1 Utility for Games

[Han]

(03-11-2014 06:42 PM)patrice Wrote:  

(03-11-2014 03:18 PM)Han Wrote:  Even shorter is to code multiple $'s in the form n$. For example, $$$ should be coded as 3$. It shouldn't be all that hard to create a small program to convert from matrix over to a string -- it's easier to edit a matrix than it is to create a string and keep all the codes correct.

You perfectly right, if you look at the decoder, you will see that it handle the case already.
I didn't handled it in the coder for 3 reason:

• 3 or more empty lines in a row is not very common over the 40 levels.
• it complicate the encoder for only a very little gain.
• I like to see all the $ , it help me do manual checking of the coding.
  

If you like it, feel free to tweak the various levels, it saves a bout 30 chars across the 40 levels.
When I first put it in Sokoban, I didn't removed the drawscreen() comand which have a side effect, it removes the pusher from the level matrix. So I have done some manual testing before I understood where was the problem.

Yeah, I guess that would be a much smaller gain if there are not many blank rows. I was basing this idea off of how I encode bitmaps as sub-bitmaps for this game. There are certain bitmaps that are used very frequently (i.e. a "default" bitmap). In your particular example, the b and w's appear most frequently. One of those (say b) can be considered the default. So rather than encoding as: "3b4w" you save a few more characters by using "34w" -- with the understanding that a digit without an alpha code refers to repeated blanks. If the digits need to be larger than 0-9, then you can just use A through Z in upper case (gives you up to 26 repeats). Or use any consecutive block of characters for the count, and a separate block of characters for the type. Thus you would have either: <ascii char for count> [<ascii char for type>] (the second value being optional). The count char should come from A through Z (for example) and the type perhaps from a through z (all lower case).

Without doing any hard analysis, a quick glance suggests that most of the code is a digit followed by b or a digit followed by w. So roughly speaking ~20% might be saved? Since it is done during initialization, you would hardly notice the extra bit of code/time needed to handle the differentiation between <count><code> vs. just <count> by itself (for the default "blank" code).