我改编了 DrBoolean 的镜头实现,这样它就可以在没有自省/鸭子打字/依赖原型身份的情况下工作。计算应完全由高阶函数的延续和函数参数确定。我已经走到了这一步:
const id = x => x;
// compose n functions (right to left)
const $$ = (f, ...fs) => x =>
fs.length === 0
? f(x)
: f($$(...fs) (x));
// function encoded Const type
const Const = x => k => k(x);
const constMap = f => fx => fx(x => Const(x));
const runConst = fx => fx(id);
// lens constructor
const objLens = k => cons => o =>
constMap(v => Object.assign({}, o, {[k]: v})) (cons(o[k]));
const arrLens = i => cons => xs =>
constMap(v => Object.assign([], xs, {[i]: v})) (cons(xs[i]));
const view = fx =>
$$(runConst, fx(Const));
const user = {name: "Bob", addresses: [
{street: '99 Maple', zip: 94004, type: 'home'},
{street: '2000 Sunset Blvd', zip: 90069, type: 'work'}]};
// a lens
const sndStreetLens =
$$(objLens("addresses"), arrLens(1), objLens("street"));
console.log(
view(sndStreetLens) (user) // 2000 Sunset Blvd
);
在这个版本中,函子约束constMap
在镜头构造函数中被硬编码。当view
调用一个镜头时,$$(runConst, fx(Const))
已经传递了正确的类型(),但是为了获得一个临时多态镜头,我也需要传递相应的函子实例。
我已经尝试了最明显的方法$$(runConst, fx(constMap) (Const))
,但是,这会干扰构图。我就是想不通。