find_heaviest_weak_symbols.py: Find the most included symbols in .o files.

find_heaviest_weak_symbols.py

#!/usr/bin/python
#
# Is your C++ project made of bloat and fail? Wouldn't it be great if you got a
# nice table of the individual symbols sorted by total size (size of one
# instance * number of instances)?
#
# Use as such:
#
#   $ find <builddir> -name "*.o" | xargs nm -S | c++filt | find_heaviest_weak_symbols.py

import sys

symbols = { }

for line in sys.stdin.readlines():
  if line.find(" W ") != -1:
    size = line[17:33]
    name = line[36:].rstrip()

    if name in symbols:
      entry = symbols[name]
    else:
      entry = {'size':0, 'count':0}
      symbols[name] = entry

    entry['size'] += int(size, 16)
    entry['count'] += 1

# Translate the symbols dictionary into an array of sortable tuples
# (total_size, count, name) for outputs.
def converter(x): return (x[1]['size'], x[1]['count'], x[0])
for_output = map(converter, symbols.items())
for_output.sort()

print "  Size Count Name"
print "------ ----- ---------------------------------------------------------------"
for row in for_output:
  # Don't print instances where there's only one copy of the symbol in the
  # entire program, no matter how big.
  if row[1] > 1:
    print "{0[0]:6} {0[1]:5} {0[2]}".format(row)