A possible workaround:
set the range element for both graphs to be proportional to each other then the axes will line up. Basically, your problem is one axis has to show negative numbers and the other doesn't. By telling y2 to show negative numbers we obtain our goal.
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
import plotly.graph_objs as go
init_notebook_mode(connected=True)
traces = [
go.Bar(
x=[1, 2, 3, 4],
y=[-1, 2, -3, 4],
name='y actual'
),
go.Bar(
x=[1],
y=[0],
name='y dummy',
hoverinfo='none',
showlegend=False
),
go.Bar(
x=[1],
y=[0],
yaxis='y2',
name='y2 dummy',
hoverinfo='none',
showlegend=False
),
go.Bar(
x=[1, 2, 3, 4],
y=[22, 2, 13, 25],
yaxis='y2',
name='y2 actual'
)
]
# layout
layout = go.Layout(
barmode='group',
yaxis=dict(
title='y actual',
rangemode="tozero",
#anchor='x',
overlaying='y2',
side="left",
range = [-4, 10]
),
yaxis2=dict(
title='y2 actual',
side='right',
rangemode = "tozero",
#anchor='x',
range = [-12, 30]
)
)
# make fig
fig = go.Figure(data=traces, layout=layout)
iplot(fig)
This may be annoying to have to keep them in proportion but it will ensure their alignment.
to help automate the process you can use the following function to produce two ranges that are proportional to each other.
def make_proportional_intervals(a, b):
"""
Given two list like objects, compute two proprotionally sized ranges.
This function assumes the max value in both lists is positive and non-zero
"""
min_a, min_b = min(a), min(b)
max_a, max_b = max(a), max(b)
if (min_a >=0) & (min_b >= 0):
# provide a 20% cushion to the scale
return [0, round(1.2*max_a)], [0, round(1.2*max_b)]
else:
if (min_a < min_b) & (max_a < max_b):
n = -(-max_b // max_a)
# n = math.ceil(max_b / max_a), if you cannot assume ints.
return [min_a, max_a], [n*min_a, n*max_a]
elif (min_b < min_a) & (max_b < max_a):
n = -(-max_a // max_b)
# n = math.ceil(max_b / max_a), if you cannot assume ints.
return [n*min_b, n*max_b], [min_b, max_b]
elif (min_b < min_a) & (max_a < max_b):
n = max( -(-max_b // max_a), -(min_b // min_a) )
return [min_b / n, max_b / n], [min_b, max_b]
elif (min_a < min_b) & (max_b < max_a):
n = max( -(-max_a // max_b), -(min_a // min_b) )
return [min_a, max_a], [min_a / n, max_a / n]
elif (min_a == min_b):
m = max(max_a, max_b)
return [min_a, m], [min_b, m]
elif max_a == max_b:
m = min(min_a, min_b)
return [m, max_a], [m, max_b]
This function assumes your values will be integers but if not you can import math and use math.ceil() instead of my integer division. I was avoiding adding any more imports. If you want to see this code in action I created an example in jupyter notebook that you can run multiple times to see how it handles different arrays.
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
import plotly.graph_objs as go
import numpy as np
def make_proportional_intervals(a, b):
"""
Given two list like objects, compute two proprotionally sized ranges.
This function assumes the max value in both lists is positive and non-zero
"""
min_a, min_b = min(a), min(b)
max_a, max_b = max(a), max(b)
if (min_a >=0) & (min_b >= 0):
# provide a 20% cushion to the scale
return [0, round(1.2*max_a)], [0, round(1.2*max_b)]
else:
if (min_a < min_b) & (max_a < max_b):
n = -(-max_b // max_a)
# n = math.ceil(max_b / max_a), if you cannot assume ints.
return [min_a, max_a], [n*min_a, n*max_a]
elif (min_b < min_a) & (max_b < max_a):
n = -(-max_a // max_b)
# n = math.ceil(max_b / max_a), if you cannot assume ints.
return [n*min_b, n*max_b], [min_b, max_b]
elif (min_b < min_a) & (max_a < max_b):
n = max( -(-max_b // max_a), -(min_b // min_a) )
return [min_b / n, max_b / n], [min_b, max_b]
elif (min_a < min_b) & (max_b < max_a):
n = max( -(-max_a // max_b), -(min_a // min_b) )
return [min_a, max_a], [min_a / n, max_a / n]
elif (min_a == min_b):
m = max(max_a, max_b)
return [min_a, m], [min_b, m]
elif max_a == max_b:
m = min(min_a, min_b)
return [m, max_a], [m, max_b]
init_notebook_mode(connected=True)
y0 = np.random.randint(-5, 35, 6)
y1 = np.random.randint(-7, 28, 6)
print(y0, y1)
range0, range1 = make_proportional_intervals(y0, y1)
traces = [
go.Bar(
x=[1, 2, 3, 4, 5, 6],
y=y0,
name='y actual'
),
go.Bar(
x=[1],
y=[0],
name='y dummy',
hoverinfo='none',
showlegend=False
),
go.Bar(
x=[1],
y=[0],
yaxis='y2',
name='y2 dummy',
hoverinfo='none',
showlegend=False
),
go.Bar(
x=[1, 2, 3, 4, 5, 6],
y=y1,
yaxis='y2',
name='y2 actual'
)
]
# layout
layout = go.Layout(
barmode='group',
yaxis=dict(
title='y actual',
rangemode="tozero",
#anchor='x',
overlaying='y2',
side="left",
range = range0
),
yaxis2=dict(
title='y2 actual',
side='right',
rangemode = "tozero",
#anchor='x',
range = range1
)
)
fig = go.Figure(data=traces, layout=layout)
iplot(fig)
Again, this is all just a work around for the fact that you have negative numbers and cannot use rangemode = "tozero" as scene here. Perhaps the developers will add something in the future to rangemode to correct for this.
