I've always wanted to go back to the concept of leximorphs, first introduced in Part V of the megadungeon training series. A leximorph is similar to a geomorph, except that instead of using a pre-drawn, scaled dungeon "piece", you use a letter as a pattern to lay out and draw that section of dungeon yourself. It's thus a little more work than using a geomorph, but the advantage is that you have a little more flexibility and can use various text sources to arbitrarily design a large dungeon.
Previously, I described only one method of creating dungeon rooms from a leximorph. There's actually three basic approaches to creating a simple leximorph, as seen in this illustration:
Method 1 is to use each section of a letter to define the boundaries of a room. In this case, the enclosed area at the top of the "A" forms a trapezoidal room, and the legs of the "A" define another room which I've squared off; the rooms are then connected in any way desired.
Method 2 is the method described in the megadungeon series: use enclosed areas to define rooms as for Method 1, but use open areas (the legs of the "A") to define corridors.
Method 3 is the complement of Method 1: the strokes of the letter define the basic structure, including connecting corridors, and rooms are placed at junctions or endpoints.
The leximorph can be rotated to any desired angle. Complex leximorphs can be built by overlaying two letters in two different orientations. The letters can either be merged, or they can be used to define two sublevels at different heights relative to the main dungeon. Complex leximorphs can also be described as two-pass, especially if your dungeon construction process involves placing one set of leximorphs in one orientation (e.g., upright,) then making a second pass over the same positions to place leximorphs in another orientation (counterclockwise 90 degrees.)
In a follow-up article, I will describe how to use two text sources in a two-pass, multi-layer process to create a random dungeon level.