sourcedemos/wave-tank/wave-tank.xtl
1⍝!/usr/bin/env xetal
2⍝# A wave tank: the wave equation on a grid. Each step, every cell
3⍝# moves towards the average of its neighbors, carrying momentum from
4⍝# the step before: next = 2 u - previous + c^2 * (neighbors - 4 u).
5⍝# Walls are a mask, a source shakes up and down, and waves spread,
6⍝# bend through slits and interfere, all as array expressions over
7⍝# the whole grid. Run it with "xetal run
8⍝# demos/wave-tank/wave-tank.xtl"; it prints the state's shape and the
9⍝# surface drawn with characters (crests, troughs, the wall). The live
10⍝# page picks the scene (a double slit, a lens, two sources, an open
11⍝# tank) and lets you click for ripples:
12⍝# https://softwarewrighter.github.io/X_eTaL-demos/wave-tank/
13
14⍝## The tank
15
16⍝# rows, cols: the grid's size. c2: the wave speed squared (at most
17⍝# 0.5 for a stable step). amp, omega: the source's strength and
18⍝# frequency. steps: how many to run.
19rows ← 36
20cols ← 72
21c2 ← 0.3
22amp ← 0.5
23omega ← 0.5
24steps ← 150
25
26⍝## Coordinates and the sponge
27
28⍝# col, row: each cell's position, by broadcasting.
29⍝# edge, damp: a cell's distance to the border; near the edges the
30⍝# surface is damped (a sponge), so waves leave instead of wrapping
31⍝# around.
32col ← (o̲ffsets rows) 'r̲ight t̲able o̲ffsets cols
33row ← (o̲ffsets rows) 'l̲eft t̲able o̲ffsets cols
34edge ← (row m̲in (rows − 1) − row) m̲in col m̲in (cols − 1) − col
35damp ← 1.0 − 0.03 × f̲loat 0 m̲ax 8 − edge
36
37⍝ -- the scene: a double slit, as masks (plain #: the page's web
38⍝ app finds this exact text to cut the program it runs, right up to
39⍝ it, so a ## here would leave a stray # in what it cuts). slits,
40⍝ wall: a wall down column 30 with two gaps. src: the source, the
41⍝ whole of column 10, a plane wave. The live page replaces this
42⍝ scene with the one you pick (a double slit, a single slit, two
43⍝ sources, a lens, an open tank), each its own wall and src built
44⍝ the same way, from row and col.
45slits ← ((row ≥ 13) ∧ row < 16) ∨ (row ≥ 20) ∧ row < 23
46wall ← f̲loat n̲ot (col = 30) ∧ n̲ot slits
47src ← f̲loat col = 10
48
49⍝# -- the Laplacian: four shifted copies minus four times the center.
50⍝# u:l_ap x: how much a cell differs from its four neighbors, up down
51⍝# left right (o_-_1, o_-_2), added and the center subtracted four
52⍝# times over (the edges wrap).
53ᵘl̲ap ← { x → ((1 o̲-₁ x) + (-1 o̲-₁ x) + (1 o̲-₂ x) + -1 o̲-₂ x) − 4.0 × x }
54
55⍝## One step for every cell at once
56
57⍝# u:s_tep (u, p, t): the state is a tuple, the surface now, the
58⍝# surface a step ago, and the time, taken apart by the pattern and
59⍝# given back as a new tuple (D7, landed); t is a plain scalar now,
60⍝# not a plane carrying the same number everywhere, so the source's
61⍝# drive reads it directly. Walls hold the surface at 0 (damp * wall
62⍝# both zero there).
63ᵘs̲tep ← { (u, p, t) →
64 drive ← amp × src × s̲in omega × t
65 nxt ← damp × wall × ((2.0 × u) − p) + (c2 × ᵘl̲ap u) + drive
66 (nxt, u, t + 1.0)
67}
68⍝ -- end of the core ---------------------------------------------------
69
70⍝# u0: the still surface. u, p, t: the state after all the steps
71⍝# (function power), taken apart by the same pattern the binding
72⍝# reads it into.
73u0 ← (rows c̲at cols) r̲eshape 0.0
74(u, p, t) ← steps 'ᵘs̲tep p̲ower (u0, u0, 0.0)
75s̲hape u
76
77⍝# The surface: = and - troughs, + and * crests, # the wall.
78h ← f̲loat 1 + f̲loor 2.5 + 2.0 × (u m̲ax -1.2) m̲in 1.2
79(f̲loor (wall × h) + (1.0 − wall) × 6.0) s̲elect "=- +*#"