#!/usr/bin/env python3

# Parse the log files output by the simulator to get the overall average
# and stddev of total bytes sent+received per epoch for each of:
# - dirauths
# - total relays
# - bootstrapping fallback relays per bw
# - non-bootstrapping fallback relays per bw
# - bootstrapping non-fallback relays per bw
# - non-bootstrapping non-fallback relays per bw
# - total clients
# - bootstrapping clients
# - non-bootstrapping clients
# - steady-state dirauths (skipping the first epoch in which all clients
#       are bootstrapping)
# - steady-state relays (as above)
# - steady-state clients (as above)
# - steady-state total relay traffic per client (as above)

# Pass the names of the log files as command-line arguments, or the log
# files themselves on stdin.
# The output will be five files named:
# - vanilla_none.dat
# - telescoping_merkle.dat
# - telescoping_threshsig.dat
# - singlepass_merkle.dat
# - singlepass_threshsig.dat

# Each file will contain one line for every network scale seen in the
# logs.  The line will consist of 27 fields:
# - the network scale (float)
# - the mean and stddev (two floats) for each of the above 13 classes,
#   in the above order; for the steady states, the stats are computed
#   over the per-epoch averages, while for the others, the stats are
#   computed on a per-entity basis

import math
import re
import fileinput

modes = [ "vanilla_none", "telescoping_merkle", "telescoping_threshsig",
          "singlepass_merkle", "singlepass_threshsig" ]

class StatAccum:
    """Accumulate (mean, stddev, N) triples to compute an overall (mean,
    stddev, N)."""
    def __init__(self):
        self.sumX = 0
        self.sumXsq = 0
        self.totN = 0

    def accum(self, mean, stddev, N):
        if N == 0:
            # Nothing to change
            return

        if N == 1:
            # stddev will be None
            this_sumX = mean
            this_sumXsq = mean*mean

        else:
            this_sumX = mean * N
            variance = stddev * stddev
            this_sumXsq = variance * (N-1) + this_sumX * this_sumX / N

        self.sumX += this_sumX
        self.sumXsq += this_sumXsq
        self.totN += N

    def stats(self):
        if self.totN == 0:
            return (None, None, 0)

        if self.totN == 1:
            return (self.sumX, None, 1)

        mean = self.sumX / self.totN
        variance = (self.sumXsq - self.sumX * self.sumX / self.totN) \
                    / (self.totN - 1)
        stddev = math.sqrt(variance)
        return (mean, stddev, self.totN)

    def __str__(self):
        mean, stddev, N = self.stats()
        if mean is None:
            mean = 0
        if stddev is None:
            stddev = 0
        return "%f %f" % (mean, stddev)

class StatBundle:
    """A bundle of 13 StatAccums, corresponding to the 13 entity classes
    listed above."""

    def __init__(self):
        self.stats = [StatAccum() for i in range(13)]

    def dirauth(self, mean, stddev, N):
        self.stats[0].accum(mean, stddev, N)

    def relay_all(self, mean, stddev, N):
        self.stats[1].accum(mean, stddev, N)

    def relay_fb_boot(self, mean, stddev, N):
        self.stats[2].accum(mean, stddev, N)

    def relay_fb_nboot(self, mean, stddev, N):
        self.stats[3].accum(mean, stddev, N)

    def relay_nfb_boot(self, mean, stddev, N):
        self.stats[4].accum(mean, stddev, N)

    def relay_nfb_nboot(self, mean, stddev, N):
        self.stats[5].accum(mean, stddev, N)

    def client_all(self, mean, stddev, N):
        self.stats[6].accum(mean, stddev, N)

    def client_boot(self, mean, stddev, N):
        self.stats[7].accum(mean, stddev, N)

    def client_nboot(self, mean, stddev, N):
        self.stats[8].accum(mean, stddev, N)

    def steady_dirauth(self, mean, stddev, N):
        self.stats[9].accum(mean, stddev, N)

    def steady_relay(self, mean, stddev, N):
        self.stats[10].accum(mean, stddev, N)

    def steady_client(self, mean, stddev, N):
        self.stats[11].accum(mean, stddev, N)

    def steady_relay_perclient(self, mean, stddev, N):
        self.stats[12].accum(mean, stddev, N)

    def __str__(self):
        return '%s %s %s %s %s %s %s %s %s %s %s %s %s' % (
            self.stats[0], self.stats[1], self.stats[2], self.stats[3],
            self.stats[4], self.stats[5], self.stats[6], self.stats[7],
            self.stats[8], self.stats[9], self.stats[10],
            self.stats[11], self.stats[12])

class LogParser:
    """A class to parse the logfiles output by simulator.py."""

    def __init__(self):
        # self.stats is a dict indexed by mode name (like
        # "singlepass_merkle") whose value is a dict indexed
        # by network scale whose value is a StatBundle
        self.stats = dict()

        self.curbundle = None
        self.fbbootstrapping = None
        self.steadystate = False

        self.startre = re.compile('Starting simulation .*?\/([A-Z]+)_([A-Z]+)_([\d\.]+)_')
        self.statperbwre = re.compile('(Relays\(N?F?B\)).*bytes=([\d\.]+)( .pm ([\d\.]+))?.*bytesperbw=([\d\.]+)( .pm ([\d\.]+))?.*N=(\d+)')
        self.statre = re.compile('(Dirauths|Relays|Clients(\(N?B\))?).*bytes=([\d\.]+)( .pm ([\d\.]+))?.*N=(\d+)')
        self.mibre = re.compile('MiB used')

    def parse_line(self, line):
        m = self.startre.search(line)
        if m:
            mode = m.group(1).lower() + "_" + m.group(2).lower()
            scale = m.group(3)
            if mode not in self.stats:
                self.stats[mode] = dict()
            if scale not in self.stats[mode]:
                self.stats[mode][scale] = StatBundle()
            self.curbundle = self.stats[mode][scale]
            self.fbbootstrapping = True
            self.steadystate = False
            return

        m = self.statperbwre.search(line)
        if m:
            enttype, means, stddevs, meanperbws, stddevperbws, Ns = \
                    m.group(1,2,4,5,7,8)
        else:
            m = self.statre.search(line)
            if m:
                enttype, means, stddevs, Ns = \
                        m.group(1,3,5,6)
                meanperbws, stddevperbws = None, None
            else:
                m = self.mibre.search(line)
                if m:
                    # We've reached steady state
                    self.steadystate = True
                return

        mean = float(means)
        if stddevs:
            stddev = float(stddevs)
        else:
            stddev = None
        if meanperbws:
            meanperbw = float(meanperbws)
        else:
            meanperbw = None
        if stddevperbws:
            stddevperbw = float(stddevperbws)
        else:
            stddevperbw = None
        N = int(Ns)
        # print('%s %s %s %s %s %s' % (enttype, mean, stddev, meanperbw, stddevperbw, N))
        if enttype == 'Dirauths':
            self.curbundle.dirauth(mean, stddev, N)
            if self.steadystate:
                self.curbundle.steady_dirauth(mean, None, 1)
        elif enttype == 'Relays':
            self.curbundle.relay_all(mean, stddev, N)
            if self.steadystate:
                self.curbundle.steady_relay(mean, None, 1)
                self.totrelaybytes = mean * N
        elif enttype == 'Relays(FB)':
            if self.fbbootstrapping:
                self.curbundle.relay_fb_boot(meanperbw, stddevperbw, N)
                self.fbbootstrapping = False
            else:
                self.curbundle.relay_fb_nboot(meanperbw, stddevperbw, N)
        elif enttype == 'Relays(B)':
            self.curbundle.relay_nfb_boot(meanperbw, stddevperbw, N)
        elif enttype == 'Relays(NB)':
            self.curbundle.relay_nfb_nboot(meanperbw, stddevperbw, N)
        elif enttype == 'Clients':
            self.curbundle.client_all(mean, stddev, N)
            if self.steadystate:
                self.curbundle.steady_client(mean, None, 1)
                self.curbundle.steady_relay_perclient(self.totrelaybytes / N, None, 1)
        elif enttype == 'Clients(B)':
            self.curbundle.client_boot(mean, stddev, N)
        elif enttype == 'Clients(NB)':
            self.curbundle.client_nboot(mean, stddev, N)
        else:
            raise ValueError('Unknown entity type "%s"' % enttype)

    def write_output(self):
        for mode in self.stats.keys():
            with open("%s.dat" % mode, "w") as datout:
                for scale in sorted(self.stats[mode].keys()):
                    datout.write("%s %s\n" % \
                            (6500*float(scale), self.stats[mode][scale]))
                datout.close()


if __name__ == '__main__':

    logparser = LogParser()

    for line in fileinput.input():
        logparser.parse_line(line)

    logparser.write_output()