import numpy as np
import pandas as pd
import gradio as gr
import plotly.graph_objs as go
from plotly.subplots import make_subplots
import datetime


# Function to load data from a text file into a numpy array of the right format
def load_data(filename):
        with open(filename, 'r') as file:
            data = [int(line.strip()) for line in file]
        data = np.array(data, dtype=int)
        data[data == -100] = -1
        return data.reshape((-1, 1))
    

def process_csv(file):
    
    if file is None:
        gr.Warning("No file uploaded") 
    
    # Load the data
    acc_data = np.load(file)
    predicted_labels = load_data('003_1_R.txt')
    
    # Append the new columns
    complete_array = np.hstack((acc_data, predicted_labels))
    
    # Define the time axis
    duration = 33.333333  # milliseconds
    num_samples = complete_array.shape[0]
    time_axis = np.arange(0, num_samples * duration, duration)

    # Convert time axis to datetime objects
    start_time = datetime.datetime.now().replace(hour=0, minute=0, second=0, microsecond=0)
    time_axis_datetime = [start_time + datetime.timedelta(milliseconds=int(t)) for t in time_axis]

    # Convert datetime objects to formatted strings
    time_axis_labels = [t.strftime('%H:%M:%S.%f')[:-3] for t in time_axis_datetime]
    
    # Calculate the total number of predicted functional and non-functional activity
    total_predicted_functional = np.sum(complete_array[:, 3] != 0)
    total_predicted_non_functional = np.sum(complete_array[:, 3] == 0)

    # Calculate percentages
    predicted_functional_percentage = (total_predicted_functional / len(complete_array)) * 100
    predicted_non_functional_percentage = (total_predicted_non_functional / len(complete_array)) * 100

    # Formulate return string
    return_string = f"Percentage of predicted functional activity: {predicted_functional_percentage:.2f}%\nPercentage of predicted non-functional activity: {predicted_non_functional_percentage:.2f}%"
    
    # Create subplots
    fig = make_subplots(rows=2, cols=1, shared_xaxes=True, row_heights=[0.6, 0.4], specs=[[{"type": "scatter"}], [{"type": "scatter"}]])

    # Add traces to the subplots
    fig.add_trace(go.Scatter(x=time_axis_labels, y=complete_array[:, 0], mode='lines', name='Acc X', line=dict(width=0.75)), row=1, col=1)
    fig.add_trace(go.Scatter(x=time_axis_labels, y=complete_array[:, 1], mode='lines', name='Acc Y', line=dict(width=0.75)), row=1, col=1)
    fig.add_trace(go.Scatter(x=time_axis_labels, y=complete_array[:, 2], mode='lines', name='Acc Z', line=dict(width=0.75)), row=1, col=1)
    fig.add_trace(go.Scatter(x=time_axis_labels, y=complete_array[:, 3], mode='lines', name='Predicted labels', line=dict(width=1)), row=2, col=1)


    # Update layout
    fig.update_layout(
        title='Accelerometer Data with Annotated Labels',
        xaxis=dict(title='Time (milliseconds)'),
        yaxis=dict(title='Accelerometer Data'),
        yaxis2=dict(title='Predicted'),
        showlegend=True,
        height=600
    )

    return return_string, fig



with gr.Blocks(theme=gr.themes.Base()) as demo:
    gr.Markdown(
    """
    # Functional Upper Limb Activity Recognition Model
    Upload your csv file containing accelerometer data to obtain a prediction on the amount of functional activity of the upper limbs.
    """)
    with gr.Row(equal_height=True):
        with gr.Column():
            input_file = gr.File(label="Upload CSV file")
            with gr.Row():
                submit_btn = gr.Button("Submit", variant='primary')
                clear_btn = gr.ClearButton(input_file, variant='secondary')
        output_text = gr.Textbox(label="Prediction statistics")
    output_plot = gr.Plot(label="CSV Plot")
    
    submit_btn.click(fn=process_csv, inputs=input_file, outputs=[output_text, output_plot])
    clear_btn.click(fn=lambda: input_file.reinit())

demo.launch()