source: benchmark/plot.py @ f835806

#!/usr/bin/python3
"""
Python Script to plot values obtained by the rmit.py script
Runs a R.I.P.L.

./plot.py
-t trials
-o option:values
"""

import argparse
import itertools
import json
import math
import numpy
import re
import statistics
import sys

import matplotlib.pyplot as plt
from matplotlib.ticker import EngFormatter

class Field:
        def __init__(self, unit, _min, _log):
                self.unit = unit
                self.min  = _min
                self.log  = _log

field_names = {
        "ns per ops"            : Field('ns'    , 0, False),
        "Number of processors"  : Field(''      , 1, False),
        "Ops per procs"         : Field('Ops'   , 0, False),
        "Ops per threads"       : Field('Ops'   , 0, False),
        "ns per ops/procs"      : Field('ns'    , 0, False),
        "Number of threads"     : Field(''      , 1, False),
        "Total Operations(ops)" : Field('Ops'   , 0, False),
        "Ops/sec/procs"         : Field('Ops'   , 0, False),
        "Total blocks"          : Field('Blocks', 0, False),
        "Ops per second"        : Field('Ops'   , 0, False),
        "Cycle size (# thrds)"  : Field('thrd'  , 1, False),
        "Duration (ms)"         : Field('ms'    , 0, False),
        "Target QPS"            : Field(''      , 0, False),
        "Actual QPS"            : Field(''      , 0, False),
        "Average Read Latency"  : Field('us'    , 0, True),
        "Median Read Latency"   : Field('us'    , 0, True),
        "Tail Read Latency"     : Field('us'    , 0, True),
        "Average Update Latency": Field('us'    , 0, True),
        "Median Update Latency" : Field('us'    , 0, True),
        "Tail Update Latency"   : Field('us'    , 0, True),
        "Update Ratio"          : Field('\%'    , 0, False),
}

def plot(in_data, x, y, out):
        fig, ax = plt.subplots()
        colors = itertools.cycle(['#0095e3','#006cb4','#69df00','#0aa000','#fb0300','#e30002','#fd8f00','#ff7f00','#8f00d6','#4b009a','#ffff00','#b13f00'])
        series = {} # scatter data for each individual data point
        groups = {} # data points for x value

        print("Preparing Data")

        for entry in in_data:
                name = entry[0]
                if not name in series:
                        series[name] = {'x':[], 'y':[]}

                if not name in groups:
                        groups[name] = {}

                if x in entry[2] and y in entry[2]:
                        xval = entry[2][x]
                        yval = entry[2][y]
                        series[name]['x'].append(xval)
                        series[name]['y'].append(yval)

                        if not xval in groups[name]:
                                groups[name][xval] = []

                        groups[name][xval].append(yval)

        print("Preparing Lines")

        lines = {} # lines from groups with min, max, median, etc.
        for name, data in groups.items():
                if not name in lines:
                        lines[name] = { 'x': [], 'min':[], 'max':[], 'med':[], 'avg':[] }

                for xkey in sorted(data):
                        ys = data[xkey]
                        lines[name]['x']  .append(xkey)
                        lines[name]['min'].append(min(ys))
                        lines[name]['max'].append(max(ys))
                        lines[name]['med'].append(statistics.median(ys))
                        lines[name]['avg'].append(statistics.mean(ys))

        print("Making Plots")

        for name, data in sorted(series.items()):
                _col = next(colors)
                plt.scatter(data['x'], data['y'], color=_col, label=name, marker='x')
                plt.plot(lines[name]['x'], lines[name]['min'], '--', color=_col)
                plt.plot(lines[name]['x'], lines[name]['max'], '--', color=_col)
                plt.plot(lines[name]['x'], lines[name]['med'], '-', color=_col)

        print("Calculating Extremums")

        mx = max([max(s['x']) for s in series.values()])
        my = max([max(s['y']) for s in series.values()])

        print("Finishing Plots")

        plt.ylabel(y)
        # plt.xticks(range(1, math.ceil(mx) + 1))
        plt.xlabel(x)
        plt.grid(b = True)
        ax.xaxis.set_major_formatter( EngFormatter(unit=field_names[x].unit) )
        if field_names[x].log:
                ax.set_xscale('log')
        else:
                plt.xlim(field_names[x].min, mx + 0.25)

        ax.yaxis.set_major_formatter( EngFormatter(unit=field_names[y].unit) )
        if field_names[y].log:
                ax.set_yscale('log')
        else:
                plt.ylim(field_names[y].min, my*1.2)

        plt.legend(loc='upper left')

        print("Results Ready")
        if out:
                plt.savefig(out)
        else:
                plt.show()


if __name__ == "__main__":
        # ================================================================================
        # parse command line arguments
        parser = argparse.ArgumentParser(description='Python Script to draw R.M.I.T. results')
        parser.add_argument('-f', '--file', nargs='?', type=argparse.FileType('r'), default=sys.stdin, help="Input file")
        parser.add_argument('-o', '--out', nargs='?', type=str, default=None, help="Output file")
        parser.add_argument('-y', nargs='?', type=str, default="", help="Which field to use as the Y axis")
        parser.add_argument('-x', nargs='?', type=str, default="", help="Which field to use as the X axis")

        options =  parser.parse_args()

        # ================================================================================
        # load data
        try :
                data = json.load(options.file)
        except :
                print('ERROR: could not read input', file=sys.stderr)
                parser.print_help(sys.stderr)
                sys.exit(1)

        # ================================================================================
        # identify the keys

        series = set()
        fields = set()

        for entry in data:
                series.add(entry[0])
                for label in entry[2].keys():
                        fields.add(label)

        if not options.out :
                print(series)
                print("fields: ", ' '.join(fields))

        wantx = "Number of processors"
        wanty = "ns per ops"

        if options.x:
                if options.x in field_names.keys():
                        wantx = options.x
                else:
                        print("Could not find X key '{}', defaulting to '{}'".format(options.x, wantx))

        if options.y:
                if options.y in field_names.keys():
                        wanty = options.y
                else:
                        print("Could not find Y key '{}', defaulting to '{}'".format(options.y, wanty))


        plot(data, wantx, wanty, options.out)