COMPOSITE
Download Flojoy Studio to try this app
Create a combination of Plotly visualizations for a given input data container. Inputs
------
primary_trace : OrderedPair|DataFrame|Matrix|Vector
the DataContainer to be visualized as the first figure
secondary_trace : OrderedPair|DataFrame|Matrix|Vector
the DataContainer to be visualized as the second figure Params: first_figure : 'bar' | 'line' | 'histogram' | 'scatter' plotly type to display as the first figure, default is 'scatter' second_figure : 'bar' | 'line' | 'histogram' | 'scatter' plotly type to display as the second figure, default is 'line' Returns: out : Plotly the DataContainer containing Plotly visualization of both figures
Python Code
from flojoy import flojoy, OrderedPair, DataFrame, Matrix, Plotly, Vector
import plotly.graph_objects as go
import pandas as pd
import numpy as np
from blocks.DATA.VISUALIZATION.template import plot_layout
from typing import Literal
@flojoy
def COMPOSITE(
primary_trace: OrderedPair | DataFrame | Matrix | Vector,
secondary_trace: OrderedPair | DataFrame | Matrix | Vector,
first_figure: Literal["bar", "line", "histogram", "scatter"] = "scatter",
second_figure: Literal["bar", "line", "histogram", "scatter"] = "line",
) -> Plotly:
"""Create a combination of Plotly visualizations for a given input data container.
Inputs
------
primary_trace : OrderedPair|DataFrame|Matrix|Vector
the DataContainer to be visualized as the first figure
secondary_trace : OrderedPair|DataFrame|Matrix|Vector
the DataContainer to be visualized as the second figure
Parameters
----------
first_figure : 'bar' | 'line' | 'histogram' | 'scatter'
plotly type to display as the first figure, default is 'scatter'
second_figure : 'bar' | 'line' | 'histogram' | 'scatter'
plotly type to display as the second figure, default is 'line'
Returns
-------
Plotly
the DataContainer containing Plotly visualization of both figures
"""
layout = plot_layout(title="COMPOSITE")
fig = go.Figure(layout=layout)
match_figure(fig, first_figure, primary_trace)
match_figure(fig, second_figure, secondary_trace)
fig.update_layout(dict(autosize=True, height=None, width=None))
return Plotly(fig=fig)
def match_figure(
fig: go.Figure,
figure_type: Literal["bar", "line", "histogram", "scatter"],
dc: OrderedPair | Matrix | DataFrame | Vector,
):
match figure_type:
case "bar":
add_bar_trace(fig, dc)
case "histogram":
add_histogram_trace(fig, dc)
case "line":
add_line_trace(fig, dc)
case "scatter":
add_scatter_trace(fig, dc)
def add_bar_trace(fig: go.Figure, dc: OrderedPair | Matrix | DataFrame | Vector):
match dc:
case DataFrame():
df = dc.m
first_col = df.iloc[:, 0]
is_timeseries = False
if is_timeseries:
for col in df.columns:
if col != df.columns[0]:
fig.add_trace(go.Bar(y=df[col].values, x=first_col, name=col))
fig.update_layout(xaxis_title=df.columns[0])
else:
for col in df.columns:
if df[col].dtype == "object":
counts = df[col].value_counts()
fig.add_trace(
go.Bar(
x=counts.index.tolist(),
y=counts.tolist(),
name=col,
)
)
else:
fig.add_trace(go.Bar(x=df.index, y=df[col], name=col))
fig.update_layout(xaxis_title="DF index", yaxis_title="Y Axis")
case OrderedPair():
x = dc.x
if isinstance(dc.x, dict):
dict_keys = list(dc.x.keys())
x = dc.x[dict_keys[0]]
y = dc.y
fig.add_trace(go.Bar(x=x, y=y))
case Matrix():
m = dc.m
num_rows, num_cols = m.shape
x_ticks = np.arange(num_cols)
for i in range(num_rows):
fig.add_trace(go.Bar(x=x_ticks, y=m[i, :], name=f"Row {i+1}"))
fig.update_layout(xaxis_title="Column", yaxis_title="Value")
case Vector():
y = dc.v
x = np.arange(len(y))
fig.add_trace(go.Bar(x=x, y=y))
def add_histogram_trace(fig: go.Figure, dc: OrderedPair | Matrix | DataFrame | Vector):
match dc:
case DataFrame():
df = dc.m
for col in df.columns:
fig.add_trace(go.Histogram(x=df[col], name=col))
fig.update_layout(xaxis_title="Value", yaxis_title="Frequency")
case OrderedPair():
y = dc.y
fig.add_trace(go.Histogram(x=y))
case Matrix():
m = dc.m
histogram_trace = go.Histogram(x=m.flatten())
fig.add_trace(histogram_trace)
case Vector():
y = dc.v
fig.add_trace(go.Histogram(x=y))
def add_line_trace(fig: go.Figure, dc: OrderedPair | Matrix | DataFrame | Vector):
match dc:
case DataFrame():
df = dc.m
first_col = df.iloc[:, 0]
is_timeseries = False
if pd.api.types.is_datetime64_any_dtype(first_col):
is_timeseries = True
if is_timeseries:
for col in df.columns:
if col != df.columns[0]:
fig.add_trace(
go.Scatter(
y=df[col].values,
x=first_col,
mode="lines",
name=col,
)
)
else:
for col in df.columns:
fig.add_trace(
go.Scatter(
y=df[col].values,
x=df.index,
mode="lines",
name=col,
)
)
case OrderedPair():
x = dc.x
if isinstance(dc.x, dict):
dict_keys = list(dc.x.keys())
x = dc.x[dict_keys[0]]
y = dc.y
fig.add_trace(go.Scatter(x=x, y=y, mode="lines"))
case Matrix():
m = dc.m
num_rows, num_cols = m.shape
x_ticks = np.arange(num_cols)
for i in range(num_rows):
fig.add_trace(
go.Scatter(x=x_ticks, y=m[i, :], name=f"Row {i+1}", mode="lines")
)
fig.update_layout(xaxis_title="Column", yaxis_title="Value")
case Vector():
y = dc.v
x = np.arange(len(y))
fig.add_trace(go.Scatter(x=x, y=y, mode="lines"))
def add_scatter_trace(fig: go.Figure, dc: OrderedPair | Matrix | DataFrame | Vector):
match dc:
case OrderedPair():
x = dc.x
if isinstance(dc.x, dict):
dict_keys = list(dc.x.keys())
x = dc.x[dict_keys[0]]
y = dc.y
fig.add_trace(go.Scatter(x=x, y=y, mode="markers", marker=dict(size=4)))
case DataFrame():
df = dc.m
first_col = df.iloc[:, 0]
is_timeseries = False
if pd.api.types.is_datetime64_any_dtype(first_col):
is_timeseries = True
if is_timeseries:
for col in df.columns:
if col != df.columns[0]:
fig.add_trace(
go.Scatter(x=first_col, y=df[col], mode="markers", name=col)
)
else:
for col in df.columns:
fig.add_trace(
go.Scatter(x=df.index, y=df[col], mode="markers", name=col)
)
case Matrix():
m = dc.m
num_rows, num_cols = m.shape
x_ticks = np.arange(num_cols)
for i in range(num_rows):
fig.add_trace(
go.Scatter(x=x_ticks, y=m[i, :], name=f"Row {i+1}", mode="markers")
)
fig.update_layout(xaxis_title="Column", yaxis_title="Value")
case Vector():
y = dc.v
x = np.arange(len(y))
fig.add_trace(go.Scatter(x=x, y=y, mode="markers", marker=dict(size=4)))
Example App
Having problems with this example app? Join our Discord community and we will help you out!
In this example weāre simulating data from āLINSPACEā, āPLOTLY_DATASETā and visualizing them with COMPOSITE node which creates a Plotly combination of visualizations depending on userās choice from line, bar, histogram, scatter for two inputs required by the node.