# una2csv.py extracts a comma-separated-valye (csv) record
# from a *una.txt file in stdin as generated by waveformAnalyzer.ck
# (from a .wav file) for csc558 assignment 1, spring 2018, D. Parson
# reads stdin, writes stdout`
# Updated 1/28/2018 to put the random serial number out as 1st attribute.

import sys
import re
__hdr_re__ = re.compile(r'^lazy1_([A-Za-z]+)_([0-9]+)_([01]\.[0-9]+)_([01]\.[0-9]+)_([0-9]+).wav.una.txt ([0-9]\.[0-9]+) ([0-9]\.[0-9]+) ([0-9]\.[0-9]+) ([0-9]\.[0-9]+) ([0-9]\.[0-9]+)$')
__bin_re__ = re.compile(r'^([0-9]\.[0-9]+) :\(float\)')
__fftbins__ = 512
_reducedbins__ = 20
# (osc) (freq) (oscgain) (noisegain) (rnd) (centroid) (rms) (25th) (50th) (75th)

if __name__ == '__main__':
    hdr = ''
    m = None
    line = sys.stdin.readline()
    lineno = 1
    while line:
        line = line.strip()
        m = __hdr_re__.match(line)
        if m:
            hdr = line
            break
        line = sys.stdin.readline()
        lineno += 1
    if not m:
        sys.stderr.write("ERROR: No header line in input file.\n")
        sys.exit(1)
    osc = m.group(1).strip()
    freq = float(m.group(2).strip())
    oscgain = float(m.group(3).strip())
    noisegain = float(m.group(4).strip())
    rnd = int(m.group(5).strip())
    centroid = float(m.group(6).strip())
    rms = float(m.group(7).strip())
    roll25 = float(m.group(8).strip())
    roll50 = float(m.group(9).strip())
    roll75 = float(m.group(10).strip())
    line = sys.stdin.readline()
    lineno += 1
    histogram = []
    sys.stdout.write(str(rnd) + ",'" + osc + "'," + str(freq) + ","
        + str(oscgain) + "," + str(noisegain) + "," + str(centroid)
        + "," + str(rms)
        + "," + str(roll25) + "," + str(roll50) + "," + str(roll75))
    while line:
        line = line.strip()
        m = __bin_re__.match(line)
        if not m:
            sys.stderr.write("ERROR: Line " + str(lineno)
                + " is not a valid line: " + line + '\n')
            sys.stderr.write("ERROR file: " + hdr + '\n')
            sys.exit(1)
        histogram.append(float(m.group(1).strip()))
        line = sys.stdin.readline()
        lineno += 1
    if len(histogram) != __fftbins__:
            sys.stderr.write("ERROR: Line count " + str(lineno-1)
                + " does not have " + str(__fftbins__) + " bins: "
                + str(len(histogram)) + '\n')
            sys.stderr.write("ERROR file: " + hdr + '\n')
            sys.exit(1)
    # Skip first two bins as out-of-band data, lower than the frequency range
    # [100,2000] used by the generator, then try to find the fundamental as
    # the highest level.
    hibin = 2
    hival = histogram[2]
    for i in range(2,len(histogram)):
        if histogram[i] > hival:
            hival = histogram[i]
            hibin = i
    # Store the fundamental freq amplitude, then look at its harmonics.
    def __findnearest__(histogram, lookat):     # Look for the nearest peak.
        ix = lookat - 3
        if ix >= len(histogram):
            return(-1, 0.0)
        bigix = ix
        bigval = histogram[ix]
        while ix < len(histogram) and ix <= (lookat + 3):
            if histogram[ix] >= bigval:
                bigval = histogram[ix]
                bigix = ix
            ix += 1
        return(bigix, bigval)
    scalebins = [hival]
    debugbins = [hibin]
    # Number of bins and actual bin offset of fundamental for freq extraction.
    sys.stdout.write(",44100," + str(__fftbins__) + ","
        + str(_reducedbins__) + "," + str(hibin))
    jumpbin = hibin if (hibin >= 4) else 4
    # Needed to add jumpbin for debugging because if fundamental
    # frequency is too close to bottom, __findnearest__'s +/- 3
    # gets stuck at a single location. Seems to be a problem only
    # for 100 hz sine wave.
    nearbin, nextval = __findnearest__(histogram, 2*jumpbin)
    while nearbin > 0 and nearbin < len(histogram):
        scalebins.append(nextval)
        debugbins.append(nearbin)
        nearbin, nextval = __findnearest__(histogram, nearbin+jumpbin)
        # sys.stderr.write("DEBUG NEARBIN: " + str(nearbin) + "," + str(jumpbin) + '\n')
    biggest = 0.0
    for i in range(0, len(scalebins)):
        # scalebins[i] = 10.0 ** scalebins[i] NO! # convert from log to linear
        if scalebins[i] > biggest:
            biggest = scalebins[i]
    for i in range(0, len(scalebins)):          # normalize to 1.0
        scalebins[i] = scalebins[i] / biggest
    if len(scalebins) > _reducedbins__:
        scalebins = scalebins[:_reducedbins__]
        # normalize to _reducedbins__ bins
    else:
        while len(scalebins) < _reducedbins__:
            scalebins.append(0.0)
    sys.stdout.write("," + ("%1.12f" % (1.0 / biggest))) # amplitude scale
    for sc in scalebins:
        # sys.stdout.write("," + str(sc))
        sys.stdout.write("," + ("%1.12f" % sc))
    sys.stdout.write("\n")
    sys.stderr.write("DEBUG BINS: " + str(debugbins) + " IN " + hdr + "\n")