more text compression, 7054 bytes free

1 files changed, 24 insertions, 1 deletions

diff --git a/textcomp.c b/textcomp.c
index 1ff769f..bb0b4c0 100644
--- a/textcomp.c
+++ b/textcomp.c
@@ -4,7 +4,28 @@
 	Example: "Taipan" (7 bytes, including null terminator) encodes as
 	0xb8 0x12 0x50 0x04 0xe0 0x00 (6 bytes).
 
-	Longer strings approach 75% compression ratio.
+	Longer strings approach 75% compression ratio. Sadly, the result
+	has to be padded to an 8-bit byte boundary, or else we'd get 75%
+	for every string.
+
+	Input length | Encoded length | Ratio
+	(incl. null) |                |
+	2            | 2              | 100%, don't bother
+	3            | 3              | 100%, don't bother
+	4            | 3              | 75%
+	5            | 4              | 80%
+	6            | 5              | 83%
+	7            | 6              | 86%
+	8            | 6              | 75%
+	9            | 7              | 78%
+	10           | 8              | 80%
+	11           | 9              | 82%
+	12           | 9              | 75%
+	13           | 10             | 77%
+	14           | 11             | 79%
+	15           | 12             | 80%
+	16           | 12             | 75%
+	...etc etc
 
 	No encoded string can be over 256 bytes long, as the decompressor
 	can't currently handle it.
@@ -38,6 +59,8 @@
 	print an encoded message. The decoding step slows down printing a bit,
 	but it's not really noticeable. cputc() is used for printing, so it
 	respects the reverse video setting (set by rvs_on() and rvs_off()).
+	The task of replacing cputs("some string") with print_msg(M_some_string)
+	is done manually.
 
 	When a newline is printed, the decoder always prints a carriage
 	return first. Any \r sequences listed in messages.pl are discarded