plot_relationship.py

def color_selector(i, s):
    """
    Select a color index based on the model name.

    This function returns a color index based on the specified model name.
    If the model name is recognized, a predefined index is returned;
    otherwise, the input index is returned.

    Args:
        i (int): The default index to return if the model name is not recognized.
        s (str): The name of the model. Possible values include:
            - 'EMOD'
            - 'malariasimulation'
            - 'OpenMalaria'

    Returns:
        int: The color index corresponding to the model name.
    """

    if s == 'EMOD':
        return 0
    elif s == 'malariasimulation':
        return 1
    elif s == 'OpenMalaria':
        return 2
    else:
        return i

def convert_to_date(x):
    """
    Convert a number of days since January 1, 2005, to a date.

    This function takes an integer representing the number of days
    since January 1, 2005, and returns the corresponding date.

    Args:
        x (int): The number of days since January 1, 2005.

    Returns:
        date: A datetime.date object representing the corresponding date.
    """

    import datetime
    return datetime.date(2005, 1, 1) + datetime.timedelta(days=x)

def custom_sort_key(age_group):
    """
    Custom sort key function for sorting age groups.

    This function extracts the lower bound of an age group represented as
    a string in the format 'X-Y' and returns it as an integer. It is
    primarily used for sorting age groups in ascending order based on
    their lower bounds.

    Args:
        age_group (str): The age group string in the format 'X-Y',
                         where X is the lower bound and Y is the upper bound.

    Returns:
        int: The lower bound of the age group as an integer.
    """

    return int(age_group.split('-')[0])

def eir_to_outcome(fdir, df, sweepvar='cm_clinical', facet_var='seasonality', eir_val='simulatedEIR',
                   channel='prevalence_2to10', agegrp='0-5'):
    """
    Generate line plots for EIR (Entomological Inoculation Rate) and a requested
    outcome variable, with models represented as colors and sweep variables as panels.

    This function creates line plots where the x-axis represents the EIR and the
    y-axis represents an outcome variable, with different models indicated by color
    and organized into panels based on specified facets.

    Args:
        fdir (str): Directory where the generated plot will be saved.
        df (pd.DataFrame): DataFrame that includes combined model results.
        sweepvar (str, optional): Variable to group the data and create multiple
            panels on the plot. Default is 'cm_clinical'.
        eir_val (str, optional): The EIR (Entomological Inoculation Rate) value to
            be used for plotting. Default is 'simulatedEIR'.
        channel (str, optional): The outcome variable to compare to EIR. Default is
            'prevalence_2to10'.
        agegrp (str, optional): Limits the resulting graphs to the selected age group.
            Default is '0-5'.

    Returns:
        None: The function saves the generated plots to disk.
    """

    figure_vars = [eir_val, channel] + [sweepvar, facet_var]
    df, caption_txt = subset_dataframe_for_plot(df, figure_vars, agegrp)
    color_palette = sns.color_palette('colorblind', max(len(df[sweepvar].unique()), 4))

    firstPlot = True
    unique_facets = df[facet_var].unique()
    nx = len(df[sweepvar].unique())
    ny = len(unique_facets)
    f = 1
    fig = plt.figure(figsize=(10 * nx, 10 * ny))

    for fi in unique_facets:
        ax = fig.add_subplot(ny, nx, f)
        ax.set_title(fi)
        f += 1
        fdf = df[(df[facet_var] == fi)]

        for i, (p, pdf) in enumerate(fdf.groupby([sweepvar])):
            color_key = color_selector(i, p)
            xmean, ymean = get_x_y(pdf, 'output_target', eir_val, channel)
            merge_df = pd.merge(left=xmean, right=ymean, on='output_target')
            merge_df.sort_values(by=eir_val, inplace=True)
            ax.plot(merge_df[eir_val], merge_df[channel], '-', linewidth=0.8, label=f"{p}",
                    color=color_palette[color_key])
            ax.fill_between(merge_df[eir_val], merge_df[f'{channel}_min'], merge_df[f'{channel}_max'], alpha=0.1,
                            color=color_palette[color_key])
        if firstPlot:
            lg = ax.legend(loc='upper left', bbox_to_anchor=(0, 1))
            firstPlot = False
        ax.set_xlim(0.1, 1000)
        ax.set_xscale('symlog')
        ax.set_xlabel(f'{eir_val.replace("EIR", " annual EIR")}', fontsize=14)
        ax.set_ylabel(get_ylab(channel))

    fname = f'{eir_val}_{channel}_{agegrp}_by_{sweepvar}_{facet_var}'
    if channel == 'prevalence_2to10':
        fname = f'{eir_val}_{channel}_by_{sweepvar}_{facet_var}'
    fname = fname.replace('modelname', 'model')

    fig.savefig(os.path.join(fdir, f'{fname}.png'), bbox_extra_artists=(lg,), bbox_inches='tight')
    plt.close()

def get_output_df(wdir, modelname, yr=False, mth=False, daily=False, custom_name=None,
                  save_combined=False):
    """
    Load and combine data from the model output files.

    This function reads model output files from a specified working directory
    and combines the data into a single DataFrame. It supports different data
    formats based on the specified parameters for yearly, monthly, or daily
    data.

    Args:
        wdir (str): Working directory where the data files are located.
        modelname (str or list of str): Name of models for which result CSVs
                                         should be loaded (case sensitive).
        yr (bool, optional): Set to True if the data files have yearly data.
                             Defaults to False.
        mth (bool, optional): Set to True if the data files have monthly data.
                             Defaults to False.
        daily (bool, optional): Set to True if the data files have daily timestep
                                data. Defaults to False. If both mth and daily
                                are True, only daily will be processed.
        custom_name (str, optional): Custom filename to use instead of the default
                                      based on the time period. Defaults to None.
        save_combined (bool, optional): Set to True to save the combined DataFrame
                                         to a CSV file. Defaults to False.

    Returns:
        tuple: A tuple containing:
            - df (DataFrame): Combined DataFrame containing the combined data
                              for the models listed in modelname.
            - wdir (str): Updated working directory (if applicable).

    Raises:
        ValueError: If an invalid modelname value is specified.
    """

    cols_to_keep = None  # default read all
    fname = 'mmmpy_timeavrg.csv'
    if yr:
        fname = 'mmmpy_yr.csv'
    if mth:
        fname = 'mmmpy_mth.csv'
    if daily:
        fname = 'mmmpy_daily.csv'
        # cols_to_keep = ['index', 'timestep', 'ageGroup', 'simulatedEIR', 'prevalence_2to10', 'prevalence',
        #                'clinical_incidence', 'severe_incidence', 'seed']
    if custom_name:
        fname = f'{custom_name}.csv'

    file_paths = [os.path.join(wdir, fname)]

    for model in modelname:
        file_paths.append(os.path.join(wdir, model, fname))

    existing_files = [path for path in file_paths if os.path.isfile(path)]

    if not existing_files:
        return pd.DataFrame(), wdir

    if os.path.isfile(os.path.join(wdir, fname)):
        df = pd.read_csv((os.path.join(wdir, fname)), low_memory=False)
    else:

        dfs = []
        for model in modelname:
            model_path = os.path.join(wdir, model, fname)
            try:
                if os.path.isfile(model_path):
                    df = pd.read_csv(model_path, usecols=cols_to_keep)
                    df['modelname'] = model
                    if model == 'EMOD':
                        df['seed'] = df['seed'] + 1
                    dfs.append(df)
                else:
                    print(f"File not found for {model}: {model_path}")
            except Exception as e:
                print(f"Error reading {model_path}: {e}")

        if not dfs:
            return pd.DataFrame(), wdir

        df = pd.concat(dfs, ignore_index=True)

        if 'ageGroup' in df.columns:
            try:
                age_grps = sorted(list(df['ageGroup'].unique()), key=custom_sort_key)
            except:
                age_grps = list(df['ageGroup'].unique())
            df['ageGroup'] = df['ageGroup'].astype('category')
            df['ageGroup'] = df['ageGroup'].cat.reorder_categories(age_grps)

        warning_df = df[df['simulatedEIR'] == 0]
        if len(warning_df) > 0 and daily is False:  # we don't want to include simulations were eir was 0 or less, because we won't get any outcome measures and that crashes the system
            print('Warning: some eirs had simulated EIRS of 0, and were removed')
            df = df[df['simulatedEIR'] > 0]
            df = df[df['simulatedEIR'].notnull()]

        if not daily and save_combined:
            df.to_csv(os.path.join(wdir, f'{fname}'), index=False)
    return df, wdir

def get_x_y(df, grpvar, x_channel, y_channel):
    """
    Calculate x-axis and y-axis values for each plot.

    This function groups the input DataFrame by a specified variable and
    calculates the mean values for the specified x and y channels. It also
    computes the 95% confidence interval for the y values.

    Args:
        df (DataFrame): The DataFrame used to group and calculate x and y values.
        grpvar (str): The variable in the DataFrame used to group the x and y values.
        x_channel (str): The variable serving as the x-axis in the graph.
        y_channel (str): The variable serving as the y-axis in the graph.

    Returns:
        tuple: A tuple containing:
            - xmean (DataFrame): A DataFrame containing values for the x-axis.
            - ymean (DataFrame): A DataFrame containing values for the y-axis,
                                 including the 95% confidence interval (min and max).
    """

    xmean = df.groupby(grpvar)[x_channel].agg(np.mean).reset_index()
    ymean = df.groupby(grpvar)[y_channel].agg(np.mean).reset_index()
    p_df = pd.DataFrame(columns=[grpvar, f'{y_channel}_min', f'{y_channel}_max'])
    for i, row in ymean.iterrows():
        p = df[df[grpvar] == row[grpvar]]
        pmin = np.nanpercentile(p[y_channel], 2.5, axis=0)
        pmax = np.nanpercentile(p[y_channel], 97.5, axis=0)
        new_row = pd.DataFrame([{grpvar: row[grpvar], f'{y_channel}_min': pmin, f'{y_channel}_max': pmax}])
        p_df = pd.concat([p_df, new_row], axis=0, ignore_index=True)
    ymean = pd.merge(left=ymean, right=p_df, on=grpvar)
    return xmean, ymean

def get_ylab(channel):
    """
    Retrieve the y-axis label for a given channel.

    This function returns a formatted string representing the y-axis label
    based on the specified channel name. The labels correspond to specific
    epidemiological measures.

    Args:
        channel (str): The name of the channel for which to retrieve the label.
            Possible values include:
            - 'prevalence_2to10'
            - 'prevalence'
            - 'clinical_incidence'
            - 'severe_incidence'
            - 'simulatedEIR'

    Returns:
        str: The corresponding y-axis label for the channel, or None if the
        channel is not recognized.
    """

    channel_labels = {'prevalence_2to10': '$\it{Pf}$PR$_{2-10}$ (%)',
                      'prevalence': '$\it{Pf}$PR (%)',
                      'clinical_incidence': 'Clinical incidence (pppy)',
                      'severe_incidence': 'Severe incidence (pppy)',
                      'simulatedEIR': 'simulated EIR'
                      }

    return channel_labels.get(channel)

def input_to_simulated_eir(fdir, df, sweepvar='cm_clinical', facet_var='seasonality'):
    """
    Generate line plots comparing input EIR to simulated EIR.

    This function creates line plots where the x-axis represents the input EIR values
    and the y-axis represents the simulated annual EIR. Different models are represented
    by different lines on the plot, and the plots are organized into panels based on
    specified facets.

    Args:
        fdir (str): Directory where the generated plot will be saved.
        df (pd.DataFrame): DataFrame that includes combined model results.
        sweepvar (str, optional): Variable to group the data and create multiple lines
            on the plot. Default is 'cm_clinical'.
        facet_var (str, optional): Variable to create multiple panels on the plot.
            Default is 'seasonality'.

    Returns:
        None: The function saves the generated plots to disk.
    """

    xyvars = ['simulatedEIR', 'output_target']
    figure_vars = xyvars + [sweepvar, facet_var]
    df, caption_txt = subset_dataframe_for_plot(df, figure_vars)
    unique_facets = df[facet_var].unique()
    unique_groups = df[sweepvar].unique()
    unique_groups = np.sort(unique_groups)
    num_colors = len(unique_groups)
    color_palette = sns.color_palette('colorblind', max(num_colors, 4))
    firstPlot = True
    nx = 1
    ny = len(unique_facets)
    f = 1
    fig = plt.figure(figsize=(10 * nx, 6 * ny))
    for fi in unique_facets:
        fdf = df[df[facet_var] == fi]

        ax = fig.add_subplot(ny, nx, f)
        ax.set_title(fi)
        f += 1
        for i, (p, pdf) in enumerate(fdf.groupby([sweepvar])):
            # plot mean, min and max of seeds and age groups
            color_key = color_selector(i, p)
            if p == 'EMOD':
                model = 'em'
            elif p == 'OpenMalaria':
                model = 'om'
            elif p == 'malariasimulation':
                model = 'ms'
            else:
                print(f'Warning: Model not recognized: {p}')
            xmean, ymean = get_x_y(pdf, 'output_target', f'model_input_{model}', 'simulatedEIR')
            merge_df = pd.merge(left=xmean, right=ymean, on='output_target')
            merge_df.sort_values(by=f'model_input_{model}', inplace=True)
            ax.plot(merge_df[f'model_input_{model}'], merge_df['simulatedEIR'], '-', linewidth=0.8, label=f"{p}",
                    color=color_palette[color_key])
            ax.fill_between(merge_df[f'model_input_{model}'], merge_df['simulatedEIR_min'],
                            merge_df['simulatedEIR_max'], alpha=0.1, color=color_palette[color_key])
        if firstPlot:
            lg = ax.legend(loc='upper left', bbox_to_anchor=(0, 1))
            firstPlot = False
        ax.set_ylim(0.1, 10000)
        ax.set_xlim(0.1, 10000)
        ax.set_yscale('symlog')
        ax.set_xscale('symlog')

    plt.xlabel('Input EIR')
    plt.ylabel('Simulated annual EIR')

    fname = f'input_to_simulated_eir_by_{sweepvar}_{facet_var}'
    fname = fname.replace('modelname', 'model')
    fig.savefig(os.path.join(fdir, f'{fname}.png'), bbox_extra_artists=(lg,), bbox_inches='tight')
    plt.close()

def parse_args():
    """
    Parses command-line arguments for simulation specifications.

    This function uses the argparse library to handle command-line inputs
    required for running simulation experiments. It defines required and optional
    arguments, including the job directory and model names.

    Returns:
        argparse.Namespace: An object containing the parsed command-line arguments.

    Command Line Arguments:
        -d/--directory (str): The job directory where the exp.obj file is located. This argument is required.
        -m/--modelname (str): One or more model names to compare. This argument is optional
                              and defaults to ['EMOD', 'OpenMalaria', 'malariasimulation'].
    """

    description = "Simulation specifications"
    parser = argparse.ArgumentParser(description=description)

    parser.add_argument(
        "-d",
        "--directory",
        type=str,
        required=True,
        help="Job Directory where exp.obj is located",
    )
    parser.add_argument(
        "-m",
        "--modelname",
        nargs='+',
        type=str,
        required=False,
        help="Name of models to compare",
        default=['EMOD', 'OpenMalaria', 'malariasimulation']
    )

    return parser.parse_args()

def prevalence2to10_to_incidence(fdir, df, sweepvar='modelname', facet_var='seasonality', channel='clinical_incidence',
                                 agegrps=None):
    """
    Generate line plots for PfPR2to10 and either clinical or severe incidence,
    grouped by the specified sweep variable and faceted by another variable.

    This function creates a series of line plots where the x-axis represents
    the prevalence of PfPR2to10, and the y-axis represents either clinical or
    severe incidence. Each line corresponds to a model, and the plots can be
    faceted by a specified variable (e.g., seasonality).

    Args:
        fdir (str): Directory where the generated plot will be saved.
        df (pd.DataFrame): DataFrame containing the model results for plotting.
        sweepvar (str, optional): Variable to group the data and create multiple panels
            on the plot. Default is 'modelname'.
        facet_var (str, optional): Variable used to create subplots based on its unique
            values. Default is 'seasonality'.
        channel (str): Variable representing the y-axis data to be plotted (e.g.,
            'clinical_incidence').
        agegrps (list of str, optional): Specific age groups to filter the DataFrame
            before plotting. Defaults to None, which means all age groups will be used.

    Returns:
        None: The function saves the generated plots to disk.

    Raises:
        ValueError: If an error occurs while filtering the DataFrame or during plotting.
    """

    figure_vars = ['prevalence_2to10', channel] + [sweepvar, facet_var]
    df, caption_txt = subset_dataframe_for_plot(df, figure_vars, agegrps)

    unique_facets = df[facet_var].unique()
    nx = len(df['ageGroup'].unique())
    ny = 1
    f = 1
    firstPlot = True
    fig = plt.figure(figsize=(10 * nx, 10 * ny))

    unique_groups = df[sweepvar].unique()
    unique_groups = np.sort(unique_groups)
    num_colors = len(unique_groups)
    color_palette = sns.color_palette('colorblind', max(num_colors, 4))

    # for each age group, plot x = prevalence 2_to_10, y = incidences (or prevalence)
    for fi in unique_facets:
        fdf = df[df[facet_var] == fi]

        for agroup in fdf['ageGroup'].unique():
            ax = fig.add_subplot(ny, nx, f)
            ax.set_title(f'{fi}, {agroup}')
            f += 1
            adf = fdf[fdf.ageGroup == agroup]

            for mi, model in enumerate(unique_groups):
                pdf = adf[adf[sweepvar] == model]
                color_key = color_selector(mi, model)
                if not pd.isnull(pdf['prevalence_2to10']).all():
                    # plot mean, min and max of seeds
                    xmean, ymean = get_x_y(pdf, 'output_target', 'prevalence_2to10', channel)
                    merge_df = pd.merge(left=xmean, right=ymean, on='output_target')
                    if channel == 'simulatedEIR':
                        merge_df.sort_values(by='output_target', inplace=True)
                    else:
                        merge_df.sort_values(by='prevalence_2to10', inplace=True)
                    ax.plot(merge_df['prevalence_2to10'], merge_df[channel], label=model,
                            color=color_palette[color_key])
                    ax.fill_between(merge_df['prevalence_2to10'], merge_df[f'{channel}_min'],
                                    merge_df[f'{channel}_max'], alpha=0.1, color=color_palette[color_key])
            if channel == "prevalence":
                y_lim = 1
            else:
                y_lim = max(adf[channel]) * 1.1
            ax.set_ylim(0, y_lim)
            ax.set_xlim(0, 1)
            ax.set_xticks([0, 0.2, 0.4, 0.6, 0.8, 1])
            ax.title.set_text(f'{agroup}')
            ax.set_ylabel(get_ylab(channel))
            ax.set_xlabel('$\it{Pf}$PR$_{2-10}$ (%)')
            if firstPlot:
                lg = ax.legend(loc='upper left', bbox_to_anchor=(0.5, 1.2))
                firstPlot = False

    fname = f'pfpr2to10_to_{channel}'
    fig.savefig(os.path.join(fdir, f'{fname}.png'), bbox_extra_artists=(lg,), bbox_inches='tight')
    plt.close()

def prevalence2to10_to_incidence_by_age_and_model(fdir, df, sweepvar='modelname', channel='clinical_incidence'):
    """
    Generate line plots for PfPR2to10 and either clinical or severe incidence,
    separated by model (columns) and age groups (rows).

    This function creates a grid of line plots where each plot corresponds to a
    specific model and age group for the prevalence of PfPR2to10 on the x-axis
    and either clinical or severe incidence on the y-axis.

    Args:
        fdir (str): Directory where the generated plot will be saved.
        df (pd.DataFrame): DataFrame containing the model results for plotting.
        sweepvar (str, optional): Variable to group the data and create multiple
            panels on the plot. Default is 'modelname'.
        channel (str): Variable representing the y-axis data to be plotted (e.g.,
            'clinical_incidence').

    Returns:
        None: The function saves the generated plots to disk.
    """

    seasonality = list(df['seasonality'].unique())
    nx = 3
    ny = 3
    f = 1
    fig = plt.figure(figsize=(10 * nx, 10 * ny))

    unique_groups = df[sweepvar].unique()
    unique_groups = np.sort(unique_groups)
    num_colors = len(unique_groups)
    color_palette = sns.color_palette('colorblind', max(num_colors, 4))
    young_child = ['0-0.5', '0.5-1', '1-2', '2-5']
    older_child = ['5-10', '10-15']
    adult = ['15-20', '20-100']
    ages_list = [young_child, older_child, adult]
    mean_vars = [channel, 'prevalence_2to10']

    # for each age group, plot x = prevalence 2_to_10, y = incidences (or prevalence)
    for si, season in enumerate(seasonality):
        sdf = df[df.seasonality == season]
        for ages in ages_list:
            adf = sdf[sdf['ageGroup'].isin(ages)]
            for mi, model in enumerate(unique_groups):
                color_key = color_selector(mi, model)
                ax = fig.add_subplot(ny, nx, f)
                ax.set_title(f'{season} {model}')
                f += 1
                mdf = adf[adf.modelname == model]
                mdf = mdf.groupby(['seed', 'output_target'])[mean_vars].agg(np.mean).reset_index()
                if not pd.isnull(mdf['prevalence_2to10']).all():
                    # plot mean, min and max of seeds
                    xmean, ymean = get_x_y(mdf, 'output_target', 'prevalence_2to10', channel)
                    merge_df = pd.merge(left=xmean, right=ymean, on='output_target')
                    merge_df.sort_values(by='prevalence_2to10', inplace=True)
                    ax.plot(merge_df['prevalence_2to10'], merge_df[channel], label=model,
                            color=color_palette[color_key])
                    ax.fill_between(merge_df['prevalence_2to10'], merge_df[f'{channel}_min'],
                                    merge_df[f'{channel}_max'], alpha=0.1, color=color_palette[color_key])
                if channel == "prevalence":
                    y_lim = 1
                else:
                    y_lim = max(adf[channel]) * 1.01
                ax.set_ylim(0, y_lim)
                ax.set_xlim(0, 1)
                ax.set_xticks([0, 0.2, 0.4, 0.6, 0.8, 1])
                ax.title.set_text(f'{model}: {ages}')
                ax.set_ylabel(get_ylab(channel))
                ax.set_xlabel('$\it{Pf}$PR$_{2-10}$ (%)')
    fname = f'pfpr2to10_to_{channel}_by_age_by_model'
    fig.savefig(os.path.join(fdir, f'{fname}.png'), bbox_inches='tight')
    plt.close()

def subset_dataframe_for_plot(df, figure_vars, agegrp=None, filter_target=True):
    """
    Filter the input DataFrame for plotting based on specified criteria.

    This function filters the DataFrame according to the provided figure variables,
    optional age groups, and other selection criteria to prepare the data for
    visualization. It also returns a string summarizing the filtering applied.

    Args:
        df (pd.DataFrame): The input DataFrame containing simulation results.
        figure_vars (list of str): List of variables used for plotting, which influences
            the filtering process.
        agegrp (str or list of str, optional): Specific age group(s) to filter by.
            If provided, only the data for these age groups will be retained.
            Defaults to None, meaning no filtering by age group will occur.
        filter_target (bool, optional): If True, the function will filter the DataFrame
            to retain the maximum output target value if certain variables are not
            present in `figure_vars`. Defaults to True.

    Returns:
        tuple: A tuple containing:
            - pd.DataFrame: The filtered DataFrame.
            - str: A summary string describing the filtering that was applied.

    Raises:
        ValueError: If 'modelname' is not included in `figure_vars` and there
        are multiple unique model names in the DataFrame.
    """

    txt = 'Filtered dataset by: '

    if agegrp is not None:
        if 'ageGroup' not in figure_vars:
            if isinstance(agegrp, list):
                df = df[df['ageGroup'].isin(agegrp)]
                txt += f'ageGroup in {agegrp}, '
            else:
                df = df[df['ageGroup'] == agegrp]
                txt += f'ageGroup {agegrp}, '

    if 'cm_clinical' not in figure_vars:
        selected_cm = df['cm_clinical'].min()
        df = df[df['cm_clinical'] == selected_cm]
        txt += f'cm_clinical {selected_cm}, '

    if 'seasonality' not in figure_vars:
        selected_season = 'seasonal' if 'seasonal' in df['seasonality'].unique() else df['seasonality'].unique()[0]
        df = df[df['seasonality'] == selected_season]
        txt += f'seasonality {selected_season}, '

    if filter_target:
      if not any(var in figure_vars for var in ['simulatedEIR', 'prevalence_2to10', 'output_target']) :
          selected_output = df['output_target'].max()
          df = df[df['output_target'] == selected_output]
          txt += f'output_target {selected_output}, '

    if 'modelname' not in figure_vars and df['modelname'].nunique() > 1:
        raise ValueError('modelname needs to be specified in plot if results were combined for more than 1 model')

    # Remove trailing comma and space if any filtering has been done
    if txt.endswith(', '):
        txt = txt[:-2]

    return df, txt