#!/bin/bash
# Generate code count plots by iterating over tagged revisions and running cloc.
# Requires cloc and gnuplot.

# Output images.
loc="`hg root`/docs/images/loc.png"
languages="`hg root`/docs/images/languages.png"

# Clone a temporary repository and generate code count data for gnuplot to plot.
tmp=/tmp/count
hg clone `hg root` $tmp && pushd $tmp
plotfile=gnuplot.dat
for rev in `hg tags | awk '{print $2}' | cut -d: -f1 | tac`; do
  date=`hg log -r $rev | grep ^date | cut -d: -f2- | tr + - | cut -d- -f1`
  hg update -r $rev -q
  timestamp=`date -d "$date" "+%s"`
  counts=`cloc --force-lang=C,h --include-lang=C,Lua,make --quiet --csv \
    --exclude-dir=doc,docs,scripts,themes,test,.github --not-match-f=adeptsensedoc\\|tadoc . | \
    tail -n +3 | head -n -1 | cut -d, -f2- | sort | tr '\n' ,`
  echo $timestamp,$counts
done | lua -e "
  -- Filter counts from cloc into a data format readable by gnuplot.
  -- Input is of the form:
  --  timestamp1,lang1,blanks1,comments1,code2,lang2,blanks2,comments2,code2,...
  --  timestamp2,lang1,blanks1,comments1,code2,lang2,blanks2,comments2,code2,...
  -- Output is of the form:
  --  Lang1
  --  Time Blanks Comments Code
  --  ts1  bl1    cm1      co1
  --  ts2  bl2    cm2      co2
  --  ...
  --
  --
  --  Lang2
  --  Time Blanks Comments Code
  --  ts1  bl1    cm1      co1
  --  ts2  bl2    cm2      co2
  --  ...

  -- Read in counts.
  local counts, langs = {}, {}
  for line in io.lines() do
    local time, data = line:match('^(%d+),(.+)$')
    counts[#counts + 1], counts[time] = time, {}
    for lang, data in data:gmatch('([^,]+),([^,]+,[^,]+,[^,]+)') do
      lang = lang:match('%S+$')
      if not langs[lang] then langs[#langs + 1], langs[lang] = lang, true end
      counts[time][lang] = data
    end
  end
  table.sort(langs)
  -- Output a data series for each language counted.
  for i = 1, #langs do
    print(langs[i])
    print('Time', 'Blanks', 'Comments', 'Code')
    for j = 1, #counts do
      local data = (counts[counts[j]][langs[i]] or '0,0,0'):gsub(',', '\t')
      print(counts[j], data)
    end
    print('\n') -- double-newline needed to delimit gnuplot data series
  end
  -- Output a data series for code, comments, and blanks counted.
  for i = 1, #counts do
    local sum_code, sum_comments, sum_blanks = 0, 0, 0
    for j = 1, #langs do
      local data = counts[counts[i]][langs[j]] or '0,0,0'
      local blanks, comments, code = data:match('([^,]+),([^,]+),([^,]+)')
      sum_code = sum_code + tonumber(code)
      sum_comments = sum_comments + tonumber(comments)
      sum_blanks = sum_blanks + tonumber(blanks)
    end
    counts[counts[i]] = {sum_code, sum_comments, sum_blanks}
  end
  for i, measure in ipairs{'Code', 'Comments', 'Blanks'} do
    print(measure)
    print('Time', measure)
    for j = 1, #counts do print(counts[j], counts[counts[j]][i]) end
    print('\n') -- double-newline needed to delimit gnuplot data series
  end
" > $plotfile
nlangs=$((`grep ^Time $plotfile | wc -l` - 3)) # ignore blanks, comments, code

# Define gnuplot plot settings and plot commands.
plotcmd=gnuplot.plt
echo "
  set term png transparent font 'DejaVu Sans,10' size 450,275;
  set border linewidth 1.5;
  set style line 1 linecolor rgb '#99994d' linewidth 2;
  set style line 2 linecolor rgb '#4d99e6' linewidth 2;
  set style line 3 linecolor rgb '#994d4d' linewidth 2;
  set style line 4 linecolor rgb '#4d994d' linewidth 2;
  set style line 5 linecolor rgb '#994d99' linewidth 2;
  set grid linecolor rgb '#cccccc' linetype 1;
  set tics scale 0;
  set key left top horizontal Left reverse samplen 1;
  set xdata time; set timefmt '%s'; set format x \"%Y\n%b\";
  set autoscale xfix;

  set title 'Code Summary'
  set output '$loc'
  plot for [i=0:2] '$plotfile' index $nlangs+i using 1:2 with lines title columnhead(1) linestyle i+1
  set title 'Code Line Counts'
  set output '$languages'
  plot for [i=0:$nlangs-1] '$plotfile' index i using 1:4 with lines title columnhead(1) linestyle i+1
" > $plotcmd

# Invoke gnuplot.
gnuplot $plotcmd

# Cleanup.
popd && rm -r $tmp