Re: cuteDoc -New DDOC theme
Nick Sabalausky Wrote: Robik szad...@gmail.com wrote in message news:j8f14s$1o78$1...@digitalmars.com... Hi. I'd like to share with new theme for DDOC named CuteDoc. It can be found here: https://github.com/robik/cuteDoc . Live demo can be foudn here: http://cutedoc.dav1d.de/ . (Thanks to dav1d (from #d) for hosting it). Cute theme! (A ha! A pun!) Little suggestion: If you leave the modules and jump to already-expanded in the HTML, and then use JS to flatten them upon page load, then it'll work for everyone, not just those who have JS enabled. It's never good to assume JS. One other issue: When both modules and jump to are expanded, the bottom half of the jump to is hidden under the bottom frame of the browser window and completely inaccessible, and scrolling dosn't help (And that's even when I have the browser maximized). Frames have never really been considered good style on the web, and floating boxes are not far off from frames. You may want to consider making them non-floaty: it'll take a *lot* less work to make it work right for everyone. Thanks for cool suggestions. About second issue, which web browser do you use?
Re: cuteDoc -New DDOC theme
Robik szad...@gmail.com wrote in message news:j8g8oj$qj7$1...@digitalmars.com... Nick Sabalausky Wrote: Robik szad...@gmail.com wrote in message news:j8f14s$1o78$1...@digitalmars.com... Hi. I'd like to share with new theme for DDOC named CuteDoc. It can be found here: https://github.com/robik/cuteDoc . Live demo can be foudn here: http://cutedoc.dav1d.de/ . (Thanks to dav1d (from #d) for hosting it). Cute theme! (A ha! A pun!) Little suggestion: If you leave the modules and jump to already-expanded in the HTML, and then use JS to flatten them upon page load, then it'll work for everyone, not just those who have JS enabled. It's never good to assume JS. One other issue: When both modules and jump to are expanded, the bottom half of the jump to is hidden under the bottom frame of the browser window and completely inaccessible, and scrolling dosn't help (And that's even when I have the browser maximized). Frames have never really been considered good style on the web, and floating boxes are not far off from frames. You may want to consider making them non-floaty: it'll take a *lot* less work to make it work right for everyone. Thanks for cool suggestions. About second issue, which web browser do you use? FF
Re: AI Challenge - Ants
On 10/25/2011 1:44 PM, Trass3r wrote: I'm working on a bot in D. I'm currently done implementing the A* algorithm for path finding Dump A*, D* Lite ftw ;) Hellooo. Correct me if I'm wrong, but in A*, I can just find a path, store it (let's say as a string) and find a new one if it's blocked for some reason. As far as I could see from what I've read about Lifelong A*, D*, Focused D* and D* Lite, I would have to store all nodes used in the algorithm - so, as opposed to A*, I have to conserve the state of the entire search algorithm throughout ticks, for each ant. Is the environment in this AI challenge really noisy enough to warrant this? Obviously, if the path needs to be re-planned almost every tick, D* Lite seems like a better choice to me. But if you have 100+ ants, that would be a lot of allocated heap memory, wouldn't it? I really don't have a clue how the processing vs allocating should be weighed here performance-wise. /Max
Xinok Sort Update
I recently put some time into updating my implementation of xinok sort for D. Major changes include support for random-access ranges and custom predicates (ab). You can download the new version here: http://sourceforge.net/projects/xinoksort/files/D%202.0/2011-10-29/xinoksort.d/download For those that are unaware, I posted about a new sorting algorithm a few weeks ago. Xinok sort is a *stable* sorting algorithm with good performance while only requiring a small amount of constant additional memory. The current stable sort in Phobos is broken and much slower, so I hope to contribute my algorithm to Phobos. But I'm new to this, so I'm not really sure of all what I need to do. I would appreciate if a few people could review my code and suggest any changes or improvements, as well as test for bugs.
Re: AI Challenge - Ants
If you want to cheat, there have been books published on ant colony optimization. I'm sure the related papers could be dug up. Sent from my iPhone On Oct 29, 2011, at 3:12 AM, Max Wolter awishform...@gmail.com wrote: On 10/25/2011 1:44 PM, Trass3r wrote: I'm working on a bot in D. I'm currently done implementing the A* algorithm for path finding Dump A*, D* Lite ftw ;) Hellooo. Correct me if I'm wrong, but in A*, I can just find a path, store it (let's say as a string) and find a new one if it's blocked for some reason. As far as I could see from what I've read about Lifelong A*, D*, Focused D* and D* Lite, I would have to store all nodes used in the algorithm - so, as opposed to A*, I have to conserve the state of the entire search algorithm throughout ticks, for each ant. Is the environment in this AI challenge really noisy enough to warrant this? Obviously, if the path needs to be re-planned almost every tick, D* Lite seems like a better choice to me. But if you have 100+ ants, that would be a lot of allocated heap memory, wouldn't it? I really don't have a clue how the processing vs allocating should be weighed here performance-wise. /Max
Re: AI Challenge - Ants
Hi. I don't think, that you have to use some advanced strategies and path finding. I'm currently 261 with this simple code: http://pastebin.com/1Nsb81rj With hill defense and ant grouping, you could be imho easilly in first 100 without A* :) 2011/10/29 Sean Kelly s...@invisibleduck.org: If you want to cheat, there have been books published on ant colony optimization. I'm sure the related papers could be dug up. Sent from my iPhone On Oct 29, 2011, at 3:12 AM, Max Wolter awishform...@gmail.com wrote: On 10/25/2011 1:44 PM, Trass3r wrote: I'm working on a bot in D. I'm currently done implementing the A* algorithm for path finding Dump A*, D* Lite ftw ;) Hellooo. Correct me if I'm wrong, but in A*, I can just find a path, store it (let's say as a string) and find a new one if it's blocked for some reason. As far as I could see from what I've read about Lifelong A*, D*, Focused D* and D* Lite, I would have to store all nodes used in the algorithm - so, as opposed to A*, I have to conserve the state of the entire search algorithm throughout ticks, for each ant. Is the environment in this AI challenge really noisy enough to warrant this? Obviously, if the path needs to be re-planned almost every tick, D* Lite seems like a better choice to me. But if you have 100+ ants, that would be a lot of allocated heap memory, wouldn't it? I really don't have a clue how the processing vs allocating should be weighed here performance-wise. /Max
Re: Xinok Sort Update
On 10/29/2011 5:53 PM, Timon Gehr wrote:
Looks good =). Thank you. How does this implementation of your algorithm
compare to the the unstable sort that is currently in Phobos,
performance wise?
I posted some benchmarks here. These benchmarks used the specialized
code for arrays. There would likely be a larger gap when using ranges.
https://sourceforge.net/p/xinoksort/blog/2011/10/another-update--benchmarks/
One comment:
while(temp is null){
try temp.length = len;
catch(Exception err){ // Reduce memory usage and try again
len /= 2;
if(len = 8) continue;
else throw err;
}
}
temp.length = len cannot throw an Exception.
I think you are trying to catch an OutOfMemoryError here?
Yes I was. What should I do/use instead?
Re: Xinok Sort Update
On 10/30/2011 12:19 AM, Xinok wrote:
On 10/29/2011 5:53 PM, Timon Gehr wrote:
Looks good =). Thank you. How does this implementation of your algorithm
compare to the the unstable sort that is currently in Phobos,
performance wise?
I posted some benchmarks here. These benchmarks used the specialized
code for arrays. There would likely be a larger gap when using ranges.
https://sourceforge.net/p/xinoksort/blog/2011/10/another-update--benchmarks/
Ok, very nice.
One comment:
while(temp is null){
try temp.length = len;
catch(Exception err){ // Reduce memory usage and try again
len /= 2;
if(len = 8) continue;
else throw err;
}
}
temp.length = len cannot throw an Exception.
I think you are trying to catch an OutOfMemoryError here?
Yes I was. What should I do/use instead?
You could use catch(Error err) or catch(OutOfMemoryError err) or not
catch the Error at all.
Re: Xinok Sort Update
On Sun, 30 Oct 2011 01:56:23 +0300, Timon Gehr timon.g...@gmx.ch wrote: You could use catch(Error err) or catch(OutOfMemoryError err) or not catch the Error at all. Note that (IIRC) an OutOfMemoryError will be thrown only when: 1) There is no space on the managed heap 2) A garbage collection cycle failed to free enough memory for the requested allocation 3) The operating system could not allocate any more memory, even from swap. Some operating systems (Windows) will even expand the swap file automatically when it nears being full. I don't think that there's any point in doing anything sensible in an OutOfMemory handler. -- Best regards, Vladimirmailto:vladi...@thecybershadow.net
Re: Xinok Sort Update
On 10/29/2011 7:19 PM, Vladimir Panteleev wrote: On Sun, 30 Oct 2011 01:56:23 +0300, Timon Gehr timon.g...@gmx.ch wrote: You could use catch(Error err) or catch(OutOfMemoryError err) or not catch the Error at all. I'll use OutOfMemoryError. If any other error occurs, it's probably best to let the function fail. Note that (IIRC) an OutOfMemoryError will be thrown only when: 1) There is no space on the managed heap 2) A garbage collection cycle failed to free enough memory for the requested allocation 3) The operating system could not allocate any more memory, even from swap. Some operating systems (Windows) will even expand the swap file automatically when it nears being full. I don't think that there's any point in doing anything sensible in an OutOfMemory handler. 32-bit processes on Windows can only have up to 2GiB of addressable memory. Even if there's enough available memory, there may not be a large enough area of contiguous free space. I've gotten out of memory errors when working in D. I handle the error because I can. My algorithm doesn't require any minimum amount of memory to be allocated, so I can reduce the memory usage for a small loss in performance.
