import numpy as np
import sqlalchemy as sa
from apptax.taxonomie.models import Taxref
from bokeh.embed import json_item
from bokeh.embed.standalone import StandaloneEmbedJson
from bokeh.layouts import column
from bokeh.models import (
AnnularWedge,
ColumnDataSource,
CustomJS,
Legend,
LegendItem,
Range1d,
Select,
)
from bokeh.palettes import Plasma256, Turbo256, linear_palette
from bokeh.plotting import figure
from geonature.core.gn_synthese.models import Synthese
from geonature.utils.env import db
[docs]
def taxon_distribution_plot(imprt) -> StandaloneEmbedJson:
"""
Generate a plot of the taxonomic distribution (for each rank) based on the import.
The following ranks are used:
- group1_inpn
- group2_inpn
- group3_inpn
- sous_famille
- tribu
- classe
- ordre
- famille
- phylum
- regne
Parameters
----------
imprt : TImports
The import object to generate the plot from.
Returns
-------
dict
Returns a dict containing data required to generate the plot
"""
taxon_ranks = "regne phylum classe ordre famille sous_famille tribu group1_inpn group2_inpn group3_inpn".split()
figures = []
# Generate the plot for each rank
for rank in taxon_ranks:
# Generate the query to retrieve the count for each value taken by the rank
c_rank_taxref = getattr(Taxref, rank)
query = (
sa.select(
sa.func.count(sa.distinct(Synthese.cd_nom)).label("count"),
c_rank_taxref.label("rank_value"),
)
.select_from(Synthese)
.outerjoin(Taxref, Taxref.cd_nom == Synthese.cd_nom)
.where(Synthese.id_import == imprt.id_import)
.group_by(c_rank_taxref)
)
data = np.asarray(
[r if r[1] != "" else (r[0], "Non-assigné") for r in db.session.execute(query).all()]
)
# if data is empty
if not data.size:
continue
# Extract the rank values and counts
rank_values, counts = data[:, 1], data[:, 0].astype(int)
# Get angles (in radians) where start each section of the pie chart
angles = np.cumsum(
[2 * np.pi * (count / sum(counts)) for i, count in enumerate(counts)]
).tolist()
# Generate the color palette
palette = (
linear_palette(Turbo256, len(rank_values))
if len(rank_values) > 5
else linear_palette(Plasma256, len(rank_values))
)
colors = {value: palette[ix] for ix, value in enumerate(rank_values)}
# Store the data in a Bokeh data structure
browsers_source = ColumnDataSource(
dict(
start=[0] + angles[:-1],
end=angles,
colors=[colors[rank_value] for rank_value in rank_values],
countvalue=counts,
rankvalue=rank_values,
)
)
# Create the Figure object
fig = figure(
x_range=Range1d(start=-3, end=3),
y_range=Range1d(start=-3, end=3),
title=f"Distribution des taxons selon {rank}",
tooltips=[("Number", "@countvalue"), (rank, "@rankvalue")],
toolbar_location=None,
)
# Add the Pie chart
glyph = AnnularWedge(
x=0,
y=0,
inner_radius=0.9,
outer_radius=1.8,
start_angle="start",
end_angle="end",
line_color="white",
line_width=3,
fill_color="colors",
)
r = fig.add_glyph(browsers_source, glyph)
# Add the legend
legend = Legend(location="top_center")
for i, name in enumerate(colors):
legend.items.append(LegendItem(label=name, renderers=[r], index=i))
fig.add_layout(legend, "below")
fig.legend.ncols = 3 if len(colors) < 10 else 5
# ERASE the grid and axis
fig.grid.visible = False
fig.axis.visible = False
fig.title.text_font_size = "16pt"
# Hide the unselected rank plot
if rank != "regne":
fig.visible = False
# Add the plot to the list of figures
figures.append(fig)
if not figures:
return {}
# Generate the layout with the plots and the rank selector
plot_area = column(figures)
select_plot = Select(
title="Critère",
value=0, # Default is "regne"
options=[(ix, rank) for ix, rank in enumerate(taxon_ranks)],
width=fig.width,
)
# Update the visibility of the plots when the taxonomic rank selector changes
select_plot.js_on_change(
"value",
CustomJS(
args=dict(s=select_plot, col=plot_area),
code="""
for (const plot of col.children) {
plot.visible = false
}
col.children[s.value].visible = true
""",
),
)
column_fig = column(plot_area, select_plot, sizing_mode="scale_width")
return json_item(column_fig)
