#!/bin/bash -e
 
CSV_FILE=loaded.csv
CSV_COLUMN=2
 
SAMPLE_RATE=5000000
 
ROOT_FILENAME=${CSV_FILE%.*}
 
# end of configuration
 
cat ${CSV_FILE} \
  | tail -n+3 \
  | cut -d , -f ${CSV_COLUMN} \
  | sed -r \
    -e 's/^\+//' \
    -e 's/^(-?[0-9]+\.[0-9]+)[eE]([+-])([0-9]+)$/\1*(10^\2\3)/' \
    -e 's/\*\(10\^\+00\)$//' \
  | bc -l \
  | sed -re 's/^(-?)\./\10./' \
  | tee .${CSV_FILE}.normal \
  | sort -n \
  | uniq \
  > .${CSV_FILE}.sorted
 
cat .${CSV_FILE}.sorted | sed -n "1p;$(wc -l .${CSV_FILE}.sorted | cut -d ' ' -f 1)p" > .${CSV_FILE}.extremes
 
LO=$(cat .${CSV_FILE}.extremes | head -n 1)
if [ $(echo "${LO} < 0" | bc) == "1" ]; then
  OP=-
else
  OP=+
fi
HI=$(echo "($(cat .${CSV_FILE}.extremes | tail -n 1))${OP}(${LO})" | bc -l)
ML=$(echo "255/${HI}" | bc -l)
 
cat .${CSV_FILE}.normal \
  | sed -r \
    -e 's/^/((/' \
    -e 's/$/)/' \
    -e "s/$/${OP}(${LO})/" \
    -e "s/$/)*${ML}/" \
  | bc -l \
  | sed -r \
    -e 's/^([0-9]*).([0-9]?[0-9]?[0-9]?).*$/\1.\2/' \
  | tee .${CSV_FILE}.adjusted \
  | awk '{printf "%c", strtonum($0)}' \
  > ${ROOT_FILENAME}.raw
 
DATA_SIZE=$(stat -c%s ${ROOT_FILENAME}.raw)
CHUNK_SIZE=$((DATA_SIZE + 36))
 
WAV=${ROOT_FILENAME}.wav
 
# writeNumAsBin
# arg1: bytes used to represent the word
# arg2: value stored in the word
function wnab() {
  WORD_SIZE=$1; shift
  WORD_VALUE=$1; shift
  printf "%0$((WORD_SIZE * 2))x" ${WORD_VALUE} \
    | fold -w2 \
    | sed '1!G;h;$!d' \
    | awk '{printf "%c", strtonum("0x" $0)}'
}
 
# ChunkID
echo -n "RIFF" > ${WAV}
# ChunkSize
wnab 4 ${CHUNK_SIZE} >> ${WAV}
# Format & Subchunk1ID
echo -n "WAVEfmt " >> ${WAV}
# Subchunk1Size
wnab 4 16 >> ${WAV}
# AudioFormat
wnab 2 1 >> ${WAV}
# NumChannels
wnab 2 1 >> ${WAV}
# SampleRate
wnab 4 ${SAMPLE_RATE} >> ${WAV}
# ByteRate
wnab 4 ${SAMPLE_RATE} >> ${WAV}
# BlockAlign
wnab 2 1 >> ${WAV}
# BitsPerSample
wnab 2 8 >> ${WAV}
# Subchunk2ID
echo -n "data" >> ${WAV}
# Subchunk2Size
wnab 4 ${DATA_SIZE} >> ${WAV}
# data!
cat ${ROOT_FILENAME}.raw >> ${WAV}