root_handler.py

 
 
import numpy as np
import uproot
 
import pandas as pd
 
 
class TrackingValidationResult:
    """This class represents a loaded validation root file. It has methods for plotting the typically needed graphs."""
 
    def __init__(self, filename, label=None, color_index=0, additional_information=None):
        """Create a new validation result from the given filename.
           Additional options for plotting (e.g. color or label) can be given."""
        
        self.filename = filename
        
        self.pr_data = uproot.open(self.filename)["pr_tree/pr_tree"].arrays(library="pd")
        self.pr_data["is_prompt"] = (
            np.sqrt(self.pr_data.x_truth ** 2 + self.pr_data.y_truth ** 2) < 0.5) & (self.pr_data.is_primary == 1)
        
        self.mc_data = uproot.open(self.filename)["mc_tree/mc_tree"].arrays(library="pd")
        self.mc_data["is_prompt"] = (
            np.sqrt(self.mc_data.x_truth ** 2 + self.mc_data.y_truth ** 2) < 0.5) & (self.mc_data.is_primary == 1)
 
        
        self.mc_prompts = self.mc_data[self.mc_data.is_prompt == 1]
        
        self.pr_prompts = self.pr_data[self.pr_data.is_prompt == 1]
 
        import seaborn as sb
 
        colors = sb.color_palette()
 
        
        self.label = label
        
        self.color = colors[color_index % len(colors)]
 
        
        self.finding_efficiency = None
        
        self.hit_efficiency = None
        
        self.fake_rate = None
        
        self.clone_rate = None
        self.get_figure_of_merits()
 
        
        self.additional_information = additional_information
 
    @staticmethod
    def from_calculations(calculations, key="output_file_name", parameter_part=None):
        """Create validation results from an ipython calculation."""
        if parameter_part:
            return [
                TrackingValidationResult(
                    c.get(key),
                    label=c.get_parameters()[parameter_part],
                    color_index=i) for i,
                c in enumerate(calculations)]
        else:
            return [
                TrackingValidationResult(
                    c.get(key),
                    label=c.get_parameters(),
                    color_index=i) for i,
                c in enumerate(calculations)]
 
    def get_figure_of_merits(self):
        """Return the figures of merit from the file. Mosty used for internal seeting of the properties."""
        if self.finding_efficiency is None:
            overview = uproot.open(
                self.filename)["ExpertMCSideTrackingValidationModule_overview_figures_of_merit"].arrays(
                library="pd")
            self.finding_efficiency = overview.finding_efficiency[0]
            self.hit_efficiency = overview.hit_efficiency[0]
 
            overview = uproot.open(
                self.filename)["ExpertPRSideTrackingValidationModule_overview_figures_of_merit"].arrays(
                library="pd")
            self.clone_rate = overview.clone_rate[0]
            self.fake_rate = overview.fake_rate[0]
 
        return dict(finding_efficiency=self.finding_efficiency,
                    hit_efficiency=self.hit_efficiency,
                    clone_rate=self.clone_rate,
                    fake_rate=self.fake_rate)
 
    def get_figures_of_merit_latex(self):
        """Print out the figures of merit as a LaTeX-ready table."""
        results = self.get_figure_of_merits()
 
        latex_string = r'\begin{table}' + "\n"
        latex_string += r'  \begin{tabular}{cc} \toprule' + "\n"
        latex_string += r'    & \\ \midrule' + "\n"
        latex_string += r'    Finding Efficiency & ' + f"{100 * results['finding_efficiency']:.2f}" + r' \% \\' + "\n"
        latex_string += r'    Hit Efficiency & ' + f"{100 * results['hit_efficiency']:.2f}" + r' \% \\' + "\n"
        latex_string += r'    Fake Rate & ' + f"{100 * results['fake_rate']:.2f}" + r' \% \\' + "\n"
        latex_string += r'    Clone Rate & ' + f"{100 * results['clone_rate']:.2f}" + r' \% \\ \bottomrule' + "\n"
        latex_string += r'  \end{tabular}' + "\n"
        latex_string += r'\end{table}'
 
        return latex_string
 
    def plot_efficiency_point(self):
        """Plot a oint in the finding-efficiency/hit-efficiency plane."""
        import matplotlib.pyplot as plt
        self.plot(100 * self.finding_efficiency, 100 * self.hit_efficiency, loc=3)
        plt.xlabel("finding efficiency")
        plt.ylabel("hit efficiency")
 
    def grouped_by_pt_data(self, mc_data=None):
        """Convenience function to return the input data (or the internal mc_data) grouped by pt."""
        if mc_data is None:
            mc_data = self.mc_data
 
        pt_values = pd.cut(mc_data.pt_truth, np.linspace(mc_data.pt_truth.min(), mc_data.pt_truth.max(), 10))
        grouped = mc_data.groupby(pt_values)
 
        return grouped
 
    def plot(self, data_x, data_y, loc=4, yerr=None):
        """Plot data_y over data_x with the correct settings for this result. Mostly used internally."""
        import matplotlib.pyplot as plt
        if yerr is not None:
            plt.errorbar(data_x, data_y, ls="-", marker="o",
                         color=self.color, label=self.label, yerr=yerr, lw=4)
        else:
            plt.plot(data_x, data_y, ls="-", marker="o",
                     color=self.color, label=self.label, lw=4)
 
        if self.label is not None:
            plt.legend(loc=loc, frameon=True)
 
    def plot_finding_efficiency(self, data=None):
        """Plot the finding efficiency over pt."""
        import matplotlib.pyplot as plt
        grouped = self.grouped_by_pt_data(data)
 
        self.plot(grouped.median().pt_truth, grouped.mean().is_matched, yerr=1 / np.sqrt(grouped.count().is_matched))
        plt.xlabel(r"$p_T$ of the MC tracks (in GeV)")
        plt.ylabel("Finding Efficiency")
 
    def plot_hit_efficiency(self, data=None):
        """Plot the hit efficiency over pt."""
        import matplotlib.pyplot as plt
        grouped = self.grouped_by_pt_data(data)
 
        self.plot(grouped.median().pt_truth, grouped.mean().hit_efficiency, yerr=1 / np.sqrt(grouped.sum().mc_number_of_hits))
        plt.xlabel(r"$p_T$ of the MC tracks (in GeV)")
        plt.ylabel("Hit Efficiency")
 
    def print_useful_information(self):
        """Print mostfully useful information about this result."""
        pr_data = self.pr_data
        mc_data = self.mc_data
        primaries = pr_data[self.pr_data.is_prompt == 1]
        primaries_mc = mc_data[self.mc_data.is_prompt == 1]
 
        print(self.label)
        print("Fake", 100 * primaries.is_fake.mean(), 100 * pr_data.is_fake.mean())
        print("Clone", 100 * primaries.is_clone.mean(), 100 * pr_data.is_clone.mean())
        print("Ghost", 100 * primaries.is_ghost.mean(), 100 * pr_data.is_ghost.mean())
        print("Fitted", 100 * primaries.is_fitted.mean(), 100 * pr_data.is_fitted.mean())
        print("Found", 100 * primaries_mc.is_matched.mean(), 100 * mc_data.is_matched.mean())
        print("Found2", 100.0 - 100 * primaries_mc.is_missing.mean(), 100.0 - 100 * mc_data.is_missing.mean())
        print("Merged", 100 * primaries_mc.is_merged.mean(), 100 * mc_data.is_merged.mean())
        print("Hit-Eff", 100 * primaries_mc.hit_efficiency.mean(), 100 * mc_data.hit_efficiency.mean())
        print("Wrong Hits", primaries.number_of_wrong_hits.mean(), pr_data.number_of_wrong_hits.mean())
 
    def append_to_dataframe(self, df):
        """Append the main results to a already consisting dataframe."""
        result = {"finding_efficiency": self.finding_efficiency,
                  "hit_efficiency": self.hit_efficiency,
                  "clone_rate": self.clone_rate,
                  "fake_rate": self.fake_rate,
                  "file_name": self.filename}
        if self.additional_information:
            result.update(self.additional_information)
        return df.append(result, ignore_index=True)