Status of Python integration

Yoghurt · **Posted:** Sun Aug 31, 2003 10:55 am

Yesterday I finally managed to take some time working on FO, the Python part mainly. After experimenting and trying to understand the bad (IMHO) documentation of Boost.Python, I finally managed to get started. I exposed TechLevel and TechManager to Python, here is an sample run:

Code:

>>> from balla import *
>>> TM = TechManager.instance()
>>> TM.LoadTechTree("dummy")
1
>>> for tech in TM:
...     print tech.GetName(), tech.GetMinPts()
...
Defense Base 350
Mark II 250
Mark III 750
Mark IV 1650

Right now, I mirror the C++-interface, but later on I will do it in a more "pythonic way", for example, TM=TechManager.instance() would not be neccesary, TechManager could be used instead. And tech.GetName will of course become (read-only) tech.Name.
I will work on a more complex demo how python can help us. For now, I think I will code the System-generation

jbarcz1 · **Posted:** Sun Aug 31, 2003 5:16 pm

I thought that we were only thinking of using python to implement logic things that we want to be able to tweak (like effects of technology, specials, etc) Why do system generation in python? Particularly when it's already coded in C++ (for irregular galaxies)

Yoghurt · **Posted:** Sun Aug 31, 2003 6:48 pm

jbarcz1 wrote:

I thought that we were only thinking of using python to implement logic things that we want to be able to tweak (like effects of technology, specials, etc) Why do system generation in python? Particularly when it's already coded in C++ (for irregular galaxies)

Yes, it would be only an example. I thought of using system generation so we all have a comparison between how things are done in Python and what the relevant C++-code would look like

Yoghurt · **Posted:** Wed Sep 03, 2003 9:54 pm

Current status: stalled (again)

This stupid ReiserFS randomized it's tree structure again. I will have to get a second hard-disk to do a backup before I can try to repair the filesystem.

This might take a little time, sorry.

Tyreth · **Posted:** Sat Sep 06, 2003 3:52 pm

I personally am still very keen to see python integrated - or at least evaluated fairly once we can see first hand how it will benefit us.

That's why I'd like to not lay this issue to rest yet until after Yoghurt completes his demo. I'm currently working through http://www.diveintopython.org and it's looking pretty good.

Yoghurt · **Posted:** Sun Sep 07, 2003 12:03 am

My computer works again!

I will now continue to work on the Python thing. Suggestions are welcome.

tzlaine · **Joined:** Thu Jun 26, 2003 1:33 pm **Posts:** 1092

My only suggestion is ext3fs. Heh.

Tyreth · **Posted:** Sun Sep 07, 2003 6:20 am

tzlaine wrote:

My only suggestion is ext3fs. Heh.

Bah, Yoghurt must be incompetent! reiserfs never hurt anyone

I give a thumbs up to journalling in either fs though, very nice. Especially with the crashing my new video card gives

Yoghurt · **Posted:** Sun Sep 07, 2003 12:23 pm

ReiserFS is more stable than ext2, and IRRC, also faster.

It is also very hard to get Reiser to data corruption, but if it happens, it's really f*cked up, I needed to do a rebuild-tree 3 times till I have installed Reiser, and 2 of the 3 times I "lost" about half of my data (which was backuped of course)

Zach, I will implement ext3fs in Python for FO, but will we really need it? :twisted:

Tyreth · **Posted:** Sun Sep 07, 2003 1:24 pm

Yoghurt wrote:

ReiserFS is more stable than ext2, and IRRC, also faster.
It is also very hard to get Reiser to data corruption, but if it happens, it's really f*cked up, I needed to do a rebuild-tree 3 times till I have installed Reiser, and 2 of the 3 times I "lost" about half of my data (which was backuped of course)

What is a rebuild-tree? I've never used that with reiserfs. However, I have noticed something disturbing a couple of times - XFree has crashed, and my XF86Config file has been completely borked. Almost never happens - has happened twice. Also vim config file was borked once.

ReiserFS is supposed to be a fair bit faster with filesystems containing lots of smaller files, iirc. From the glance of a beginner's eyes at the future of reiserfs, looks pretty exciting.

Yoghurt wrote:

Zach, I will implement ext3fs in Python for FO, but will we really need it? :twisted:

Hmm, that gives me an idea. Why don't we just make FO as a plugin for Emacs? I'm sure there's some code in there we could build off :twisted:

Otherwise, we could have FO compete with Emacs.

tzlaine · **Joined:** Thu Jun 26, 2003 1:33 pm **Posts:** 1092

Mmmm, Emacs... [drools on self].

Yoghurt · **Posted:** Mon Sep 08, 2003 12:34 pm

tyreth wrote:

Hmm, that gives me an idea. Why don't we just make FO as a plugin for Emacs? I'm sure there's some code in there we could build off :twisted:

Otherwise, we could have FO compete with Emacs.

Either that, or we could you Emacs-like key-combinations. I can alreay imagine the first page of the FreeOrion Tutorial.

Quote:

First of all, we will have to select a scout ship to explore. We achieve this by Pressing C-M-X M-T Y z (thats a small z!) C-U S 1.
... To quit the game, simple press C-x C-f C-q q y RET

Maybe we can get the FSF-Award!
:roll:

Yoghurt · **Posted:** Wed Sep 10, 2003 12:19 pm

Status: on-going.

Yesterday I finally managed to checkout the CVS version of Pyste (the wrapper-generator) which works better than the version in boost 1.30.0, however, I stumbled upon some misterious compilation errors that lead to an internal g++ compiler error :-(

As you might now, the error messages that are produced when STL and other template-heavy stuff (which Boost.Python is) are no more infomative than line noise (I got around 20K of error messages)

Nevertheless, I'm confident that Python-integration will reach alpha in a few weeks, hopefully less.

Yoghurt · **Posted:** Sun Sep 21, 2003 1:43 am

Update: with the new build system. Python should fit in nicely.

It still will take some time before the alpha (sorry, sorry, sorry), but the FreeOrion files are a stress-test for the automatic wrapper generators. I stopped counting the "internal compiler errors" of GCCXML.

But it'll be done soon.

PowerCrazy · **Posted:** Tue Sep 23, 2003 1:35 pm

Hey Yoghurt,

This weekend (after all my tests etc.) I'll have a bit of free time, and wanted to start on some primitive colonizing AI. I'll be doing it in python and so i need to know how to structure the program. It will need 2 inputs, one is easy, The number of stars. the second one is more difficult. It needs to know which stars are connected by starlanes to which other stars. The output will consist of 2 NxN matricies where N is the number of stars.

The idea is to create an adjacency matrix that will give the AI a way to tell distance (in starlane hops) between any two stars on the galaxy map as an aid in colonizing and eventually attack effectively. This module coupled with a set of weights or something similiar ought to be pretty good until races are integrated and the tech tree is expanded. Just give me an outline to follow and i'll (try) to take care of the rest.

FreeOrion

Status of Python integration

Who is online