utils.py

import streamlit as st
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import requests
import json
from datetime import datetime

def get_series_name_and_unit(series, dataset_description):
    """
    Extract the name and unit from a time series using its dataset description.
    
    Args:
        series: Dictionary containing series data
        dataset_description: Dictionary containing dataset field descriptions
        
    Returns:
        tuple: (name, unit) of the series
    """
    field_id = series['datacellar:datasetFieldID']
    field = next((f for f in dataset_description['datacellar:datasetFields'] 
                 if f['datacellar:datasetFieldID'] == field_id), None)
    
    name = field['datacellar:fieldName'] if field else f'Series {field_id}'
    unit = field['datacellar:type']['datacellar:unit'] if field else 'Unknown'
    
    # Override name if metadata contains loadType
    if 'datacellar:timeSeriesMetadata' in series:
        metadata = series['datacellar:timeSeriesMetadata']
        if 'datacellar:loadType' in metadata:
            name = metadata['datacellar:loadType']
    
    return name, unit

def process_series(series, dataset_description, is_input=False):
    """
    Process a single time series into a pandas DataFrame.
    
    Args:
        series: Dictionary containing series data
        dataset_description: Dictionary containing dataset field descriptions
        is_input: Boolean indicating if this is input data
        
    Returns:
        tuple: (DataFrame, unit, name) of the processed series
    """
    name, unit = get_series_name_and_unit(series, dataset_description)
    df = pd.DataFrame(series['datacellar:timeseriesDataPoints'])
    
    # Convert timestamp to datetime and ensure values are numeric
    df['datacellar:timeStamp'] = pd.to_datetime(df['datacellar:timeStamp'])
    df['datacellar:value'] = pd.to_numeric(df['datacellar:value'], errors='coerce')
    
    # Add series identifier
    df['series_id'] = f'{name} (Input)' if is_input else name
    
    return df, unit, name

def load_and_process_data(json_data, input_data=None):
    """
    Load and process time series from the JSON data, filtering out empty series.
    """
    series_by_unit = {}
    try:
        dataset_description = json_data['datacellar:datasetSelfDescription']
    except:
        dataset_description = {
            "@type": "datacellar:DatasetField",
            "datacellar:datasetFieldID": 0,
            "datacellar:fieldName": "anomaly",
            "datacellar:fieldDescription": "Anomalies",
            "datacellar:type": {
                "@type": "datacellar:boolean",
                "datacellar:unitText": "-"
            }
        }
    
    # Process output series
    try:
        for series in json_data['datacellar:timeSeriesList']:
            # Check if series has any data points
            if series.get('datacellar:timeseriesDataPoints'):
                df, unit, _ = process_series(series, dataset_description)
                # Additional check for non-empty DataFrame
                if not df.empty and df['datacellar:value'].notna().any():
                    if unit not in series_by_unit:
                        series_by_unit[unit] = []
                    series_by_unit[unit].append(df)
    except Exception as e:
        st.error(f"Error processing series: {str(e)}")
    
    # Process input series if provided
    if input_data:
        input_description = input_data['datacellar:datasetSelfDescription']
        for series in input_data['datacellar:timeSeriesList']:
            if series.get('datacellar:timeseriesDataPoints'):
                df, unit, _ = process_series(series, input_description, is_input=True)
                if not df.empty and df['datacellar:value'].notna().any():
                    if unit not in series_by_unit:
                        series_by_unit[unit] = []
                    series_by_unit[unit].append(df)
    
    # Concatenate and filter out units with no valid data
    result = {}
    for unit, dfs in series_by_unit.items():
        if dfs:  # Check if there are any DataFrames for this unit
            combined_df = pd.concat(dfs)
            if not combined_df.empty and combined_df['datacellar:value'].notna().any():
                result[unit] = combined_df
    
    return result

def create_time_series_plot(df, unit, service_type=None,fig=None):
    """
    Create visualization for time series data, handling empty series appropriately.
    """
    if service_type == "Anomaly Detection":

        if not fig:
            fig = go.Figure()
        
        # Filter for non-empty input data
        input_data = df[df['series_id'].str.contains('Input')]
        input_data = input_data[input_data['datacellar:value'].notna()]
        
        if not input_data.empty:
            fig.add_trace(go.Scatter(
                x=input_data['datacellar:timeStamp'], 
                y=input_data['datacellar:value'],
                mode='lines',
                name='Energy Consumption',
                line=dict(color='blue')
            ))
        
            # Handle anomalies
            anomalies = df[(~df['series_id'].str.contains('Output')) & 
                          (df['datacellar:value'] == True) &
                          (df['datacellar:value'].notna())]
            if not anomalies.empty:
                anomaly_values = []
                for timestamp in anomalies['datacellar:timeStamp']:
                    value = input_data.loc[input_data['datacellar:timeStamp'] == timestamp, 'datacellar:value']
                    anomaly_values.append(value.iloc[0] if not value.empty else None)
                
                # fig.add_trace(go.Scatter(
                #     x=anomalies['datacellar:timeStamp'],
                #     y=anomaly_values,
                #     mode='markers',
                #     name='Anomalies',
                #     marker=dict(color='red', size=10)
                # ))
        
        fig.update_layout(
            title=f'Time Series Data with Anomalies ({unit})',
            xaxis_title="Time",
            yaxis_title=f"Value ({unit})",
            hovermode='x unified',
            legend_title="Series"
        )
        return fig
    else:
        # Filter out series with no valid data
        valid_series = []
        for series_id in df['series_id'].unique():
            series_data = df[df['series_id'] == series_id]
            if not series_data.empty and series_data['datacellar:value'].notna().any():
                valid_series.append(series_id)
        
        # Create plot only for valid series
        if valid_series:
            filtered_df = df[df['series_id'].isin(valid_series)]
            return px.line(
                filtered_df,
                x='datacellar:timeStamp',
                y='datacellar:value',
                color='series_id',
                title=f'Time Series Data ({unit})'
            ).update_layout(
                xaxis_title="Time",
                yaxis_title=f"Value ({unit})",
                hovermode='x unified',
                legend_title="Series"
            )
        else:
            # Return None or an empty figure if no valid series
            return None

def display_statistics(dfs_by_unit):
    """
    Display statistics only for non-empty series.
    """
    for unit, df in dfs_by_unit.items():
        st.write(f"## Measurements in {unit}")
        for series_id in df['series_id'].unique():
            series_data = df[df['series_id'] == series_id]
            # Check if series has valid data
            if not series_data.empty and series_data['datacellar:value'].notna().any():
                st.write(f"### {series_id}")
                
                cols = st.columns(4)
                metrics = [
                    ("Average", series_data['datacellar:value'].mean()),
                    ("Max", series_data['datacellar:value'].max()),
                    ("Min", series_data['datacellar:value'].min()),
                    ("Total", series_data['datacellar:value'].sum() * 6/3600)
                ]
                
                for col, (label, value) in zip(cols, metrics):
                    with col:
                        unit_suffix = "h" if label == "Total" else ""
                        st.metric(label, f"{value:.2f} {unit}{unit_suffix}")

def call_api(file_content, token, service_endpoint):
    """
    Call the analysis API with the provided data.
    
    Args:
        file_content: Binary content of the JSON file
        token: API authentication token
        service_endpoint: String indicating which API endpoint to call
        
    Returns:
        dict: JSON response from the API or None if the call fails
    """
    try:
        url = f'https://loki.linksfoundation.com/datacellar/{service_endpoint}'
        response = requests.post(
            url,
            headers={'Authorization': f'Bearer {token}'},
            files={'input_file': ('data.json', file_content, 'application/json')}
        )
        
        if response.status_code == 401:
            st.error("Authentication failed. Please check your API token.")
            return None
            
        return response.json()
    except Exception as e:
        st.error(f"API Error: {str(e)}")
        return None

def get_dataset_type(json_data):
    """
    Determine the type of dataset from its description.
    
    Args:
        json_data: Dictionary containing the JSON data
        
    Returns:
        str: "production", "consumption", or "other"
    """
    desc = json_data.get('datacellar:description', '').lower()
    if 'production' in desc:
        return "production"
    elif 'consumption' in desc:
        return "consumption"
    return "other"

def get_forecast_horizon(json_data):
    """
    Determine the forecast horizon from dataset description.
    
    Args:
        json_data: Dictionary containing the JSON data
        
    Returns:
        str: "long", "short", or None
    """
    desc = json_data.get('datacellar:description', '').lower()
    if 'long term' in desc:
        return "long"
    elif 'short term' in desc:
        return "short"
    return None