Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Mego]

Thanks, Mego, to_string worked perfectly :3 Much better than having to write my own function that carefully dissected ints into vector output.

Welcome!

[Nadaka]

Ha! Maybe a weekend of writing unit tests wasn't so bad after all.

After an intense 4 hour refactor/ feature add session, I broke 2 of my 32 unit tests, but was able to identify the single source of the error within minutes.

Now I just have to get my code coverage up another 15% to cover the new code.

Then I can refactor the tokenizer. And write more unit tests.

Then I can refactor the parser. And write more unit tests.

Then I can get bored and forget that I was originally trying to write a building generator.

And I can forget that that generator might have been the first step in providing procedural content for an as of yet undetermined video game.

Then I can get depressed and do nothing but work and watch netflix.

[Skyrunner]

I was going to try and learn to use tinyXML, but then I was like "nah, I don't need this right now anyways, so I'll learn it when I need it."

A bit of searching around, and I have a question. What's the difference between Boost's Random header, and good ol' rand()?

Also, what might be a good way to do weighted drawing thingies? Such as having a higher chance of drawing a certain number than other numbers?

[MorleyDev]

Random allows for more sophisticated or appropriate algorithms to be used, rand isn't random enough for a lot of uses. It's typically just a simple LCG implementation whilst random allows for say Mersenne's twister to be used. :-)

Also it allows for multiple sets of random number generators to be used, which helps with keeping code modular and testable, the next random number from a specific seed can be predicted for your physics system without having to worry about the audio system mucking up the order by getting a random number.

[Skyrunner]

Hmmmm, is it hard to install? I've heard Boost's libraries tend to have tons of complicated dependencies. Also, last time I tried to use a library I ended up messing it up and doing nothing with it at all.

As an aside, is XML good for storing data? Also, is TinyXML 2 a good header file for it? O_o

[Nadaka]

XML is "good" in that it is standard and has a lot of existing tools built for/around it. It is also clunky and bloated.

XML is good if you need structured trees of data and expect to share them with other developers/programs, particularly public API's accessible over the internet. It is either bad, or not very good at everything else.

My annoyances with XML is about half the reason I decided to play around developing a lightweight alternative MNML. Note: not advocating mnml over xml at the moment, mnml is still a ways away from being useful to anyone other than myself.

[Mego]

Hmmmm, is it hard to install? I've heard Boost's libraries tend to have tons of complicated dependencies. Also, last time I tried to use a library I ended up messing it up and doing nothing with it at all.

As an aside, is XML good for storing data? Also, is TinyXML 2 a good header file for it? O_o

Boost isn't terribly difficult to install under Linux or Cygwin. Also, if you want to use the extra-special random stuff, install gcc 4.3.4. Boost::Random is in the standard library in C++11.

[MorleyDev]

Boost isn't terribly difficult to install under Linux or Cygwin. Also, if you want to use the extra-special random stuff, install gcc 4.3.4. Boost::Random is in the standard library in C++11.

..probably want to use a newer one than that xD Get the fun stuff like variadic templates and nullptr to play with then >:-D What's it at now, GCC 4.7? I personally use http://nuwen.net/ for GCC for Windows since it's typically quite up-to-date. Plus it also comes with the latest (or near-latest) versions of Boost and what-not pre-installed

I did make some further modifications, copied libboost_thread.a and renamed it to libpthread.a and rewrote the standard thread-related headers to pull in boost threads and mutexes into the std namespace (adding some extra functionality boost threads was(is?) missing like interoperability with std::chrono) xD That way I can pretend I'm using the new C++11 threading library in Windows And more importantly don't need to change any settings in Eclipse's project to recompile on Linux

[Mego]

4.3.4 has variadic templates and nullptr. I can't get anything newer than that to work in Cygwin properly.

[MorleyDev]

Erm, as far as I know nullptr wasn't supported until 4.6 of GCC (to the point that I specifically remember writing my own implementation for cross-compiling between VS2010 and GCC 4.5) :S Visual Studio 2010 supported it surprisingly long before GCC did. And 4.3.4 only supports the wip version of the specification for variadic templates, full support is in 4.4.

[Mego]

Erm, as far as I know nullptr wasn't supported until 4.6 of GCC (to the point that I specifically remember writing my own implementation for cross-compiling between VS2010 and GCC 4.5) :S Visual Studio 2010 supported it surprisingly long before GCC did. And 4.3.4 only supports the wip version of the specification for variadic templates, full support is in 4.4.

My bad. I'm gonna bitch at some people for not making a GCC 4.7 package for Cygwin yet.

EDIT: Turns out I actually have 4.5.3. Still not 4.7.

[Skyrunner]

So... are you saying boost.random is easy to use in Visual C++ 2010? >.>
Also, what in the world is a nullptr >.>

[MadocComadrin]

It's a null pointer...or do you what the whole explanation?

Edit:

Spoiler: Full Explanation (click to show/hide)

[Mego]

So... are you saying boost.random is easy to use in Visual C++ 2010? >.>
Also, what in the world is a nullptr >.>

Nothing is easy to use in VC++. That's a flaw with the IDE, not the library.

A nullptr is like the NULL macro of standards past, only it can't be treated as an integral type. Along with some other fancy stuff that I don't care to look up right now.

[MorleyDev]

Turns out that in SQL that ROW_NUMBER() degrades majorly in performance after you get past 250000 rows. This makes implementing a paging system for just one user's data something of a bitch with large data sets...

Update: Got it to run a query on a massive data-set getting the tail end of over 700000 (so say getting the items at the row positions between 700100 and 700200) and items in about 20 seconds. Some more optimisations could be made (we have to get the total count for the user anyway, so could possibly inverse the ordering so it only needs to search about half the data (Well "half + c / 2" where c is the number of items being retrieved, in the worst case) but even then, damn huge SQL databases can be bitches to work with.