Generating my training set takes a little longer than I had thought...
on average each character takes 3 seconds to generate. Since it's bedtime anyways, I'll just let it run until I wake up :)
And thanks shaun!
Fuck off windows, why don't you have case sensitive file names. The capital letters are overwriting the non-capital letters now
Pretty much done my project. You can add series and issues to a series with those buttons. Double clicking on anything on the lists deletes them. When you add them they are sorted from smallest to largest for issue numbers and A-Z for series names. Just found a problem where lower case names are treated differently from uppercase. Other then that I need to mess around with button placement and stuff so it looks better!
Once I submit it I'm going to take a break from it, then maybe in a week or two I'll come back and see if I can make it better! Prices and stuff like that would be great.
DOS was case-insensitive, so changing it to a case-sensitive system at any point would have broken everything.
And that's the reason you should treat every design decision as though you would be stuck with it forever.
Made some progress in my shit ass indie as fuck platformer crapfest game. I put a pixelation shader over the game's graphics to mask how ugly the art is.
I'll give you a dollar if you can remember what the two midi songs are from.
Started work on a console
For some reason XNA uses really weird ASCII codes. A-Z 0-9 follow the normal ones, but that's it.
This, my friends, is a triangle rendered in openGL ES 2.0, on Android.
...and it only took all day to make!
What it means is that all that staring at tutorials finally did something.
Now that I have this figured out, I can get on with the real fun in converting my engine from using the android canvas to openGL - everything else!
When I first switched to Linux case sensitivity was a bit of a struggle, I don't actually remember why though. When I first switched to Linux a lot of things upset me, mostly because it wasn't Windows and it just gave me this feeling like being in a forgein country and having to adapt.
Does anybody have experience with using POSIX on Windows, through the POSIX subsystem or Interix? Or is Cygwin better?
Finally got this working:
For some reason, if you do
Keys pressedKeys = KeyboardState.GetPressedKeys();
a-z and 0-9 keys are regular ASCII codes, pretty much every other key are completely different.
The other issue is that frontslash is the standard command switch on Windows, e.g. "dir /O".
Doom 3 is an example of game where everything is parsed from text. Loading times are awful.
And maps are not intended to be ported across games, they are intended to define objects for a specific game.
The same goes for the rest.
I doubt serialization will be your bottleneck
Ever wanted to code something that interests you so bad and you're chained by a job and a buttload of work that needs to be done?
It sucks to be me right now.
2) That's what I'm talking about: increased loading time.
And JSON consumes more disc space than binary data.
It works like this:
The PC and Android are running my own Java programs and the Arduino is running my own Arduino program too. The only (non-standard) library I used was Amarino to make communication between the Arduino and Android easier. For video streaming I'm just sending tons of low-quality JPEGs as fast as I can (I think this is also called Motion-JPEG )
Now I'm thinking about making it automatically follow a track using my phone's GPS.
For small documents, parsing simple textual formats (it doesn't have to be JSON, exactly) takes little time compared to actually seeking to and reading from a file on our big slow, dumb mechanical spinning disk drives. Mostly the seeking to part, because once you're there a sequential read is nothing. A JSON document of this nature would very rarely be any larger than 1MB, and probably most of the time under a few tens of kilobytes or so.
This library claims to be able to parse a ~600kB document in under 2ms. If a drive takes 15-20ms just to find and load the data from the file, you can see that the actual deserialization processes is actually the smaller part of the problem.
So the performance difference between text and binary in this case would be negligible.
The difference between binary and plaintext formats becomes more significant when you're working with tons of very repetitive numerical data.
Oh, also, consider how much content of each type you have to load. When you go through a level transition, you only need to load one list of entities, then you have to load hundreds of other assets (meshes, textures, etc.). Most of which are even larger and more complex than the entity and map properties list itself.
So not only does it not matter whether it's plaintext or binary, but if the JSON version was twenty times slower (which it isn't), it wouldn't make a damn bit of difference overall because that isn't where the loading time cost comes from.
And this is why Knuth said that premature optimization was the root of all evil.
Oh wow, I actually forgot to post what I was working on
Making a little while function with tail calls. (Not optimised, but that's down on the list somewhere!)
The actual code that gets executed is this:
Not properly formatted because I'm only reading and executing one line at a time at the minute. Quoted forms work properly now, and I've added in a load of special operators and stuff. I'm currently reading about three different Lisp books and thinking about quasi-quoting and macros, wow Lisp is weird.Code:(define while (lambda (cond expr) (if (and (eval cond) (or (eval expr) true)) (while cond expr) nil))) (while false '(print 5)) (define i 0) (while '(< i 10) '(setq i (print (+ i 1))))
Fuck it, even zlib-compressed bitmap frames would give a better quality/size ratio, and trust me I'm implementing this stupid flash video codec right now, even it is better than MJPEG.