Legend In Plots #119
Legend In Plots #119
Conversation
|
On faceted charts, do we want the legend under each plot? By default, if not under each plot, the legend appears all the way at the bottom of the faceted plots. 
|
|
I prefer, in order:
|
|
The problem I am having is when I do independent 
I tried querying AI somewhat on this but it wasn't able to help really and propered solutions that did not work / were seemingly not sensible. I will keep at it. |
The legend doesn't really add much to the plot, in my opinion. I think we should abandon the PR, if it is proving difficult. |
|
I don't think there is any way for the transparencies to match the legend. By definition the 50% CI is overlayed on top of the the 90%, etc. The legend will have no concept of this. We could swap to different colors for each band to resolve this. To mimic the R plots, use |
Sounds great. This commit 7185da2 was clean and had Forecast in steelblue wo/ actual CI labels, but what you just suggested sounds better. |
|
@AFg6K7h4fhy2 I suggest getting the colors programmatically with >>> import colorbrewer
... colorbrewer.Blues[3]
[(222, 235, 247), (158, 202, 225), (49, 130, 189)] |
hubverse_annotator/utils.py
Outdated
| labels = [f"{round((high - low) * 100)}% CI" for low, high in ci_pairs] | ||
|
|
||
| palette_rgb255 = list( | ||
| colorbrewer.Blues[max(3, min(MAX_NUM_CIS, len(labels)))] |
There was a problem hiding this comment.
MAX_NUM_CIS is not used to limit ci_pairs. Seems like an oversight.
hubverse_annotator/utils.py
Outdated
| palette_rgb255 = list( | ||
| colorbrewer.Blues[max(3, min(MAX_NUM_CIS, len(labels)))] | ||
| ) | ||
| palette = [(r / 255, g / 255, b / 255) for r, g, b in palette_rgb255] |
There was a problem hiding this comment.
This approch, using rgb_strings, feels more straightforward than division + hex conversion.
import altair as alt
import polars as pl
import colorbrewer
rgb_tuples = colorbrewer.Blues[3]
rgb_strings = [f"rgb({r},{g},{b})" for r, g, b in rgb_tuples]
df = pl.DataFrame({
'x': [1, 2, 3],
'y': [4, 5, 6]
})
chart = alt.Chart(df).mark_point(
color=rgb_strings[2],
size=100
).encode(
x='x',
y='y'
)
chart
hubverse_annotator/utils.py
Outdated
| "low": f"{low:.3f}".rstrip("0").rstrip("."), | ||
| "high": f"{high:.3f}".rstrip("0").rstrip("."), | ||
| "color": to_hex(color), |
There was a problem hiding this comment.
Again, let's use format_percent from babel.
| bands = [ | ||
| band("0.025", "0.975", 0.10), | ||
| band("0.1", "0.9", 0.20), | ||
| band("0.25", "0.75", 0.30), | ||
| band(spec["low"], spec["high"], label) | ||
| for label, spec in ci_specs.items() | ||
| if spec["low"] in df_wide.columns and spec["high"] in df_wide.columns |
There was a problem hiding this comment.
Structure overall feels a bit off.
What about
- We have
get_available_ci(list_of_quantiles)which returns all of the available ci's (width only, or width, lower, and upper) - Then somewhere later we have the availability to subsample the available ci's (either by giving the list of widths we want or by limiting to some max number).
- Then we get the colors based on the length of the subsampled list of ci's.
- Then we make the plot with legend.
There was a problem hiding this comment.
I will try this out.
There was a problem hiding this comment.
"Get the colors" step may be obviated by https://vega.github.io/vega/docs/schemes/#blues
hubverse_annotator/utils.py
Outdated
| from typing import Literal | ||
|
|
||
| import altair as alt | ||
| import colorbrewer |
There was a problem hiding this comment.
Rather than colorbrewer, we should try to take advantage of https://vega.github.io/vega/docs/schemes/#blues
hubverse_annotator/utils.py
Outdated
| ] | ||
|
|
||
| num_colors = max(3, min(MAX_NUM_CIS, len(labels))) | ||
| blues_map = colormaps["Blues"] |
There was a problem hiding this comment.
I thought the plan was to use vega's built-in implementation of the blues palette. Is there some reason that doesn't work? https://vega.github.io/vega/docs/schemes/#blues
3 participants
For the full scope of this PR, please refer to issue #117 .
Specifically, this PR adds: