from __future__ import annotations
from typing import List
import gradio as gr
from state.app_state import AppState
from controllers.linear_regression_controller import run_linear_regression
def build(state: AppState) -> None:
"""
Build the Linear Regression tab.
This recreates the behaviour and layout of the monolithic implementation,
while delegating statistics to the `stats.linear_regression` module and
wiring / validation to `regression_controller`.
"""
gr.Markdown("## 📈 Linear Regression")
# Use the numeric columns discovered by the Data tab
numeric_cols = state.numeric_cols or []
# ------------------------------------------------------------------
# Top row: column selection
# ------------------------------------------------------------------
with gr.Row(elem_id="row_centered"):
refresh_columns_button = gr.Button("🔄 Refresh Numeric Columns")
dependent_dropdown = gr.Dropdown(
label="Dependent Variable",
choices=numeric_cols,
interactive=True,
elem_classes="data_related",
elem_id="custom_dropdown",
)
independent_dropdown = gr.Dropdown(
label="Independent Variable(s)",
multiselect=True,
choices=numeric_cols,
interactive=True,
elem_classes="data_related",
elem_id="custom_dropdown",
)
# ------------------------------------------------------------------
# Formula controls
# ------------------------------------------------------------------
with gr.Row():
formula_check = gr.Checkbox(
label="Would you like to write down the regression formula?",
value=False,
interactive=True,
)
formula_text = gr.Textbox(
label="Write the formula",
placeholder="Y ~ X + np.sin(X) + I((X-5)**2)",
interactive=True,
visible=False,
)
formula_latex = gr.Textbox(
label="Write the formula in LaTeX (Optional)",
placeholder=r"Y = X + \sin(X) + (X-5)^2",
interactive=True,
visible=False,
)
# ------------------------------------------------------------------
# Global options
# ------------------------------------------------------------------
with gr.Row():
alpha_input = gr.Textbox(
label="Confidence level (e.g. 0.95)",
value=0.95,
interactive=True,
)
intercept_check = gr.Checkbox(
label="Include intercept",
value=True,
interactive=True,
)
graph_check_reg = gr.Checkbox(
label="Create graph",
value=True,
interactive=True,
)
# ------------------------------------------------------------------
# Graph options
# ------------------------------------------------------------------
with gr.Row() as graph_options:
graph_dropdown = gr.Dropdown(
label="Graph",
choices=["Simple Regression", "Observed vs Predicted"],
value="Simple Regression",
interactive=True,
)
show_ci_check = gr.Checkbox(
label="Include CI",
value=True,
interactive=True,
)
show_pi_check = gr.Checkbox(
label="Include PI",
value=True,
interactive=True,
)
fit_to_obs_check = gr.Checkbox(
label="Fit to observations",
value=True,
interactive=True,
)
x_vect_input = gr.Textbox(
label="Minimum and maximum of dependent variable ",
value="",
visible=False,
interactive=True,
)
# ------------------------------------------------------------------
# Run button
# ------------------------------------------------------------------
with gr.Column(elem_id="column_centered"):
run_regression_button = gr.Button(
value="🚀 Run Linear Regression",
elem_id="run_button",
)
# ------------------------------------------------------------------
# Results blocks
# ------------------------------------------------------------------
with gr.Row(visible=False) as output_table_row:
output_table = gr.HTML(label="Regression Summary")
with gr.Row(visible=False) as output_plot_row:
output_plot = gr.Plot(label="Regression Plot")
# ------------------------------------------------------------------
# Wiring helpers
# ------------------------------------------------------------------
def on_toggle_formula(check: bool):
# When the user chooses to write a formula, we show the formula
# text inputs and hide the intercept checkbox (the formula
# decides the intercept).
return (
gr.update(visible=check), # formula_text
gr.update(visible=check), # formula_latex
gr.update(visible=not check, value=not check), # intercept_check
)
def on_update_graph_choices(independent_vars: List[str]):
# With a single predictor we can show both graphs; otherwise only
# the "Observed vs Predicted" graph is available.
if len(independent_vars) == 1:
return gr.update(
choices=["Simple Regression", "Observed vs Predicted"],
value="Simple Regression",
)
return gr.update(
choices=["Observed vs Predicted"],
value="Observed vs Predicted",
)
def on_toggle_graph_options(graph_type: str, fit_to_obs: bool):
# For the Simple Regression graph we allow CI/PI and fit_to_obs;
# for Observed vs Predicted no interval / range options are used.
if graph_type == "Simple Regression":
return (
gr.update(visible=True), # show_ci_check
gr.update(visible=True), # show_pi_check
gr.update(visible=True, value=fit_to_obs), # fit_to_obs_check
)
else:
return (
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False, value=True),
)
def on_toggle_graph_block(check: bool):
return gr.update(visible=check)
def on_toggle_range_input(fit_to_obs: bool):
# Range input only makes sense when not fitting strictly to observed X.
return gr.update(visible=not fit_to_obs)
def on_refresh_columns():
cols = state.numeric_cols or []
return (
gr.update(choices=cols),
gr.update(choices=cols),
)
# ------------------------------------------------------------------
# Run callback
# ------------------------------------------------------------------
def on_run(
formula_check_val,
formula_text_val,
formula_latex_val,
dep_var_val,
indep_vars_val,
alpha_input_val,
intercept_val,
graph_check_val,
graph_type_val,
show_ci_val,
show_pi_val,
fit_to_obs_val,
x_range_text_val,
):
# Ensure we have a dataset
if state.df is None:
return (
gr.update(visible=True),
gr.update(visible=True, value="Error: No dataset loaded."),
gr.update(visible=False),
gr.update(visible=False, value=None),
)
try:
summary_html, params_df, fig = run_linear_regression(
df=state.df,
filtered_df=state.filtered_df,
formula_check=bool(formula_check_val),
formula_text=str(formula_text_val or ""),
formula_latex=str(formula_latex_val or ""),
dependent_var=dep_var_val,
independent_vars=list(indep_vars_val or []),
alpha_input=str(alpha_input_val),
intercept=bool(intercept_val),
graph_check=bool(graph_check_val),
graph_type=str(graph_type_val),
show_ci=bool(show_ci_val),
show_pi=bool(show_pi_val),
fit_to_obs=bool(fit_to_obs_val),
x_range_text=str(x_range_text_val or ""),
round_digits=4,
)
except Exception as e: # noqa: BLE001
# Surface a friendly HTML error message
error_html = f"Regression failed: {e}"
return (
gr.update(visible=True),
gr.update(visible=True, value=error_html),
gr.update(visible=False),
gr.update(visible=False, value=None),
)
# Store the coefficient table for potential downloads elsewhere
state.export_table = params_df
return (
gr.update(visible=True),
gr.update(visible=True, value=summary_html),
gr.update(visible=fig is not None),
gr.update(visible=fig is not None, value=fig),
)
# ------------------------------------------------------------------
# Wire up callbacks
# ------------------------------------------------------------------
formula_check.change(
on_toggle_formula,
inputs=formula_check,
outputs=[formula_text, formula_latex, intercept_check],
)
refresh_columns_button.click(
fn=on_refresh_columns,
inputs=[],
outputs=[dependent_dropdown, independent_dropdown],
)
independent_dropdown.change(
on_update_graph_choices,
inputs=independent_dropdown,
outputs=graph_dropdown,
)
graph_check_reg.change(
on_toggle_graph_block,
inputs=graph_check_reg,
outputs=graph_options,
)
graph_dropdown.change(
on_toggle_graph_options,
inputs=[graph_dropdown, fit_to_obs_check],
outputs=[show_ci_check, show_pi_check, fit_to_obs_check],
)
fit_to_obs_check.change(
on_toggle_range_input,
inputs=fit_to_obs_check,
outputs=x_vect_input,
)
run_regression_button.click(
on_run,
inputs=[
formula_check,
formula_text,
formula_latex,
dependent_dropdown,
independent_dropdown,
alpha_input,
intercept_check,
graph_check_reg,
graph_dropdown,
show_ci_check,
show_pi_check,
fit_to_obs_check,
x_vect_input,
],
outputs=[output_table_row, output_table, output_plot_row, output_plot],
)