1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
|
require 'rake/rdoctask'
require 'rake/testtask'
require 'rake/packagetask'
require 'rake/gempackagetask'
require 'rbconfig'
require 'fileutils'
$:.unshift 'lib'
require 'ole/storage'
PKG_NAME = 'ruby-ole'
PKG_VERSION = Ole::Storage::VERSION
task :default => [:test]
Rake::TestTask.new do |t|
t.test_files = FileList["test/test_*.rb"]
t.warning = true
t.verbose = true
end
begin
require 'rcov/rcovtask'
# NOTE: this will not do anything until you add some tests
desc "Create a cross-referenced code coverage report"
Rcov::RcovTask.new do |t|
t.test_files = FileList['test/test*.rb']
t.ruby_opts << "-Ilib" # in order to use this rcov
t.rcov_opts << "--xrefs" # comment to disable cross-references
t.verbose = true
end
rescue LoadError
# Rcov not available
end
Rake::RDocTask.new do |t|
t.rdoc_dir = 'doc'
t.rdoc_files.include 'lib/**/*.rb'
t.rdoc_files.include 'README', 'ChangeLog'
t.title = "#{PKG_NAME} documentation"
t.options += %w[--line-numbers --inline-source --tab-width 2]
t.main = 'README'
end
spec = Gem::Specification.new do |s|
s.name = PKG_NAME
s.version = PKG_VERSION
s.summary = %q{Ruby OLE library.}
s.description = %q{A library for easy read/write access to OLE compound documents for Ruby.}
s.authors = ['Charles Lowe']
s.email = %q{aquasync@gmail.com}
s.homepage = %q{http://code.google.com/p/ruby-ole}
s.rubyforge_project = %q{ruby-ole}
s.executables = ['oletool']
s.files = ['README', 'Rakefile', 'ChangeLog', 'data/propids.yaml']
s.files += FileList['lib/**/*.rb']
s.files += FileList['test/test_*.rb', 'test/*.doc']
s.files += FileList['test/oleWithDirs.ole', 'test/test_SummaryInformation']
s.files += FileList['bin/*']
s.test_files = FileList['test/test_*.rb']
s.has_rdoc = true
s.extra_rdoc_files = ['README', 'ChangeLog']
s.rdoc_options += [
'--main', 'README',
'--title', "#{PKG_NAME} documentation",
'--tab-width', '2'
]
end
Rake::GemPackageTask.new(spec) do |t|
t.gem_spec = spec
t.need_tar = true
t.need_zip = false
t.package_dir = 'build'
end
desc 'Run various benchmarks'
task :benchmark do
require 'benchmark'
require 'tempfile'
require 'ole/file_system'
# should probably add some read benchmarks too
def write_benchmark opts={}
files, size = opts[:files], opts[:size]
block_size = opts[:block_size] || 100_000
block = 0.chr * block_size
blocks, remaining = size.divmod block_size
remaining = 0.chr * remaining
Tempfile.open 'ole_storage_benchmark' do |temp|
Ole::Storage.open temp do |ole|
files.times do |i|
ole.file.open "file_#{i}", 'w' do |f|
blocks.times { f.write block }
f.write remaining
end
end
end
end
end
Benchmark.bm do |bm|
bm.report 'write_1mb_1x5' do
5.times { write_benchmark :files => 1, :size => 1_000_000 }
end
bm.report 'write_1mb_2x5' do
5.times { write_benchmark :files => 1_000, :size => 1_000 }
end
end
end
=begin
1.2.1:
user system total real
write_1mb_1x5 73.920000 8.400000 82.320000 ( 91.893138)
revision 17 (speed up AllocationTable#free_block by using
@sparse attribute, and using Array#index otherwise):
user system total real
write_1mb_1x5 57.910000 6.190000 64.100000 ( 66.207993)
write_1mb_2x5266.310000 31.750000 298.060000 (305.877203)
add in extra resize_chain fix (return blocks to avoid calling
AllocationTable#chain twice):
user system total real
write_1mb_1x5 43.140000 5.480000 48.620000 ( 51.835942)
add in RangesIOResizeable fix (cache @blocks, to avoid calling
AllocationTable#chain at all when resizing now, just pass it
to AllocationTable#resize_chain):
user system total real
write_1mb_1x5 29.770000 5.180000 34.950000 ( 39.916747)
40 seconds is still a really long time to write out 5 megs.
of course, this is all with a 1_000 byte block size, which is
a very small wite. upping this to 100_000 bytes:
user system total real
write_1mb_1x5 0.540000 0.130000 0.670000 ( 1.051862)
so it seems that that makes a massive difference. so i really
need buffering in RangesIO if I don't want it to really hurt
for small writes, as all the resize code is kind of expensive.
one of the costly things at the moment, is RangesIO#offset_and_size,
which is called for each write, and re-finds which range we are in.
that should obviously be changed, to a fixed one that is invalidated
on seeks. buffering would hide that problem to some extent, but i
should fix it anyway.
re-running the original 1.2.1 with 100_000 byte block size:
user system total real
write_1mb_1x5 15.590000 2.230000 17.820000 ( 18.704910)
so there the really badly non-linear AllocationTable#resize_chain is
being felt.
back to current working copy, running full benchmark:
user system total real
write_1mb_1x5 0.530000 0.150000 0.680000 ( 0.708919)
write_1mb_2x5227.940000 31.260000 259.200000 (270.200960)
not surprisingly, the second case hasn't been helped much by the fixes
so far, as they only really help multiple resizes and writes for a file.
this could be pain in the new file system code - potentially searching
through Dirent#children at creation time.
to test, i'll profile creating 1_000 files, without writing anything:
user system total real
write_1mb_2x5 16.990000 1.830000 18.820000 ( 19.900568)
hmmm, so thats not all of it. maybe its the initial chain calls, etc?
writing 1 byte:
user system total real
write_1mb_1x5 0.520000 0.120000 0.640000 ( 0.660638)
write_1mb_2x5 19.810000 2.280000 22.090000 ( 22.696214)
weird.
100 bytes:
user system total real
write_1mb_1x5 0.560000 0.140000 0.700000 ( 1.424974)
write_1mb_2x5 22.940000 2.840000 25.780000 ( 26.556346)
500 bytes:
user system total real
write_1mb_1x5 0.530000 0.150000 0.680000 ( 1.139738)
write_1mb_2x5 77.260000 10.130000 87.390000 ( 91.671086)
what happens there? very strange.
=end
|
