[Freeciv-Dev] Re: [RFC] Path finding interface

["Ross W. Wetmore" <rwetmore@xxxxxxxxxxxx>]

At 04:44 PM 02/04/11 +0100, Gregory Berkolaiko wrote:
>On Thu, 11 Apr 2002, Ross W. Wetmore wrote:
>> 
>> You've got it mostly right here. There is some 6 month old code
>> kicking around in GB's early prototypes that does most of this.
>> 
>
>I have given up any hope of pushing it through on my own.  I don't think 
>any maintainer will have the gut to commit it in any shape.  If even a 
>10-liner waits 2 months before being looked at, updated and tagged "with 
>changes by me". ;)

As Tony once said ... such comments reflect a sad state of affairs.

>But if Raimar writes the warmap code, commiting will come much easier. ;)  
>And I will try to help him avoid the pitfalls that are known to me and 
>various performance bombs that he is trying to create.  My name won't be 
>on the patch but I can live with it. :)

When in Nero's Rome, enjoy the fires, dance through them when commanded ... 

>I hope that in fights with him a truly resillient code will be born :)
>And the setup is almost ideal: he knows the agent and client demands and I 
>know the AI and partly client demands.

The possiblities have always been really appealing.

>BTW, Ross, the examples you quoted were not all written by me.  Did you 
>use our warmap code in other places too?  BTW, it's not really "GB's" 
>patch, as the backbone of the patch came from the warmap.c file you sent 
>me a while ago.

The examples are just a rewrite of your 3 line while block as a for. The
dynamic warmap code behind it is your earlier CVS submission (probably v.1 
or 2) from last fall that went into the corecleanups at that time and has
rotted there ever since without being really used.

I was chasing down goto_is_sane() (it isn't) issues and rewrote a lot
of search code to just use warmaps as a sideline to amuse myself. For
example Parts II and III of manage_explorer is a single warmap loop about
equivalent to the huts example. I think I suggested you do this in one
review comment, thought I'd try my own medecine at some point.

BTW, one of the reasons why exploring types don't head into the unknown 
is that if goto_is_sane() (once fixed) is used to select target squares
you aren't allowed to select unknowns for land moving. Thus you skirt
the edges :-(. Also, gotohand.c: do_unit_goto() is completely busted
if you fix goto_is_sane(). Exploreres and city founders really sort of 
depend on being able to go into unknown territory, and the AI really
doesn't have a good local explorer program to map immediate sites, even
if the goto_is_sane() known_AND_SEEN() condition let them be visible
long enough to get there :-).

As for warmap.c, it is probably on a CD backup disk somewhere. but the
bucket and local warmap concepts are all that are really useful in it
compared to your subsequent submissions and rewrite efforts.

If you want to see this stuff live, I'll probably upload a CVS resync of
the corecleanups by end of next weekend.

>Best wishes,
>G.

Cheers,
RossW
=====