Originally by: Xessej

---

Originally by: Liang Nuren

---

Edited by: Liang Nuren on 18/08/2010 05:42:41

Originally by: Xessej

---

You intentionally wrote bad code as if that would prove anything. Any code you can write in Python I can write in C++ and it will run faster (Python is written in C++ and is typeless so it has to produce less efficient code than well written C++ that is a simple fact.)

---

Congratulations for completely missing the point. See, the point was that its all about the algorithm. People say "Switch to C++!" like its some magic cure-all but the truth of the matter is that its likely to be an enormous time investment for very little performance gain - if any. And that argument totally neglects how easy it is to maintain one over the other.

---

You do know that your metrics are meaningless since you are intentionally writing inefficient C++ code and comparing it to efficient (well as efficient as Python can get) Python.

---

Originally by: CCP Tuxford

---

my bad. Rest assured I'm being ridiculed by my co-workers.

---

Originally by: Xessej

---

You do know that your metrics are meaningless since you are intentionally writing inefficient C++ code and comparing it to efficient (well as efficient as Python can get) Python.

---

Originally by: Xessej

---

Originally by: Liang Nuren

---

A bunch of BS-Liang
Ed: code tags. Also, I don't claim either implementation is necessarily great - in fact I deliberately did a bad job with C++ to illustrate my point.

---

You intentionally wrote bad code as if that would prove anything. Any code you can write in Python I can write in C++ and it will run faster (Python is written in C++ and is typeless so it has to produce less efficient code than well written C++ that is a simple fact.)

---

Originally by: AccesiViale

---

If someone wanted the leather seats they should have had them in the spec before hand. Good planning makes for a good product. Adding and cutting features right up until release is awful but done anyway as we see with CCP all the time. :)

---

Quote:

---

Cool to see someone even spend the time throwing a small test to illustrate a point. Code <3

---

Originally by: Liang Nuren

---

Edited by: Liang Nuren on 18/08/2010 06:00:46

Originally by: AccesiViale

---

If someone wanted the leather seats they should have had them in the spec before hand. Good planning makes for a good product. Adding and cutting features right up until release is awful but done anyway as we see with CCP all the time. :)

---

Ok, you've done a bang up job and released Eve Apocrypha in well factored pure C++. But those damn customers still aren't happy! They want ****ing fuel bays on that god damn black ops and they're *****ing about supercarriers being underpowered and doomsdays being overpowered. The game dev also came up with these ideas to squeeze into the next expansion:
<insert ideas which totally 100% violate the entire premise behind your well factored C++>

Now - do you want to rewrite 250K lines of C++ or 2K lines of Python? Also, don't forget about all that lovely error handling in C++...
(

---

Originally by: AccesiViale

---

Lol lets be honest now, it can be done and every expansion would not be a total rewrite if you had built a system that was meant to expand (lots of work). Eve is not the only video game out there to have expansions come out. By ccp...maybe not. Not on 6 month time table and with resources(time/money) they have and I would assume that is pretty much the entire reason it is not done.

---

Quote:

---

This conversation is useless though...as if anybody at ccp cares about how the community believes they should or should not have developed something.

---

Originally by: Liang Nuren

---

Congratulations for completely missing the point. See, the point was that its all about the algorithm. People say "Switch to C++!" like its some magic cure-all but the truth of the matter is that its likely to be an enormous time investment for very little performance gain - if any. And that argument totally neglects how easy it is to maintain one over the other.

---

#include <algorithm>
#include <iostream>

unsigned int construct_palindrome(unsigned int n) {
 unsigned pal = n;
 while (n) {
 pal = pal * 10 + n % 10;
 n /= 10;
 }
 return pal;
}

bool is_divisible_by_three_digit_numbers(unsigned int n) {
 for (unsigned int i = 999; i >= 100; i--)
 if (n % i == 0 && n / i < 1000 && n / i >= 100)
 return true;

 return false;
}

int main(int argc, char** argv) {
 unsigned int n = 1000;

 while (--n >= 100) {
 unsigned int pal = construct_palindrome(n);
 
 if (is_divisible_by_three_digit_numbers(pal)) {
 std::cout << "Found pal: " << pal << "." << std::endl;
 return 0;
 }
 }

 return -1;
}

Originally by: Tres Farmer

---

And if you get hit by a bus and some other poor dev has to decipher your part of the code we'll all be in BIG FAT ASS KICKING trouble down the road.

You armchair programmers and code *****s can say what you want.
Python is easier to maintain after x-years of several people contributing to the code base and that's what counts here for CCP.

Also, from the newest dev blogs it doesn't look like we've got trouble in form of lag because of the programming language or database used by ccp.. we're facing trouble because of the n^(n-1) nature problem by players staying close and poking each other.
<whatever prog language> might work faster, but the cap would be there nonetheless..

---

Originally by: Rakshasa Taisab

---

You solved the problem in the stupidest way possible in the C++ version, and try to use it as an argument. That's just stupid.

---

Originally by: Xessej

---

I am a developer and project manager with an expertise in saving failed projects. I have refactored entire projects written in Python/Perl/VB into C/C++. It is a major pain but in systems that do not perform and have pervasive execution speed problems it is always the best solution.

---

Quote:

---

CCP is likely trying their best to profile their running code but the simple fact is they cannot put sufficient performance measuring yools on the live server without drastically affecting performance and testing on Sisi will likely never find the problem if they haven't found it already. Sisi is simply too different from TQ to be a reasonable test environment.

---

Quote:

---

Therefore one thing to do is to refactor the whole codebase to gain all the performance gains you can get. If CCP doesn't have the people in house who can do that job there are hundreds of coders in the US and Europe out of work who can. This isn't a job for guys who are primarily graphics or game coders. Eve is fundamentally a mission critical distributed real time application and the only things roughly equivalent are some of the weather sim supercomputers and the high volume financial world. Those are the guys I'd be hiring.

---

Originally by: Liang Nuren

---

#include <fstream>

using namespace std;

int main(int argc, char ** argv)
{
 ofstream of("output_file.txt", ios::out);
 for (int i = 0; i < 1000000; ++i)
 of << i << endl;
 of.close();
 return EXIT_SUCCESS;
}

---

Originally by: pmchem

---

from http://docs.python.org/dev/whatsnew/2.7.html

http://bugs.python.org/issue4074 scalability improvement in list building
http://bugs.python.org/issue4688 garbage collection (memory management) speedup
http://bugs.python.org/issue5084 pickling memory+speed optimization
http://bugs.python.org/issue5670 pickling speedup when dealing with dicts
http://bugs.python.org/issue3582 windows-specific threading optimization

or maybe it's just groundwork for multicore programming since all the python parallel programming related toys tend to do their development with recent releases of python, e.g. http://mpi4py.scipy.org/

---

Originally by: Dianabolic

---

Your tears are absolutely divine, like a fine fine wine, rolling down your cheeks until they flow down the river of LOL.

Originally by: Liang Nuren

---

What you just did is normally called a "straw man". You constructed a position that I did not take, and then attacked it. Then you attributed it to me, and called my position wrong. To quote you: "That's just stupid."

The position that I actually took was to show that good design is far more important than writing something in a "more performant language" and "close to the machine code". The number of people getting stuck on this point is utterly staggering.

---

Originally by: Hurtado Soneka

---

erm so what is python anyways? :D

---

Originally by: Liang Nuren

---

Edited by: Liang Nuren on 18/08/2010 05:45:28

Originally by: HenkieBoy

---

Oracle being the only real other option, they suck. Their product may be good but they are really bad for any company. They ask bucket loads of money for even the most trival options.

---

Excuse me, but PostgreSQL and many other open source (free!) databases would like a word with you. And yes I know for fact that you can achieve amazing performance out of a single instance Postgres DB

-Liang

Ed: Um. I decided that I would delete part of my post. It is sufficient to /fanboi without going over the top with technical details.

---

Originally by: CCP Dropbear

---

rofl

edit: ah crap, dev account. Oh well, official rofl at you sir.

---

Originally by: Barakkus

---

The only other possibilty for them is Oracle, and well that's a bloated piece of crap.

---

Originally by: Hemp Invader

---

have you tried:
register int i;
for(i=1000000; i>0; i--)
{
....;
}
?
comparison with 0 is better, the i value always in your register if it has room...

If only your coding skills would match your bragging skills...i'm just sayin'

---

Originally by: Barakkus

---

I work with postgresql, mysql and oracle, yes you can get good performance out of postgresql, that's why we migrated to it, but you can not cluster postgresql, and postgresql will not handle the load ccp puts on their database, plain and simple.

---

Quote:

---

postgresql 9

---

Originally by: CCP Fallout

---

More information regarding Python 2.7 will be coming in a blog early next week, actually.

---

Originally by: Liang Nuren

---

Originally by: Barakkus

---

I work with postgresql, mysql and oracle, yes you can get good performance out of postgresql, that's why we migrated to it, but you can not cluster postgresql, and postgresql will not handle the load ccp puts on their database, plain and simple.

---

Meh, I've put together a postgres cluster before - but honestly it was maintained in the application. There's also multiple commercial implementations, but they simultaneously didn't quite handle some of our custom code and were bordering on Oracle expensive.

Quote:

---

postgresql 9

---

Yeah, its sounding pretty awesome. :swoon:

-Liang

---

Originally by: CCP Dropbear

---

rofl

edit: ah crap, dev account. Oh well, official rofl at you sir.

---

Quote:

---

A poster above mentioned DB2, I honestly hadn't even considered DB2 when we migrated from Paradox to PostgreSQL (6 years ago), I went and checked it out, it actually looks pretty damn sweet...I figured you still had to run it on some sort of IBM OS, but it appears to work cross platform now a days.

---