I tried two Macintosh tools for
capturing schematic floorplans composed of rectangles:
1.) Form.Z in drafting mode, and 2.) Canvas.
I found that Canvas is more convenient for this task.
Form.Z is a true 3D solid modeling tool and thus far more powerful
than Canvas; but this also means it is harder to learn,
you have to watch and set more menus, and more things can go wrong.
The two tools compare like a helicopter (Form.Z) to a town car (Canvas);
when you just want to go to the post office, the car is more suitable.
Start with a small and simple floor, and get your basic image set-up right.
Choose the proper snap-on grid resolution (1 foot = 1/16 of an inch).
Choose proper line widths (1 pixel for abutting corridor tiles;
2 pixels for rooms surrounded mostly by walls;
3 pixels for inaccessible areas such as vertical HVAC shafts).
Choose proper fonts for labeling the rectangles (Helvetica, size 12, bold).
Choose proper colors (e.g., blue for offices, green for seminar rooms,
brown for labs, purple for utilities ...)
Stick with these settings for all floors.
Once you have done one floor and are happy with the basic style and set-up, it will be much easier to do the other ones. When you do another floor in the same building, always start by copying a previously completed floor; this will make sure you have all the same settings and colors. Then you mostly just use these commands: ungroup (cmd-u); cut; copy; paste; resize rectangles; edit labels; group (cmd-g). Only rarely would you draw a new rectangle from scratch, because it is easier to use a suitable existing rectangle and modify it; this guarantees that you will have the right color, line style, pattern, label. Another advantage of using an existing floor to make a new drawing is that you already have a reference frame, and many items may lie in exactly the same locations, elevators and HVAC shafts, or certain groops of rooms.
Spend some time thinking about the hierarical organization of your plans!
Goupings are crucial. If they are done well, they will help you tremendously
when you have to edit your drawings later; if they are done badly,
you will encounter extreme frustration ...
At the lowest level, I grouped labels such as room numbers or descriptors
to the rectangle that they refer to. But I did this grouping only after
I was fairly sure that the rectangle was of the final size I wanted,
because - unfortunately - as you resize a shape, the grouped text scales with it.
At the next level, I grouped collections of labeled rectangles
that might get re-used as clusters, e.g. groups of offices, or the elements
in the core of each floor: elevators, HVAC shafts, electrical closets.
When I started on a new, modified floor, I grouped the old floor from which
I started and then put an ungrouped copy of the same elements on top of it.
This gave me a fixed indestructable reference which could easily be removed
at the end when the new floor plan was designed, and enough free elements
to modify and to play with.
(You could also put the reference layout into a separate layer.)
When you create abstractions of such floor plans, you proceed in the same way and start with the more complicated floorplan. Now you just enlarge a few of the rectangles to become representatives of a whole group of spaces.
I saved the schematics generated in Canvas as PICT files. These I read in with Photoshop and saved them as TIFF files (either format, ibm-pc or mac, seems to work, as long as you ship them in "raw" mode over the net).
The black-and-white (crummy-looking) floorplans
were scanned in on a Macintosh scanner from copies
of annotated copies of some old floorplans.
They were read in with Photoshop as PICT files.
They were converted to "gray level" images,
so that they could then be cropped to just the interesting part of the image.
If necessary, the image was rotated 90 degrees
and then saved as a 1-bit PICT image.
These were then read into Photoshop again and saved as TIFF files.
The program Fetch was used to ship them to the Unix world. When transferring the TIFF files, it was crucial to select the "raw data" transfer mode to create a file that was readable by the image converter program xv.
In the Unix environment, these TIFF files were then loaded into xv and, after possible scaling, saved as gif-files. These files are not too big, there is no visible loss of quality, an Mosaic as well as netscape can readily render these files.