Point by point, with axes that grow to fit

The most general case, combining Point by point, raster order’s true one-measurement-at-a-time raster order with a window that isn’t known up front: the underlying grid exists, but nothing is drawn beyond the bounding box of what’s actually been measured so far, so the visible axes widen a step at a time right along with the data.

The code below is exactly what you’d write against the real plotpress.qt.LiveArtist: a callback that receives whatever new points the stage reported since the last tick and pushes the updated grid to the plot, fed by a loop simulating a raster scan stage. Only read_next_points() is meant to be replaced, with your own instrument call.

import numpy as np
import plotpress


NY, NX = 15, 20
gx = np.arange(NX + 1, dtype=float)
gy = np.arange(NY + 1, dtype=float)
VMIN, VMAX = 0.0, 10.0   # the instrument's own known reading range

fig, ax = plotpress.subplots(figsize=(7, 5.5))
grid = np.full((NY, NX), np.nan)
mesh = LiveArtist(ax, cmap="cividis", vmin=VMIN, vmax=VMAX)
bbox = {"r_min": None, "r_max": None, "c_min": None, "c_max": None}

# One-time colorbar: the scale is fixed, so it never needs a per-frame
# refresh the way plot_02_systematic_fill's autoscaled one does.
m0 = ax.pcolormesh(gx, gy, grid, cmap="cividis", vmin=VMIN, vmax=VMAX)
fig.colorbar(m0, ax=ax)


def on_new_points(points):
    """Called once per acquisition tick with whatever ``(row, col, value)``
    points the stage reported since the last one -- push them into the
    grid and widen the visible window to the bounding box of everything
    measured so far.
    """
    for r, c, v in points:
        grid[r, c] = v
        bbox["r_min"] = r if bbox["r_min"] is None else min(bbox["r_min"], r)
        bbox["r_max"] = r if bbox["r_max"] is None else max(bbox["r_max"], r)
        bbox["c_min"] = c if bbox["c_min"] is None else min(bbox["c_min"], c)
        bbox["c_max"] = c if bbox["c_max"] is None else max(bbox["c_max"], c)

    mesh.update(gx, gy, grid)
    ax.set_xlim(bbox["c_min"], bbox["c_max"] + 1)   # cla() wiped this
    ax.set_ylim(bbox["r_min"], bbox["r_max"] + 1)   # -- bbox of what's revealed
    ax.set_aspect("equal")
    ax.set_xlabel("x index"); ax.set_ylabel("y index")
    n_seen = int(np.count_nonzero(~np.isnan(grid)))
    ax.set_title(f"Point by point, growing window: {n_seen}/{NY * NX}")
    fig.tight_layout()


# ---------------------------------------------------------------------------
# Data acquisition -- replace this with your own raster stage controller.
# Everything above only needs a list of (row, col, value) points handed to
# on_new_points() as they're measured.
# ---------------------------------------------------------------------------
rng = np.random.default_rng(5)
cols, rows = np.meshgrid(np.arange(NX), np.arange(NY))
field = np.clip(
    np.exp(-((rows - 7) ** 2 + (cols - 10) ** 2) / 35.0) * 10.0
    + 0.15 * rng.standard_normal((NY, NX)), 0.0, None)

# Same raster order as plot_05_point_by_point: column by column, bottom to
# top within each column.
order = [(r, c) for c in range(NX) for r in range(NY)]
POINTS_PER_TICK = 6


def read_next_points(lo, hi):
    """Stand-in for the stage reporting whichever points it measured this
    tick, in raster order.
    """
    return [(r, c, float(field[r, c])) for r, c in order[lo:hi]]


for lo in range(0, len(order), POINTS_PER_TICK):
    hi = min(lo + POINTS_PER_TICK, len(order))
    on_new_points(read_next_points(lo, hi))