Update abstract.jl solves #1045

[amynang]

The culprit behind #1045 was the use of poly! for user defined polygons. As a first step, I replaced it with scatter!(p, p.pos; p.color, p.marker, p.markersize, p.agentsplotkwargs...) and generating the plot with the custom polygon worked. In that case I think the whole if else statement is no longer necessary so I removed it.

if user_used_polygons(p.agent_marker[], p.marker[])
poly!(p, p.marker; p.color, p.marker, p.markersize, p.agentsplotkwargs...)
else
scatter!(p, p.pos; p.color, p.marker, p.markersize, p.agentsplotkwargs...)
end

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 86.53%. Comparing base (8b5b456) to head (d902e39).
Report is 125 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1059       +/-   ##
===========================================
+ Coverage   70.12%   86.53%   +16.41%     
===========================================
  Files          42       39        -3     
  Lines        2718     2578      -140     
===========================================
+ Hits         1906     2231      +325     
+ Misses        812      347      -465     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[amynang]

I just realized that while this change removes the error that abmvideo was throwing, the end result is not quite right; the generated video seems to show only part of the space. This is despite the fact that the x and y axes suggest that the dimensions are correct. I make a model with 100 agents but see only few of them and it is also clear that they don't reappear where you'd expect them to given a periodic space. I do not understand what is going on here.

[Tortar]

I didn't test anything but I feel like removing poly! shouldn't be the solution, what does the error say when you try to create the video with the previous code?

[Tortar]

ah also, thanks for your help on this :-)

[Datseris]

unfortunately i have to agree, removing poly! also doesn't sound like a solution to me. Does the scatter! actually plot the polygons..? Does the flocking video looks like flocking, with triangles moving and rotating around? If not, then we haven't solved the problem.

I suspect that this code needs to stay as is, with if:

if user_used_polygons(p.agent_marker[], p.marker[])
poly!(p, p.marker; p.color, p.marker, p.markersize, p.agentsplotkwargs...)
else
scatter!(p, p.pos; p.color, p.marker, p.markersize, p.agentsplotkwargs...)
end

but what needs to be done is that we cannot blindly propagate p.agentsplotkwargs anymore. There are some keywords that poly! doesn't accept that are otherwise valid for scatter!, so we have to filter them out.

[amynang]

@Tortar here is the full error message with the original code. This is the model where I encountered the problem.

@Datseris scatter! does actually plot the polygons. The first picture is a screenshot with the PR code, the next one is with the agent_marker argument in abmvideo commented out. This is the strange behavior I described earlier, both cases have 100 agents but the first looks zoomed into a part of the space for some reason.

[Datseris]

agent_marker argument in abmvideo commented out

Don't do that, it doesn't make sense at the moment. So we don't have to spend resources figuring out what commenting this out does.

I didn't know that scatter! plots polygons. Then we can just use scatter! for everything. However, let's make sure about one more thing: if you increase the polygon size, for example by multiplying the polygon coordinates with some factor

const bird_polygon = Makie.Polygon(5 .* Point2f[(-1, -1), (2, 0), (-1, 1)])
function bird_marker(b::Bird)
    φ = atan(b.vel[2], b.vel[1]) #+ π/2 + π
    rotate_polygon(bird_polygon, φ)
end

(notice 5 .*)

Are the polygons larger in size in scatter!? Does it respect the actual size of the polygons?

[amynang]

Yep, polygons get larger

[Tortar]

I tried your model, and with the last commit I pushed everything works fine, we don't propagate to poly! anymore markersize, but maybe it never worked? From the code of @Datseris it seems so because you define the markersize when creating the polygon

[amynang]

Great! Thanks a lot!

[Datseris]

Alright, please go into top-level Project.toml and increment the patch version (3rd number) by one. Then we can merge and immediatelly tag a new release.