Micro Mouse

MICROMOUSE is a maze-navigation simulation in which a cursor character wanders through a randomly generated obstacle field, trying to reach a fixed goal at screen position (20,30). The maze is built using a combination of inverse-video block characters and block graphics printed at randomised AT positions during a setup loop (lines 30–60). The mouse’s movement logic tests eight compass directions by issuing a PRINT AT to the candidate cell, then reading the display RAM address stored in system variables G (16398) and H (16399) to determine whether that cell is blank (character code 0), implementing a collision-detection technique entirely through PEEK of the display file pointer. A move counter Q accumulates attempts, and on reaching the goal the score is displayed briefly before the maze is regenerated.

---

Program Analysis

Program Structure

The program divides into four logical phases:

1. Initialisation (lines 1–6): Stores the display-file pointer address in G (16398) and H (16399).
2. Maze generation (lines 10–68): Draws top and bottom border rows of inverse spaces, then randomises wall segments and block graphics across the interior in a FOR A=1 TO 20 loop. A second short loop (lines 61–68) animates the goal marker at (20,30) by flashing it.
3. Movement engine (lines 70–390): Places the mouse at a random starting position, then repeatedly probes adjacent cells in up to eight directions using the display-RAM PEEK technique. A counter Q records total moves.
4. Win condition & restart (lines 2000–2070): On reaching (20,30), prints the move count, pauses with a FOR/NEXT delay, redraws the border, and branches back to line 30 to regenerate the maze.

Display-RAM Collision Detection

The most technically interesting feature is the use of system variables at addresses 16398–16399, which together form the two-byte pointer to the current cursor position in the display file. The idiom used throughout is:

• PRINT AT A+1,B; — moves the display cursor to a candidate cell without printing anything visible.
• IF PEEK (PEEK G+256*PEEK H)=0 THEN LET T=1 — dereferences the cursor pointer to read the character code at that cell. A value of 0 means a space (empty), so the move is valid.

This avoids the need for any array-based map representation; the screen itself is the maze data structure. The technique exploits the ZX81/TS1000 system variable DF_CC (display file current character address), held at 16398–16399 in little-endian format, making PEEK G + 256*PEEK H the full 16-bit address.

Movement Logic

Eight candidate directions are attempted, selected partly randomly via Y=RND*7+1 and partly through a fallback chain. The direction codes and their target lines are:

Y value	Direction	Line	Delta (row,col)
1 / default	Down	120	(+1, 0)
2	Up-right diagonal	154	(−1,+1)
3	Down-right diagonal	200	(+1,+1)
4	Up	250	(−1, 0)
5	Left	290	(0,−1)
6	Right	169	(0,+1)
7	Up-left diagonal	330	(−1,−1)

If none of the tested directions is free, T remains 0 and the engine loops back to line 110 to retry with a new random direction. The previous position E,F is erased with a space at line 1000 before stamping the mouse character at the new position.

Key BASIC Idioms

• RND scaling: RND*N+1 is used throughout for integer ranges, relying on ZX81 BASIC’s 0–0.9999… RND range truncated via multiplication without an explicit INT call — valid because PRINT AT accepts fractional row/column arguments by truncating internally.
• Delay loop: FOR N=1 TO 3000: NEXT N at lines 2010–2020 provides a pause after the win message without using PAUSE.
• Flicker animation: The goal marker at (20,30) is toggled between a space and %* (inverse asterisk) six times in lines 61–68 to draw attention to the target before the mouse is released.

Maze Generation Technique

The maze interior is built entirely from random PRINT AT statements placing inverse spaces (% ), block graphics (\##, \;;\!!), and short wall segments (% % ). There is no guarantee of solvability; the mouse may occasionally become fully trapped. The left and right borders are drawn per-row inside the setup loop (lines 42 and 50), while top and bottom solid borders are printed before the loop (lines 10 and 20).

Bugs and Anomalies

• Line 106 tests IF RND>-2476, which is always true (RND is never negative). This means the random-direction branch at line 110 is never skipped, and the “straight down” default at line 120 is never taken as a direct fall-through from line 95. The conditional is effectively dead code.
• Similarly, lines 152 (RND>-2), 165 (RND<-2), 195 (RND<-6), and 245 (RND<-1) use impossible RND values (always positive), making those fallback branches either always-taken or never-taken. The movement priorities are therefore fixed rather than truly probabilistic.
• Line 2050 CLEAR and lines 2060–2070 are unreachable dead code following the GOTO 30 at line 2040.
• The boundary check at line 154 tests B=30 (the goal column) rather than B=31, which could prevent rightward diagonal moves unnecessarily near the right edge.