Major refactor for improved documentation and readability
This commit is contained in:
Landon Wark
278
src/submissions/frontend/visualizations/control_charts.py
Normal file
278
src/submissions/frontend/visualizations/control_charts.py
Normal file
@@ -0,0 +1,278 @@
|
||||
'''
|
||||
Functions for constructing controls graphs using plotly.
|
||||
'''
|
||||
import plotly.express as px
|
||||
import pandas as pd
|
||||
from pathlib import Path
|
||||
from plotly.graph_objects import Figure
|
||||
import logging
|
||||
from backend.excel import get_unique_values_in_df_column
|
||||
|
||||
logger = logging.getLogger(f"submissions.{__name__}")
|
||||
|
||||
|
||||
def create_charts(ctx:dict, df:pd.DataFrame, ytitle:str|None=None) -> Figure:
|
||||
"""
|
||||
Constructs figures based on parsed pandas dataframe.
|
||||
|
||||
Args:
|
||||
settings (dict): settings passed down from gui
|
||||
df (pd.DataFrame): input dataframe
|
||||
group_name (str): controltype
|
||||
|
||||
Returns:
|
||||
Figure: plotly figure
|
||||
"""
|
||||
from backend.excel import drop_reruns_from_df
|
||||
# converts starred genera to normal and splits off list of starred
|
||||
genera = []
|
||||
if df.empty:
|
||||
return None
|
||||
for item in df['genus'].to_list():
|
||||
try:
|
||||
if item[-1] == "*":
|
||||
genera.append(item[-1])
|
||||
else:
|
||||
genera.append("")
|
||||
except IndexError:
|
||||
genera.append("")
|
||||
df['genus'] = df['genus'].replace({'\*':''}, regex=True)
|
||||
df['genera'] = genera
|
||||
df = df.dropna()
|
||||
# remove original runs, using reruns if applicable
|
||||
df = drop_reruns_from_df(ctx=ctx, df=df)
|
||||
# sort by and exclude from
|
||||
sorts = ['submitted_date', "target", "genus"]
|
||||
exclude = ['name', 'genera']
|
||||
modes = [item for item in df.columns if item not in sorts and item not in exclude]# and "_hashes" not in item]
|
||||
# Set descending for any columns that have "{mode}" in the header.
|
||||
ascending = [False if item == "target" else True for item in sorts]
|
||||
df = df.sort_values(by=sorts, ascending=ascending)
|
||||
# actual chart construction is done by
|
||||
fig = construct_chart(ctx=ctx, df=df, modes=modes, ytitle=ytitle)
|
||||
return fig
|
||||
|
||||
|
||||
|
||||
def generic_figure_markers(fig:Figure, modes:list=[], ytitle:str|None=None) -> Figure:
|
||||
"""
|
||||
Adds standard layout to figure.
|
||||
|
||||
Args:
|
||||
fig (Figure): Input figure.
|
||||
modes (list, optional): List of modes included in figure. Defaults to [].
|
||||
|
||||
Returns:
|
||||
Figure: Output figure with updated titles, rangeslider, buttons.
|
||||
"""
|
||||
if modes != []:
|
||||
ytitle = modes[0]
|
||||
# Creating visibles list for each mode.
|
||||
fig.update_layout(
|
||||
xaxis_title="Submitted Date (* - Date parsed from fastq file creation date)",
|
||||
yaxis_title=ytitle,
|
||||
showlegend=True,
|
||||
barmode='stack',
|
||||
updatemenus=[
|
||||
dict(
|
||||
type="buttons",
|
||||
direction="right",
|
||||
x=0.7,
|
||||
y=1.2,
|
||||
showactive=True,
|
||||
buttons=make_buttons(modes=modes, fig_len=len(fig.data)),
|
||||
)
|
||||
]
|
||||
)
|
||||
fig.update_xaxes(
|
||||
rangeslider_visible=True,
|
||||
rangeselector=dict(
|
||||
buttons=list([
|
||||
dict(count=1, label="1m", step="month", stepmode="backward"),
|
||||
dict(count=3, label="3m", step="month", stepmode="backward"),
|
||||
dict(count=6, label="6m", step="month", stepmode="backward"),
|
||||
dict(count=1, label="YTD", step="year", stepmode="todate"),
|
||||
dict(count=1, label="1y", step="year", stepmode="backward"),
|
||||
dict(step="all")
|
||||
])
|
||||
)
|
||||
)
|
||||
assert type(fig) == Figure
|
||||
return fig
|
||||
|
||||
|
||||
def make_buttons(modes:list, fig_len:int) -> list:
|
||||
"""
|
||||
Creates list of buttons with one for each mode to be used in showing/hiding mode traces.
|
||||
|
||||
Args:
|
||||
modes (list): list of modes used by main parser.
|
||||
fig_len (int): number of traces in the figure
|
||||
|
||||
Returns:
|
||||
list: list of buttons.
|
||||
"""
|
||||
buttons = []
|
||||
if len(modes) > 1:
|
||||
for ii, mode in enumerate(modes):
|
||||
# What I need to do is create a list of bools with the same length as the fig.data
|
||||
mode_vis = [True] * fig_len
|
||||
# And break it into {len(modes)} chunks
|
||||
mode_vis = list(divide_chunks(mode_vis, len(modes)))
|
||||
# Then, for each chunk, if the chunk index isn't equal to the index of the current mode, set to false
|
||||
for jj, sublist in enumerate(mode_vis):
|
||||
if jj != ii:
|
||||
mode_vis[jj] = [not elem for elem in mode_vis[jj]]
|
||||
# Finally, flatten list.
|
||||
mode_vis = [item for sublist in mode_vis for item in sublist]
|
||||
# Now, make button to add to list
|
||||
buttons.append(dict(label=mode, method="update", args=[
|
||||
{"visible": mode_vis},
|
||||
{"yaxis.title.text": mode},
|
||||
]
|
||||
))
|
||||
return buttons
|
||||
|
||||
def output_figures(settings:dict, figs:list, group_name:str):
|
||||
"""
|
||||
Writes plotly figure to html file.
|
||||
|
||||
Args:
|
||||
settings (dict): settings passed down from click
|
||||
fig (Figure): input figure object
|
||||
group_name (str): controltype
|
||||
"""
|
||||
with open(Path(settings['folder']['output']).joinpath(f'{group_name}.html'), "w") as f:
|
||||
for fig in figs:
|
||||
try:
|
||||
f.write(fig.to_html(full_html=False, include_plotlyjs='cdn'))
|
||||
except AttributeError:
|
||||
logger.error(f"The following figure was a string: {fig}")
|
||||
|
||||
|
||||
|
||||
def construct_chart(ctx:dict, df:pd.DataFrame, modes:list, ytitle:str|None=None) -> Figure:
|
||||
"""
|
||||
Creates a plotly chart for controls from a pandas dataframe
|
||||
|
||||
Args:
|
||||
ctx (dict): settings passed down from gui
|
||||
df (pd.DataFrame): input dataframe of controls
|
||||
modes (list): analysis modes to construct charts for
|
||||
ytitle (str | None, optional): title on the y-axis. Defaults to None.
|
||||
|
||||
Returns:
|
||||
Figure: output stacked bar chart.
|
||||
"""
|
||||
fig = Figure()
|
||||
for ii, mode in enumerate(modes):
|
||||
if "count" in mode:
|
||||
df[mode] = pd.to_numeric(df[mode],errors='coerce')
|
||||
color = "genus"
|
||||
color_discrete_sequence=None
|
||||
elif 'percent' in mode:
|
||||
color = "genus"
|
||||
color_discrete_sequence=None
|
||||
else:
|
||||
color = "target"
|
||||
match get_unique_values_in_df_column(df, 'target'):
|
||||
case ['Target']:
|
||||
color_discrete_sequence=["blue"]
|
||||
case ['Off-target']:
|
||||
color_discrete_sequence=['red']
|
||||
case _:
|
||||
color_discrete_sequence=['blue', 'red']
|
||||
bar = px.bar(df, x="submitted_date",
|
||||
y=mode,
|
||||
color=color,
|
||||
title=mode,
|
||||
barmode='stack',
|
||||
hover_data=["genus", "name", "target", mode],
|
||||
text="genera",
|
||||
color_discrete_sequence=color_discrete_sequence
|
||||
)
|
||||
bar.update_traces(visible = ii == 0)
|
||||
fig.add_traces(bar.data)
|
||||
return generic_figure_markers(fig=fig, modes=modes, ytitle=ytitle)
|
||||
|
||||
# Below are the individual construction functions. They must be named "construct_{mode}_chart" and
|
||||
# take only json_in and mode to hook into the main processor.
|
||||
|
||||
def construct_refseq_chart(settings:dict, df:pd.DataFrame, group_name:str, mode:str) -> Figure:
|
||||
"""
|
||||
Constructs intial refseq chart for both contains and matches (depreciated).
|
||||
|
||||
Args:
|
||||
settings (dict): settings passed down from gui.
|
||||
df (pd.DataFrame): dataframe containing all sample data for the group.
|
||||
group_name (str): name of the group being processed.
|
||||
mode (str): contains or matches, overwritten by hardcoding, so don't think about it too hard.
|
||||
|
||||
Returns:
|
||||
Figure: initial figure with contains and matches traces.
|
||||
"""
|
||||
# This overwrites the mode from the signature, might get confusing.
|
||||
fig = Figure()
|
||||
modes = ['contains', 'matches']
|
||||
for ii, mode in enumerate(modes):
|
||||
bar = px.bar(df, x="submitted_date",
|
||||
y=f"{mode}_ratio",
|
||||
color="target",
|
||||
title=f"{group_name}_{mode}",
|
||||
barmode='stack',
|
||||
hover_data=["genus", "name", f"{mode}_hashes"],
|
||||
text="genera"
|
||||
)
|
||||
bar.update_traces(visible = ii == 0)
|
||||
# Plotly express returns a full figure, so we have to use the data from that figure only.
|
||||
fig.add_traces(bar.data)
|
||||
# sys.exit(f"number of traces={len(fig.data)}")
|
||||
return generic_figure_markers(fig=fig, modes=modes)
|
||||
|
||||
|
||||
def construct_kraken_chart(settings:dict, df:pd.DataFrame, group_name:str, mode:str) -> Figure:
|
||||
"""
|
||||
Constructs intial refseq chart for each mode in the kraken config settings. (depreciated)
|
||||
|
||||
Args:
|
||||
settings (dict): settings passed down from click.
|
||||
df (pd.DataFrame): dataframe containing all sample data for the group.
|
||||
group_name (str): name of the group being processed.
|
||||
mode (str): kraken modes retrieved from config file by setup.
|
||||
|
||||
Returns:
|
||||
Figure: initial figure with traces for modes
|
||||
"""
|
||||
df[f'{mode}_count'] = pd.to_numeric(df[f'{mode}_count'],errors='coerce')
|
||||
# The actual percentage from kraken was off due to exclusion of NaN, recalculating.
|
||||
df[f'{mode}_percent'] = 100 * df[f'{mode}_count'] / df.groupby('submitted_date')[f'{mode}_count'].transform('sum')
|
||||
modes = settings['modes'][mode]
|
||||
# This overwrites the mode from the signature, might get confusing.
|
||||
fig = Figure()
|
||||
for ii, entry in enumerate(modes):
|
||||
bar = px.bar(df, x="submitted_date",
|
||||
y=entry,
|
||||
color="genus",
|
||||
title=f"{group_name}_{entry}",
|
||||
barmode="stack",
|
||||
hover_data=["genus", "name", "target"],
|
||||
text="genera",
|
||||
)
|
||||
bar.update_traces(visible = ii == 0)
|
||||
fig.add_traces(bar.data)
|
||||
return generic_figure_markers(fig=fig, modes=modes)
|
||||
|
||||
|
||||
def divide_chunks(input_list:list, chunk_count:int):
|
||||
"""
|
||||
Divides a list into {chunk_count} equal parts
|
||||
|
||||
Args:
|
||||
input_list (list): Initials list
|
||||
chunk_count (int): size of each chunk
|
||||
|
||||
Returns:
|
||||
tuple: tuple containing sublists.
|
||||
"""
|
||||
k, m = divmod(len(input_list), chunk_count)
|
||||
return (input_list[i*k+min(i, m):(i+1)*k+min(i+1, m)] for i in range(chunk_count))
|
||||
Reference in New Issue
Block a user