Agent size change based on size of figure #1820
Conversation
| def calculate_space_size(): | ||
| pass | ||
|
|
||
| def calculate_agent_size(): |
There was a problem hiding this comment.
The Matplotlib marker size shouldn't depend on the number of agents. It should depend only on the grid size and the computer's screen size.
There was a problem hiding this comment.
It should be more abstract, like screen width, because in a Jupyter notebook, the width should be even smaller.
There was a problem hiding this comment.
hmm, let me think how this can be implemented. Currently the size of each agent is parsed via agent_portrayal and hence it's easier(also less compute expensive) to manipulate.
There was a problem hiding this comment.
I think it would also be important to show all agents in a smaller window size, hence the implementation can work there. I try to come up with some scaling to fit the screen width.
|
Thank you for starting to work on this 👍 |
|
I have restructured the UI layout using some of the features by Solara. The sidebar can be minimised to expand the user view: 
|
|
That looks nice. We could adaptively change the layout by detecting whether the Solara component is run in a Jupyter notebook or not: https://stackoverflow.com/questions/15411967/how-can-i-check-if-code-is-executed-in-the-ipython-notebook. But for this PR, you should focus on solving the grid render for #1806. The layout restyling should be in a separate PR. |
|
I have now updated to change the size of the figure based on number of agents. This is possibly the easiest way. This avoids agent markers to come close as the number of users increase. Changing the size of the markers itself does not look neat, although its possible. Sure I can move future layout modifications to another PR. |
|
To put it concisely:
|
ankitk50
force-pushed
the
agent_size
branch
from
1bae820 to
5ff171a
Compare
|
Thanks for the explaination, I can try this out. Meanwhile UI related changes is moved to PR #1825. |
ankitk50
changed the title
|
The change maintains the size of the agents based on size of the figure. |
| aspect_ratio_low = 0.1 # | ||
| print() | ||
| size_of_agent = aspect_ratio_high * ( | ||
| model_parameters["width"] * model_parameters["height"] |
There was a problem hiding this comment.
(Must be GitHub hiccup. I remembered already commenting about this) the size is length^1, but the width * height is length^2, so it scales quadratically instead. It makes more sense to scale linearly over min(width, height).
There was a problem hiding this comment.
yes, the idea is to have a constant ratio b/w size of agents (for e.g. area of a circle) and size of the figure (area of a square). Hence the scaling is still linear i.e. the size of agent = aspect_ratio * size of figure.
There was a problem hiding this comment.
I just realized that the s param in Matplotlib scatter is length^2: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.scatter.html. It's markersize**2 without a multiplication by pi. We should document that size refers to Matplotlib size, and it is an area.
mesa/experimental/jupyter_viz.py
Outdated
| plt.rcParams["figure.figsize"] = (fixed_size, fixed_size) | ||
|
|
||
| def calculate_agent_size(): | ||
| aspect_ratio_high = ( |
There was a problem hiding this comment.
Definition of aspect ratio is width/height: https://en.wikipedia.org/wiki/Aspect_ratio_(image). fill_fraction is a more descriptive term.
There was a problem hiding this comment.
sure, makes more sense, thanks!
| @@ -51,6 +51,23 @@ def make_model(): | |||
| set_current_step(0) | |||
| return model | |||
|
|
|||
| def calculate_space_size(): | |||
| if model_parameters["N"] > 50: | |||
There was a problem hiding this comment.
Once again, the grid size is independent of the number of agents. Having few agents and having many agents should result in the same marker size and figure size.
There was a problem hiding this comment.
Yes, agents less than 50 the figure size of default used by matplotlib. I start to scale the it more num of agents.
Ideal would be scale the image size based on screen size, but this doesn't seem feasible. Any other ideas?
There was a problem hiding this comment.
I start to scale the it more num of agents.
Number of agents shouldn't have effect on image size for the grid.
|
Really like this idea! What's needed to move this forward? |
|
|
Secondary reason for #1825 (comment) is that it is not thread safe. Anything that is global operation via |
|
@ankitk50 can we help you move this forward? |
|
Superseded by #2049. |
If the number of agents in greater than 50, I use a linearly decreasing function to calculate the size of the agent. The lower limit for the agent size is 5.