[Ohrrpgce] SVN: james/8551 Clean up perform_npc_move() refactoring the collision detection for wall

[Ralph Versteegen]

On 20 March 2017 at 15:51, James Paige <Bob at hamsterrepublic.com> wrote:

>
> On Fri, Mar 17, 2017 at 9:46 PM, Ralph Versteegen <teeemcee at gmail.com>
> wrote:
>
>> There's no need for the npc_collision_check overload without the
>> hero_collision_exception argument, because FB lets you make byref arguments
>> optional, by giving them a default value. That results in a temporary
>> variable getting initialised to the default value and passed to the
>> function. I do this in several places. One of those little things I like
>> about FB.
>>
>
> Cool! I'll fix that. I didn't realize (or had forgotten that) I could do
> that :)
>
>
>> I see you made a small back-compat breaking change here, by changing the
>> determination of NPC tile X/Y position for the purpose of wall/zone
>> checking, using the center rather than the top-left of the walkabout. Of
>> course, this only happens when the NPC finishes a step, so there's only a
>> difference if the NPC is mis-aligned, in which case collision checking is
>> quite broken anyway. I wonder, does any game actually rely on that
>> brokenness? It seems unlikely actually. I tested Siv: journey of an egg and
>> didn't notice any problems.
>>
>
> I am slowly working my way up to fixing the getting-stuck-mid-tile bug.
>
> Even though it is technically backcompat breaking, we have not yet
> identified a single game that relies on the bug. If we ever find one, I'll
> dust off the old code, wrap it in a backcompat bitset, and turn that bitset
> on with some kind of IF gamename = "game_that_needs_to_be_broken.rpg"
> check
>

OK, I guess we can worry about it later. Enabling backcompat bits for
specific games would be a whole new thing. In some cases that would be the
best way to handle back-compat problems. I think we can come up with better
ways than predicating on the file name though.


>
>>
>>
>> Actually I don't think that change to using the tile center helps at all;
>> it's still going to be broken. I suggest fixing misalignment by walking
>> NPCs/heroes less than 20 pixels (whether by walknpc or builtin movement) so
>> that it realigns at the next tile boundary, and doing collision checking
>> like a pixel-base-moving NPC/hero until then.
>>
>
> Yeah, I know that doesn't fix anything, I still figured it was more
> correct to determine which tile's walls to check based on the center rather
> than the top left corner.
>
> Yes, that does sound like a good plan for fixing misalignment.
>
>
>> Script commands would be great. I think there are two approaches: high
>> level "do NPC/hero obstacle checking exactly like the engine does" that
>> might be useful for scripting eg pathfinding, and could be affected by a
>> pixel-based-movement option, or low level "check for collisions with
>> walkabouts" that would be a building block for scripting your own movement.
>> So there could be commands for both, if they would actually get used. But
>> maybe the latter would not get used enough to bother... I know recently
>> I've said several times "use the wall checking commands and do your own
>> collision checking with 'find colliding slice'", but have have to look into
>> those specific use-cases actually were.
>> Plus, unfortunately the wall checking commands I added recently don't
>> support NPC zones or one-way walls.
>>
>
> The other thing I am doing (besides chipping away at the tile misalignment
> bug) is to add a couple new NPC movement types.
>
> Specifically I want a less random version of Chase and Avoid. I have one
> that is *almost* working but I still need to clean up some edge cases and
> test it better.
>
> I also definitely want a movement type for follow-right-hand-walls and
> follow left-hand-walls which will allow for some pretty fancy NPC movement
> using just zones and no scripts.
>
> I have also been thinking about full-on A-star pathfinding. but won't dare
> touch that until I have the other stuff working.
>

Looking forward to it :)

I guess you'll recompute the A* path each step rather than hanging onto it?
For exposing to scripting, this is something that wants arrays and garbage
collection.


>
> ---
> James
>
>
>
> _______________________________________________
> Ohrrpgce mailing list
> ohrrpgce at lists.motherhamster.org
> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.motherhamster.org/pipermail/ohrrpgce-motherhamster.org/attachments/20170320/34333b57/attachment.htm>