Document SaW map format some more

doc/formats/maps.md

@@ -517,3 +517,86 @@ Around 001841A0: mission objectives!
 
 Since all the files are exactly the same length uncompressed, I'm going to
 assume these are all a fixed number of fixed-size records when looking into it.
+
+
+
+
+## Soldiers At War
+
+All the above applies to Chaos Gate maps. Maps for Soldiers At War seem to have
+a lot of similarities, but also some differences. For a start, the maps are a
+variable size!
+
+Starting with the header, given a tiny 26x20 generated map, the first 256 bytes
+look like this:
+
+```
+00000000: 1500414d 425f4d41 50005041 52495300  ..AMB_MAP.PARIS.
+00000010: 00000000 00000000 00000000 00000000  ................
+00000020: 00000000 00000000 00000000 00000000  ................
+00000030: 00000000 00000000 00000000 00000000  ................
+00000040: 00000000 00000000 00000000 00000000  ................
+00000050: 00000000 00000000 00000000 00000000  ................
+00000060: 00000000 00000000 00000000 00000000  ................
+00000070: 00000000 00000000 00000000 00000000  ................
+00000080: 00000000 00000000 00001e00 45000100  ............E...
+00000090: 1f004600 10010000 52000000 00001b00  ..F.....R.......
+000000a0: 38000100 00000500 0a000001 00f0f9ff  8...............
+000000b0: ffb60500 00000100 ff370a00 64006400  .........7..d.d.
+000000c0: 08008501 00000000 00ff0000 1f008082  ................
+000000d0: 01000000 0000ff00 001f0080 84010000  ................
+000000e0: 000000ff 00001f00 00810100 00000000  ................
+000000f0: ff00001f 00808301 00000000 00ff0000  ................
+```
+
+Almost everything we knew is out of the window, but a few things look familiar.
+First, the header seems simplified down to just two recognisable-at-first-glance
+fields: Magic bytes (now `\x15\x00AMV_MAP\x00`) and the set name, coming
+immediately after.
+
+Like Chaos Gate, all map files are the same size once uncompressed, but they are
+smaller - at 1,214,559 bytes, they are 76% the size. This is quite significant.
+We now have 13.3 bytes per voxel, rather than the 17.5 bytes per voxel that was
+available to Chaos Gate. This means that the number of bytes *per cell* must be
+reduced, in addition to the header (and trailer?) values.
+
+Looking at data from 0x110, it seems to group naturally into 13-byte records:
+
+```
+$ xxd -s 0x110 -c 13 -l 65 -g 1 TINYMAP.MAP 
+00000110: 80 01 00 00 00 00 00 ff 00 00 1f 00 00  .............
+0000011d: 85 01 00 00 00 00 00 ff 00 00 1f 00 00  .............
+0000012a: 82 01 00 00 00 00 00 ff 00 00 1f 00 80  .............
+00000137: 82 01 00 00 00 00 00 ff 00 00 1f 00 00  .............
+00000144: 82 01 00 00 00 00 00 ff 00 00 1f 00 80  .............
+```
+
+It's a strange number. Chaos Gate cells group nicely on 16 bytes:
+
+```
+$ xxd -s 0x110 -c 16 -l 64 -g 1 Chapter01.MAP 
+00000110: 3f 00 00 00 83 01 00 00 00 00 00 ff 00 00 00 00  ?...............
+00000120: 38 00 00 00 85 01 00 00 00 00 00 ff 00 00 00 00  8...............
+00000130: 38 00 00 00 84 01 00 00 00 00 00 ff 00 00 00 00  8...............
+00000140: 38 00 00 00 8a 01 00 00 00 00 00 ff 00 00 00 00  8...............
+00000150: 38 00 00 00 83 01 00 00 00 00 00 ff 00 00 00 00  8...............
+```
+
+That grouping is very enticing, though. I feel strongly that it's the right
+number.
+
+Now we need to ask about start offset. Where is byte 0 of the per-cell data, and
+do the 13 bytes it has line up neatly to the functions of some of the 16 bytes
+seen in Chaos Gate?
+
+I generated a `BIGGESTMAP` (130x100) to investigate. It's just grass, nothing
+but grass, and 0xC0 is the first offset where it starts to look nicely grouped:
+
+```
+xxd -s 0xc0 -c 13 -l 260 -g 13 BIGGESTMAP.MAP
+000000c0: 08 80 81 01 00 00 00 00 00 ff 00 00 1f  .............
+000000cd: 00 80 81 01 00 00 00 00 00 ff 00 00 1f  .............
+000000da: 00 00 81 01 00 00 00 00 00 ff 00 00 1f  .............
+000000e7: 00 00 85 01 00 00 00 00 00 ff 00 00 1f  .............
+# ...
+```