ramda/ramda

I have been going through Ramda's source code as a learning exercise and don't understand the following:

Here is a simplified curry3 function

import _curry2 from './fns/_curry2' // copy of Ramda function
import _curry3 from './fns/_curry3' // copy of Ramda function

export function curry3(fn) {
  return function f3(a, b, c) {
    const argsLen = arguments.length
    log('curry3 argLen', argsLen)
    switch (argsLen) {
      case 0:
        return f3
      case 1:
        return _curry2(function (_b, _c) {
          return fn(a, _b, _c)
        })
      case 2:
        return _curry1(function (_c) {
          return fn(a, b, _c)
        })
      case 3:
        return fn(a, b, c)
    }
  }
}

Wrap a function in it

const c3 = curry3(function curryMe3(a, b, c) {
  return `curryMe3: ${a}, ${b}, ${c}`
})

const c3 = curry3(function curryMe3(a, b, c) {
  return `curryMe3: ${a}, ${b}, ${c}`
})

And call it

log('c3-0', c3())
// -> argLen 0
// -> c3 [Function: f3]

log('c3-1', c3(1))
// -> argLen 1
// -> c3 [Function: f2]

log('c3-2', c3(1, 2))
// -> argLen 2
// -> c3 [Function: f1]

log('c3-3', c3(1, 2, 3))
// -> argLen 3
// -> c3 curryMe3: 1, 2, 3

So c3 receives 1 argument as fn. What I dont understand is how the arguments passed to curryMe3 get into f3 as 0, 1, 2 or 3 separate arguments.