eigenclass - More about naming conventions

MAIN Index Search Changes PageRank Login

I patched ruby 1.8.4 to gather some statistics regarding naming conventions, as far as identifiers are concerned. My local build will now emit such information at parse time as follows:

 $ ruby -t -c -e "def a(x); c = x.size; def x.bar; 1 end; c + 1 end"
 [tokinfo] method def: a
 [tokinfo] tIDENTIFIER assignment for lvar: c
 [tokinfo] singleton method def: bar
 Syntax OK

These are the statistics for Ruby 1.8.4's stdlib:

type	average length	std dev	count
symbol	7.0	4.3	3214
ivar	8.6	3.9	1915
smethod	9.6	4.9	1001
gvar	9.1	4.3	377
cname	8.3	4.8	2124
cvar	11.4	4.4	103
dvar_curr	3.5	2.7	1158
constant	10.9	5.0	1660
dvar	4.7	3.7	49
method	10.5	5.6	7566
lvar	4.3	2.9	3427

I just added some code to the parser and some support functions in order to capture:

uses of symbols (symbol)
method definitions (method)
singleton method definitions in the form def object.meth (smethod)
class/module names (cname)
plain constant names (constant)
block-local variables assigned to inside a nested block (dvar)
block-local variables (dvar_curr)
local variables (lvar)
instance variables (ivar)
global variables (gvar)
class variables (cvar)

The difference between dvar and dvar_curr is fairly subtle; in the following example, x is a dvar_curr:

%w[a b c].each{|x| puts x * 2}

"curr" means that the variable was defined in the current block.

When a block reuses a variable defined in an enclosing block, it is considered a dvar:

%w[aasdad dfdsb sdfsfc].map do |x|
  max = 0
  x.each_byte{|b| max = b if b > max }
  max
end

Here max, as seen from the block passed to each_byte, is a dvar.

Comparison across Ruby versions

The average identifier length seems to have increased a bit or remained mostly stable for most categories. This increase can in part be attributed to the growth of the standard library: more specific libraries often require longer names (for more complex concepts).

Average constant name length

Average method name length

Some code

As usual, here's the code I used to generate the stats. I guess the most interesting bits would be

  allstats.each do |key, val|
    tlength, tlength2, ntoks = val.inject([0,0,0]) do |(s,s2,t),(name,count)|
      [s + name.size * count, s2 + (name.size ** 2 * count), t + count]
    end
    avg = 1.0 * tlength / ntoks
    stddev = Math.sqrt(1.0 * tlength2 / ntoks - avg ** 2)
    ret[key] = [avg, stddev, val.size]
  end

  def_proc = lambda{|h,k| h[k] = 0}
  stats = {}
   
  MATCHER.each do |key, val|
    stats[key] = Hash.new(&def_proc)
  end

Full code

MATCHER = {
  :symbol => ["symbol: (\\S+)"],
  :method => ["method def: (\\S+)"],
  :smethod => ["singleton method def: (\\S+)"],
  :cname => ["tCONSTANT for cname: (\\S+)"],
  :dvar => ["tIDENTIFIER assignment for dvar: (\\S+)"],
  :dvar_curr => ["tIDENTIFIER assignment for dvar_curr: (\\S+)"],
  :lvar => ["tIDENTIFIER assignment for lvar: (\\S+)"],
  :ivar => ["tIDENTIFIER assignment for ivar: (\\S+)"], 
  :gvar => ["tIDENTIFIER assignment for gvar: (\\S+)"],
  :cvar => ["tIDENTIFIER assignment for cvar: (\\S+)"],
  :constant => ["tIDENTIFIER assignment for constant: (\\S+)"] 
}
def naming_statistics(filename)
  def_proc = lambda{|h,k| h[k] = 0}
  stats = {}
   
  MATCHER.each do |key, val|
    stats[key] = Hash.new(&def_proc)
  end
  IO.popen("ruby -t -c #{filename} 2>&1") do |f|
    f.each do |line|
      next unless /^\[tokinfo\] / =~ line
      MATCHER.each do |key, re_arr|
        md = nil
        if re_arr.any?{|x| md = Regexp.new("^\\[tokinfo\\] #{x}").match(line) }
          stats[key][md[1]] += 1
          break
        end
      end
    end
  end
  stats
end

require 'find'
def dir_stats(dirname)
  allstats = Hash.new{|h,k| h[k] = Hash.new{|h2,k2| h2[k2] = 0} }
  Dir.chdir(dirname) do
    Find.find(".") do |fname|
      next unless /\.rb$/ =~ fname
      stats = naming_statistics(fname)
      stats.each_pair{|k,v| v.each_pair{|name,count| allstats[k][name] += count} }
    end
  end
  ret = {}
  allstats.each do |key, val|
    tlength, tlength2, ntoks = val.inject([0,0,0]) do |(s,s2,t),(name,count)|
      [s + name.size * count, s2 + (name.size ** 2 * count), t + count]
    end
    avg = 1.0 * tlength / ntoks
    stddev = Math.sqrt(1.0 * tlength2 / ntoks - avg ** 2)
    ret[key] = [avg, stddev, val.size]
  end

  ret
end

%w[/tmp/ruby-1.4.6/ /tmp/ruby-1.6.8 /tmp/ruby-1.8.0 /tmp/ruby-1.8.4].each do |dirname|
  stats = dir_stats(dirname)
  puts "=" * 20
  puts dirname
  puts "=" * 20
  stats.each_pair do |k, r|
    puts "%10s  %4.1f  %4.1f  (%5d)" % [k, *r]
  end
end

Last modified:2005/12/25 07:01:40
Keyword(s):[blog] [ruby] [naming] [convention] [1.8.4]
References:[Ruby]

More about naming conventions

Comparison across Ruby versions

Average constant name length

Average method name length

Some code

Full code

Recent

Popular