>> = T[TKind]

    type PickDiscriminatorType<TConfig extends EventEmitterConfig<any, any, any>,
        TKindValue extends string,
        TKind extends string = TConfig['kind'],        
        T extends Record<TKind, IRecordITypes> & ({} | EventShapeArgs) = TConfig['events'],
        TNever = TConfig['noEvent']> = 
            T[TKind] extends TKindValue 
            ? TNever
            : T extends IRecord<TKind, TKindValue>
                ? T extends EventShapeArgs
                    ? T['TArgs']
                    : [T]
                : TNever        

    type EventEmitterDConfig = EventEmitterConfig<'kind', {kind: string | symbol}, any>
    type EventEmitterDConfigKeys = EventEmitterConfig<any, any> // Overide the cached process of the keys.

    interface EventEmitter<TConfig extends EventEmitterConfig<any, any, any> = EventEmitterDConfig,
                TCacheEventKinds extends string = UnionDiscriminatorType<TConfig['kind'], TConfig['events']>
                > {
      on<TKey extends TCacheEventKinds, 
                    T extends Array<any> = PickDiscriminatorType<TConfig, TKey>>(
                        event: TKey, 
                        listener: (...args: T) => void): this;

     emit<TKey extends TCacheEventKinds>(event: TKey, args: PickDiscriminatorType<TConfig, TKey>): boolean;
    }

用法示例：

    interface EventA {
        KindT:'KindTA'
        EventA: 'EventA'
    }

    interface EventB {
        KindT:'KindTB'
        EventB: 'EventB'
    }

    interface EventC {
        KindT:'KindTC'
        EventC: 'EventC'
    }

    interface EventArgs {
        KindT:1
        TArgs: [string, number]    
    }
    const test :EventEmitter<EventEmitterConfig<'KindT', EventA | EventB | EventC | EventArgs>>;

    test.on("KindTC",(a, pre) => {
        
    })

更好的方法来区分类型并从映射中选择类型以缩小范围，这通常会导致更快的性能和更少的类型操作开销，并允许改进缓存。与前面的示例相比。

    type IRecordKeyValue<T extends Record<any,any>, TKey extends IRecordITypes, TValue> = 
    {
        [K in keyof T] : (TKey & K) extends never ? never : TValue
    } 

    type IRecordKeyRecord<T extends Record<any,any>, TKey extends IRecordITypes> = 
    {
        [K in keyof T] : (TKey & K) extends never ? never : T[K] // need to figure out the constrint here for both interface and records.
    } 
    
    type EventEmitterConfig<TKey extends string | symbol | number, TValue, TMap extends IRecordKeyValue<TMap, TKey, TValue>> = {
        map: TMap
    }

    type PickKey<T extends Record<any,any>, TKey extends any> = (T[TKey] extends Array<any> ? T[TKey] : [T[TKey]]) & Array<never>

    type EventEmitterDConfig = EventEmitterConfig<string | symbol, any, any>


    interface TDEventEmitter<TConfig extends EventEmitterConfig<any, any, TConfig['map']> = EventEmitterDConfig,
        TMap = TConfig['map'],
        TCacheEventKinds = keyof TMap
    > {
        
        on<TKey extends TCacheEventKinds, T extends Array<any> = PickKey<TMap, TKey>>(event: TKey, 
            listener: (...args: T) => void): this;

        emit<TKey extends TCacheEventKinds, T extends Array<any> = PickKey<TMap, TKey>>(event: TKey, ...args: T): boolean;
    }
   
    type RecordToDiscriminateKindCache<TKindType extends string | symbol | number, TKindName extends TKindType, T extends IRecordKeyRecord<T, TKindType>> = {
        [K in keyof T] : (T[K] & Record<TKindName, K>)
    }

    type DiscriminateKindFromCache<T extends IRecordKeyRecord<T, any>> = T[keyof T]

用法示例：

    
    interface EventA {
        KindT:'KindTA'
        EventA: 'EventA'
    }

    interface EventB {
        KindT:'KindTB'
        EventB: 'EventB'
    }

    interface EventC {
        KindT:'KindTC'
        EventC: 'EventC'
    }

    type EventArgs = [number, string]

    type Items = {
        KindTA : EventA,
        KindTB : EventB,
        KindTC : EventC
        //0 : EventArgs,
    }

    type DiscriminatorKindTypeUnionCache = RecordToDiscriminateKindCache<string 
    //| number,
    "KindGen", Items>;

    type CachedItemForSpeed = DiscriminatorKindTypeUnionCache['KindTB']

    type DiscriminatorKindTypeUnion = DiscriminateKindFromCache<DiscriminatorKindTypeUnionCache>;

    function example() {
        
        const test: DiscriminatorKindTypeUnion;
        switch(test.KindGen) {
            case 'KindTA':
                test.EventA
                break;
            case 'KindTB':
                test.EventB
                break;
            case 'KindTC':
                test.EventC

            case 0:
                test.toLocaleString

        }
    }


    type EmitterConfig = EventEmitterConfig<string 
    //| number
    , any, Items>;

    const EmitterInstance :TDEventEmitter<EmitterConfig>;

    EmitterInstance.on("KindTB",(a, b) => {
        
        a.

    })

想加上我的2美分；

我曾经是“界面爱好者”（除了联合、交叉等以外，我更喜欢界面而不是类型）。。。直到我开始使用类型“any key value object”，即Record＜string，unknown＞

如果键入“任意键值对象”：

function foo(data: Record<string, unknown>): void {
  for (const [key, value] of Object.entries(data)) {
    // whatever
  }
}

如果你使用接口，你可能会走到死胡同

interface IGoo {
  iam: string;
}

function getGoo(): IGoo {
  return { iam: 'an interface' };
}

const goo = getGoo(); 

foo(goo); // ERROR
// Argument of type 'IGoo' is not assignable to parameter of type 
// 'Record<string, unknown>'.
//  Index signature for type 'string' is missing in type 
// 'IGoo'.ts(2345)

虽然打字就像一个符咒：

type Hoo = {
  iam: string;
};

function getHoo(): Hoo {
  return { iam: 'a type' }; 
}

const hoo = getHoo(); 

foo(hoo); // works

这个特定的用例——IMO——产生了不同。

根据我最近看到或参与的所有讨论，类型和接口之间的主要区别在于接口可以扩展，而类型不能扩展。

此外，如果您两次声明一个接口，它们将被合并为一个接口。你不能用打字。