Add simplified histogram creation script

#!/usr/bin/env python3

import argparse

import math

import numpy as np

# These next three lines are added as a precaution in case the gitlab runner

# needs DISPLAY to render the plots, even if they are written to file.

import matplotlib

matplotlib.use("Agg")

import matplotlib.pyplot as plt

import csv

import os

from parse_xml_report import IVAS_FORMATS, EVS_FORMATS, IVAS_CATEGORIES, EVS_CATEGORIES

"""

Parses a CSV report and creates a summary report.

"""

def create_histogram(results_sorted, diff, measure, outdir):    

    if diff:

        limits_per_measure = {

            "MLD": ("MLD", None),

            "DIFF": ("MAXIMUM ABS DIFF", None),

            "SSNR": ("MIN_SSNR", None),

            "ODG": ("MIN_ODG", None),

            "DELTA_ODG": ("DELTA_ODG", None),

        }

    else:

        limits_per_measure = {

            "MLD": ("MLD", [0, 1, 2, 3, 4, 5, 10, 20, math.inf]),

            "DIFF": ("MAXIMUM ABS DIFF", [0, 16, 256, 1024, 2048, 4096, 8192, 16384, 32769]),

            "SSNR": ("MIN_SSNR", [-math.inf, 0, 10, 20, 30, 40, 40, 50, 60, 100]),

            "ODG": ("MIN_ODG", [-5, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5]),

            "DELTA_ODG": ("DELTA_ODG", [-5, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5]),

        }

    (measure_label, limits) = limits_per_measure[measure]

    if limits is None:

        vals = [float(x) for x in [m[measure_label] for m in results_sorted.values() if m[measure_label] != "None" and m[measure_label] != ""]]

        start = min(vals)

        f = 10 ** (2 - int(np.floor(np.log10(abs(start)))) - 1)

        start = np.floor(start*f)/f

        step = (max(vals) - start)/10

        f = 10 ** (2 - int(np.floor(np.log10(abs(step)))) - 1)

        step = np.ceil(step*f)/f

        limits = np.arange(start, 10*step, step)

    # Output CSV file

    with open(os.path.join(outdir, f"histogram_{measure}.csv"), "w") as fp:

        limits_labels = [f"{a:g}" for a in limits] + ["","None"] # Put None cases in separate bin

        headerline = f"Format;Category;" + ";".join(limits_labels) + "\n"

        fp.write(headerline)

        for fmt in FORMATS:

            fig, ax = plt.subplots()

            bottom = np.zeros(len(limits_labels))

            for cat in CATEGORIES:

                values = [

                    x

                    for x in [

                        m[measure_label]

                        for m in results_sorted.values()

                        if m["Format"] == fmt and m["Category"] == cat

                    ]

                ]

                # Create separate bin for None (errors)

                val = [float(x) for x in values if x != "None" and x != ""]

                none = [sum([1 for x in values if x == "None" or x == ""])]

                hist, _ = np.histogram(val, limits)

                data = np.array(list(hist) + [0] + none + [0])

                # CSV output

                line = f"{fmt};{cat};{'; '.join(map(str,data))}\n"

                fp.write(line)

                # Matplotlib histogram

                ax.bar(limits_labels, data, 1, align='edge', edgecolor='black', linewidth=0.5, label=cat, bottom=bottom)

                bottom += data

            # Histogram layout

            ax.set_title(fmt)

            ax.legend(loc="best")

            ax.set_xlabel(measure_label)

            if "DIFF" in measure_label:

                ax.set_xticks(range(len(limits_labels)), limits_labels, rotation=35)

            ax.set_ylabel("Number of test cases")

            fig.set_figheight(4)

            fig.set_figwidth(6)

            plt.savefig( os.path.join(outdir,f"histogram_{fmt}_{measure}.png"))

            plt.close()

# Main routine

if __name__ == "__main__":

    parser = argparse.ArgumentParser(

        description="Parses a CSV report and creates a summary report."

    )

    parser.add_argument(

        "csv_report",

        type=str,

        help="Input CSV report with test cases, e.g. report.csv",

    )    

    parser.add_argument(

        "--outdir",

        type=str,

        nargs=1,

        help="Output directory",

        default=["."],

    )

    parser.add_argument(""

        "--html_out",

        type=str,

        nargs=1,

        help="HTML output file, default: index.html",

        default=["index.html"],

    )

    parser.add_argument(""

        "--id",

        type=int,

        nargs=1,

        help="Job ID, default: 0",

        default=0,

    )

    parser.add_argument(""

        "--job",

        type=str,

        nargs=1,

        help="Job name, default: jobname",

        default=["jobname"],

    )        

    parser.add_argument(

        "--measures",

        type=str,

        nargs="+",

        help="Measure, any of: MLD, DIFF, SSNR, ODG, default: MLD DIFF SSNR ODG",

        default=["MLD","DIFF","SSNR","ODG"],

    )

    parser.add_argument(

        "--evs",

        action="store_true",

        help="Parse using EVS 26.444 formats",

        default=False,

    )

    parser.add_argument(

        "--diff",

        action="store_true",

        help="Use limits for diff scores",

        default=False,

    )

    args = parser.parse_args()

    csv_report = args.csv_report

    measures = args.measures

    outdir = args.outdir[0]

    if args.evs:

        FORMATS = EVS_FORMATS

        CATEGORIES = EVS_CATEGORIES

    else:

        FORMATS = IVAS_FORMATS

        CATEGORIES = IVAS_CATEGORIES

    # Load CSV report

    results_sorted = {}

    with open(csv_report, "r") as fp:

        reader = csv.reader(fp, delimiter=";")

        header = next(reader)

        keys = header[1:]

        for row in reader:

            testcase = row[0]

            results_sorted[testcase] = {}

            for k, val in zip(keys, row[1:]):

                results_sorted[testcase][k] = val

    for measure in measures:

        create_histogram(results_sorted, args.diff, measure, outdir)