Play The Search for Almazar online

Previously on this blog: “Nifty notebooks, and Almazar (2021-05-13).

In that previous post, I explained how I’d become aware of Winston M. Llamas’s circa-1981 text adventure game The Search for Almazar; googled up two different versions of its BASIC source code and gotten one of them running in a CP/M emulator; and put that source code and instructions on my GitHub (

Three days later, I’ve produced a C99 port! You can compile it natively from the C source, or — better — you can play it online. Let me know if you find any bugs!

This online version uses the same “C99 to Z-code to JavaScript” workflow that I used back in 2013 to produce online versions of Adventure, Adventure II (LUPI0440), Adventure 3 (PLAT0550), and Adventure 6 (MCDO0551). This process deserves its own blog post; but long story short, I compile the C99 code using Volker Barthelmann’s vbcc compiler attached to David Given’s Z-machine backend; this produces Z-code. Then I embed the Parchment interpreter (maintained by Dannii Willis, originally written by Atul Varma, based on Gnusto by Tom Thurman…) to turn that Z-code into something you can interact with on a web page. “It’s giants all the way down.”

Deficiencies of the Osborne 1 version

The version I’d been playing in that CP/M emulator was the one published in the SIG/M software library in 1983, credited to Bob Liddelow. (Liddelow seems to have been affiliated with a user group called “Software Tools of Australia”; he’s also the author of books including A Guide to Native Orchids of South Western Australia and Wine for All.) As I played through it, I discovered three major issues with it:

First: There were typos! Three days ago I’d found four; now I’m up to eight. I’m not talking little spelling mistaeks, either. The first bug I found was that whenever you tried to DROP something, the game would crash with a Syntax error, because that particular codepath misspelled the BASIC keyword AND as ANR. The next bug was that when you tried to LIGHT LAMP, the game would reply “You are not carrying it” — unless you happened to be carrying the ring — because that codepath said 10 (the ring) when it meant IO (the object of the current command). The next two bugs were on the critical path: there are two actions you must take in order to reach the endgame, one after the other. The first of these actions would have no effect because the magic number 79 had been misprinted as 7; after fixing that bug, the second action would still have no effect because verb number 20 had been misprinted as 29. There were also some bugs off the critical path, such as swapping 43 for 34 (breaking the verb SWIM); writing 229 for 29 (breaking the verb DRINK); writing EELSE for ELSE (somehow not a syntax error, but actually nerfing an entire puzzle); and incorrectly making BURN a synonym for KILL instead of for LIGHT.

I can’t satisfactorily explain the shape of these errors: the first two seem like the kind of errors a modern OCR program would make, but never a human (particularly, a human transcriber would never mistake IO for 10 in this context); but the others (particularly swapping 43 for 34 and changing the meaning of BURN) seem easy for a human and impossible for a computer.

Second: In order to fit the game onto the Osborne 1’s 52-column screen, Liddelow abridged some of Llamas’s original text. Also, just like Adventure and Castlequest, Llamas’s original Almazar had both long and short room descriptions; Liddelow (maybe to conserve space on the Osborne, maybe to fit on the SIG/M disk, maybe for aesthetics) removed the short descriptions. This massively affects the “feel” of the game, and also loses at least one nice pun.

Third: Just as with Castlequest and David Long’s Adventure, the critical path for Almazar leads through a combination-lock puzzle. The combination to this lock is revealed in one specific in-game message. Bizarrely, Liddelow’s version changes the numbers in this message without also changing the combination to the lock! So that’s one more obstacle to victory (in addition to the two or three bugs mentioned above).

So, it was really lucky that I also had the 80 Micro magazine version to compare against.

For my port, I did start with Liddelow’s code, because I could dump it straight into a text editor and start adding curly braces and stuff. But I took all the messages and room descriptions (including the restored short descriptions!) from Llamas’s 80 Micro version. Since both versions’ messages were full of typos (and the 80 Micro messages were even ALL CAPS), I felt free to massage their grammar and spelling. One big liberty I took was to abbreviate many of the short room descriptions in Adventure’s telegraphic style: Llamas’s YOU'RE IN A THREE WAY JUNCTION in my port becomes You're at three-way junction. So if you’re a stickler for historical fidelity, keep that in mind!

Thoughts on “simulationism”

I’ve noticed that a lot of adventure games are affected by what I think of as “the pursuit of perfect simulation.” There will be some piece of the program that isn’t really noticeable by the player, but has dozens of lines of code devoted to meticulous simulation. Crowther’s dwarf-movement code is the ur-example of this. (Even though I love Crowther’s dwarf-movement code! See “Observations on Castlequest’s code” (2021-03-21).) Any game with multiple containers (MCDO0551, for example) likely falls into this category, too.

Almazar’s simulationist urge seems to be expressed, remarkably, in its geography. Even though Llamas picked a data representation in which “leaving a room to the north does not guarantee entering the next from the south,” his actual geography is almost exactly based on a square grid. You can count your steps west, take a step north, count the same number of steps back east, and be sure that you’re exactly one step north of your starting point. Even in Llamas’s three explicitly mazey sections — the forest, the maze of twisty passages, and the diamond mine — there are no twists or turns; the “tricks” are limited to making directions into no-ops at the edges of the maze, and reusing room descriptions. I infer that the programmer took a bit of pride in his ability to make every passageway connect up in the most absolutely logical and Euclidean way.

Fun with gnomes

Llamas’s wandering monster is semi-simulationist in a way that’s new to me. Whereas Crowther’s dwarves wander realistically through the same geography as the player; and Holtzman’s gnome and werewolf (and Platt’s dwarves) simply warp into the player’s room at random; Llamas’s knife-throwing gnome tracks the player through a different geography. (Serendipitously, this reminds me of a scene in SCP-4739; but that’s probably the wrong mental image for this gnome!) This codepath runs on every turn you spend inside the cave:

7410 IF OP(21)=0 THEN OP(21)=INT(RND*35)
7420 IF OP(21)>RN THEN OP(21)=OP(21)-1
7425 IF OP(21)<RN THEN OP(21)=OP(21)+1

Visiting the gnome’s headquarters sets SC, the “chased by gnome” flag. Every turn afterward that you spend inside the cave, the gnome has a chance of spawning in a random room — possibly outside the cave! Once the gnome has spawned, every turn you spend in the cave, the gnome will move toward you along the number line. If the gnome has cornered you in the lighted room (room 15), and you move into the tool room (16), the gnome will follow you, because those rooms’ numbers differ by 1; but if you move from the lighted room to the long corridor (17), you’ll gain a one-turn lead and temporarily evade the gnome.

The gnome is in most respects a normal object, but it’s skipped by the code that prints object descriptions. In other words, while you’re outside the cave the gnome is both stationary and invisible. But it’s still there! Wander around outside the cave throwing the dagger at random, and you may eventually see “You killed a nasty knife-throwing gnome!”

Executive summary

I made a version of Winston Llamas’s Almazar you can play online. Play it here.

Posted 2021-05-16