Coarse getMoonTimes

[andersborgabiro]

It seems very coarse to check for cross-overs each 2 hours. Could this be made a function argument?

Stargazing calls for something like 10 minutes resolution to be on the safe side, provided the calculation is otherwise correct.

Thanks in advance.

[Artoria2e5]

The idea to check each 2 hours is sound, but the lack of further narrowing down is... obviously not. A proper root-finding algorithm should be used to narrow it down, something along the lines of:

let hx = (x) =>
  SunCalc.getMoonPosition(hoursLater(t, x), lat, lng).altitude - hc;

for (var i = 1; i <= 24; i += 2) {
  h2 = h(i+1);
  if (h0 * h2 <= 0) {
    // crosses zero, good to solve
    // assume a JS port of https://docs.rs/kurbo/0.8.1/kurbo/common/fn.solve_itp.html
    let root = solve_itp(
      // function and bounds
      hx, i - 1, i + 1,
      // 0.01h max error -- that's tiny!
      0.01,
      // irrelevant hyperparameters
      1, 0.2 / 2,
      // initial values
      h0, h2
    );
    if (h0 < 0) rise = root;
    else set = root;
  }
  h0 = h2;
}

(The step-size can actually be made much larger in this case, since ITP will handle the narrow-down. The tricky thing is that I haven't figured out what to do when the moon actually goes up-then-down in a two-hour window -- the quadratic thing will catch it with a middle point, while here it won't even try to solve! I guess making it two subintervals according to a quadratic guess might be the way to go:)

for (var i = 1; i <= 24; i += 2) {
  h2 = h(i+1);
  if (h0 * h2 <= 0) {
    // crosses zero, good to solve
    // assume a JS port of https://docs.rs/kurbo/0.8.1/kurbo/common/fn.solve_itp.html
    let root = solve_itp(
      // function and bounds
      hx, i - 1, i + 1,
      // 0.01h max error -- that's tiny!
      0.01,
      // irrelevant hyperparameters
      1, 0.2 / 2,
      // initial values
      h0, h2
    );
    if (h0 < 0) rise = root;
    else set = root;
  } else {
    h1 = h(i);
    // Look at h1, do two separate intervals for itp, splitting at the quadratic "xe"
  }
  h0 = h2;
}

But do note that getMoonPosition itself might be a bit wrong (#102), and without that fixed any root-finding algorithm will be bad.

[andersborgabiro]

Thanks. I'll try this. Amateur astronomers are picky about when the moon is up (or rather not up).