We had pages and pages of random description with no spatial record to prevent these sorts of collisions. We knew we were screwed by the third hallway we graphed, when we found ourselves looping back and intersecting a previously mapped hallway that was not in the earlier description. Once we got our several pounds of printout from the first successful run, we began graphing. Typos meant we had to resubmit this program many times before we got a successful execution and each time we submitted our stack of cards, the computer operator would give us the stink eye. It was the many times larger than the biggest stack of cards we’d ever seen submitted to the computer operator. The resulting mountain of punch cards wouldn’t even fit in one wide shoe box. It took hours and hours to create the program and eliminate all the typos. Our idea was to have the DM take the the computer-generated descriptions and walk through it ahead of time, manually converting text to graphical map representation and noting what monsters were present in each room. We used punch cards– PUNCH CARDS!–and our output was textual descriptions printed on that extra-wide fan-fold paper. Our only computer access was our college’s UNIVAC which our current FORTRAN 77 class was time-sharing.

A friend and I experimented with procedural dungeons on a computer back in the day, using the 1st edition DMG’s random dungeon generator (intended dice rolls).

Random dungeon map generator algorithm