├── .gitignore
├── LICENSE
├── README.md
└── fast_sample
/.gitignore:
--------------------------------------------------------------------------------
1 | /blib/
2 | /.build/
3 | _build/
4 | cover_db/
5 | inc/
6 | Build
7 | !Build/
8 | Build.bat
9 | .last_cover_stats
10 | /Makefile
11 | /Makefile.old
12 | /MANIFEST.bak
13 | /META.yml
14 | /META.json
15 | /MYMETA.*
16 | nytprof.out
17 | /pm_to_blib
18 | *.o
19 | *.bs
20 | *.dbf
21 |
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | GNU GENERAL PUBLIC LICENSE
2 | Version 2, June 1991
3 |
4 | Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
5 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
6 | Everyone is permitted to copy and distribute verbatim copies
7 | of this license document, but changing it is not allowed.
8 |
9 | Preamble
10 |
11 | The licenses for most software are designed to take away your
12 | freedom to share and change it. By contrast, the GNU General Public
13 | License is intended to guarantee your freedom to share and change free
14 | software--to make sure the software is free for all its users. This
15 | General Public License applies to most of the Free Software
16 | Foundation's software and to any other program whose authors commit to
17 | using it. (Some other Free Software Foundation software is covered by
18 | the GNU Lesser General Public License instead.) You can apply it to
19 | your programs, too.
20 |
21 | When we speak of free software, we are referring to freedom, not
22 | price. Our General Public Licenses are designed to make sure that you
23 | have the freedom to distribute copies of free software (and charge for
24 | this service if you wish), that you receive source code or can get it
25 | if you want it, that you can change the software or use pieces of it
26 | in new free programs; and that you know you can do these things.
27 |
28 | To protect your rights, we need to make restrictions that forbid
29 | anyone to deny you these rights or to ask you to surrender the rights.
30 | These restrictions translate to certain responsibilities for you if you
31 | distribute copies of the software, or if you modify it.
32 |
33 | For example, if you distribute copies of such a program, whether
34 | gratis or for a fee, you must give the recipients all the rights that
35 | you have. You must make sure that they, too, receive or can get the
36 | source code. And you must show them these terms so they know their
37 | rights.
38 |
39 | We protect your rights with two steps: (1) copyright the software, and
40 | (2) offer you this license which gives you legal permission to copy,
41 | distribute and/or modify the software.
42 |
43 | Also, for each author's protection and ours, we want to make certain
44 | that everyone understands that there is no warranty for this free
45 | software. If the software is modified by someone else and passed on, we
46 | want its recipients to know that what they have is not the original, so
47 | that any problems introduced by others will not reflect on the original
48 | authors' reputations.
49 |
50 | Finally, any free program is threatened constantly by software
51 | patents. We wish to avoid the danger that redistributors of a free
52 | program will individually obtain patent licenses, in effect making the
53 | program proprietary. To prevent this, we have made it clear that any
54 | patent must be licensed for everyone's free use or not licensed at all.
55 |
56 | The precise terms and conditions for copying, distribution and
57 | modification follow.
58 |
59 | GNU GENERAL PUBLIC LICENSE
60 | TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
61 |
62 | 0. This License applies to any program or other work which contains
63 | a notice placed by the copyright holder saying it may be distributed
64 | under the terms of this General Public License. The "Program", below,
65 | refers to any such program or work, and a "work based on the Program"
66 | means either the Program or any derivative work under copyright law:
67 | that is to say, a work containing the Program or a portion of it,
68 | either verbatim or with modifications and/or translated into another
69 | language. (Hereinafter, translation is included without limitation in
70 | the term "modification".) Each licensee is addressed as "you".
71 |
72 | Activities other than copying, distribution and modification are not
73 | covered by this License; they are outside its scope. The act of
74 | running the Program is not restricted, and the output from the Program
75 | is covered only if its contents constitute a work based on the
76 | Program (independent of having been made by running the Program).
77 | Whether that is true depends on what the Program does.
78 |
79 | 1. You may copy and distribute verbatim copies of the Program's
80 | source code as you receive it, in any medium, provided that you
81 | conspicuously and appropriately publish on each copy an appropriate
82 | copyright notice and disclaimer of warranty; keep intact all the
83 | notices that refer to this License and to the absence of any warranty;
84 | and give any other recipients of the Program a copy of this License
85 | along with the Program.
86 |
87 | You may charge a fee for the physical act of transferring a copy, and
88 | you may at your option offer warranty protection in exchange for a fee.
89 |
90 | 2. You may modify your copy or copies of the Program or any portion
91 | of it, thus forming a work based on the Program, and copy and
92 | distribute such modifications or work under the terms of Section 1
93 | above, provided that you also meet all of these conditions:
94 |
95 | a) You must cause the modified files to carry prominent notices
96 | stating that you changed the files and the date of any change.
97 |
98 | b) You must cause any work that you distribute or publish, that in
99 | whole or in part contains or is derived from the Program or any
100 | part thereof, to be licensed as a whole at no charge to all third
101 | parties under the terms of this License.
102 |
103 | c) If the modified program normally reads commands interactively
104 | when run, you must cause it, when started running for such
105 | interactive use in the most ordinary way, to print or display an
106 | announcement including an appropriate copyright notice and a
107 | notice that there is no warranty (or else, saying that you provide
108 | a warranty) and that users may redistribute the program under
109 | these conditions, and telling the user how to view a copy of this
110 | License. (Exception: if the Program itself is interactive but
111 | does not normally print such an announcement, your work based on
112 | the Program is not required to print an announcement.)
113 |
114 | These requirements apply to the modified work as a whole. If
115 | identifiable sections of that work are not derived from the Program,
116 | and can be reasonably considered independent and separate works in
117 | themselves, then this License, and its terms, do not apply to those
118 | sections when you distribute them as separate works. But when you
119 | distribute the same sections as part of a whole which is a work based
120 | on the Program, the distribution of the whole must be on the terms of
121 | this License, whose permissions for other licensees extend to the
122 | entire whole, and thus to each and every part regardless of who wrote it.
123 |
124 | Thus, it is not the intent of this section to claim rights or contest
125 | your rights to work written entirely by you; rather, the intent is to
126 | exercise the right to control the distribution of derivative or
127 | collective works based on the Program.
128 |
129 | In addition, mere aggregation of another work not based on the Program
130 | with the Program (or with a work based on the Program) on a volume of
131 | a storage or distribution medium does not bring the other work under
132 | the scope of this License.
133 |
134 | 3. You may copy and distribute the Program (or a work based on it,
135 | under Section 2) in object code or executable form under the terms of
136 | Sections 1 and 2 above provided that you also do one of the following:
137 |
138 | a) Accompany it with the complete corresponding machine-readable
139 | source code, which must be distributed under the terms of Sections
140 | 1 and 2 above on a medium customarily used for software interchange; or,
141 |
142 | b) Accompany it with a written offer, valid for at least three
143 | years, to give any third party, for a charge no more than your
144 | cost of physically performing source distribution, a complete
145 | machine-readable copy of the corresponding source code, to be
146 | distributed under the terms of Sections 1 and 2 above on a medium
147 | customarily used for software interchange; or,
148 |
149 | c) Accompany it with the information you received as to the offer
150 | to distribute corresponding source code. (This alternative is
151 | allowed only for noncommercial distribution and only if you
152 | received the program in object code or executable form with such
153 | an offer, in accord with Subsection b above.)
154 |
155 | The source code for a work means the preferred form of the work for
156 | making modifications to it. For an executable work, complete source
157 | code means all the source code for all modules it contains, plus any
158 | associated interface definition files, plus the scripts used to
159 | control compilation and installation of the executable. However, as a
160 | special exception, the source code distributed need not include
161 | anything that is normally distributed (in either source or binary
162 | form) with the major components (compiler, kernel, and so on) of the
163 | operating system on which the executable runs, unless that component
164 | itself accompanies the executable.
165 |
166 | If distribution of executable or object code is made by offering
167 | access to copy from a designated place, then offering equivalent
168 | access to copy the source code from the same place counts as
169 | distribution of the source code, even though third parties are not
170 | compelled to copy the source along with the object code.
171 |
172 | 4. You may not copy, modify, sublicense, or distribute the Program
173 | except as expressly provided under this License. Any attempt
174 | otherwise to copy, modify, sublicense or distribute the Program is
175 | void, and will automatically terminate your rights under this License.
176 | However, parties who have received copies, or rights, from you under
177 | this License will not have their licenses terminated so long as such
178 | parties remain in full compliance.
179 |
180 | 5. You are not required to accept this License, since you have not
181 | signed it. However, nothing else grants you permission to modify or
182 | distribute the Program or its derivative works. These actions are
183 | prohibited by law if you do not accept this License. Therefore, by
184 | modifying or distributing the Program (or any work based on the
185 | Program), you indicate your acceptance of this License to do so, and
186 | all its terms and conditions for copying, distributing or modifying
187 | the Program or works based on it.
188 |
189 | 6. Each time you redistribute the Program (or any work based on the
190 | Program), the recipient automatically receives a license from the
191 | original licensor to copy, distribute or modify the Program subject to
192 | these terms and conditions. You may not impose any further
193 | restrictions on the recipients' exercise of the rights granted herein.
194 | You are not responsible for enforcing compliance by third parties to
195 | this License.
196 |
197 | 7. If, as a consequence of a court judgment or allegation of patent
198 | infringement or for any other reason (not limited to patent issues),
199 | conditions are imposed on you (whether by court order, agreement or
200 | otherwise) that contradict the conditions of this License, they do not
201 | excuse you from the conditions of this License. If you cannot
202 | distribute so as to satisfy simultaneously your obligations under this
203 | License and any other pertinent obligations, then as a consequence you
204 | may not distribute the Program at all. For example, if a patent
205 | license would not permit royalty-free redistribution of the Program by
206 | all those who receive copies directly or indirectly through you, then
207 | the only way you could satisfy both it and this License would be to
208 | refrain entirely from distribution of the Program.
209 |
210 | If any portion of this section is held invalid or unenforceable under
211 | any particular circumstance, the balance of the section is intended to
212 | apply and the section as a whole is intended to apply in other
213 | circumstances.
214 |
215 | It is not the purpose of this section to induce you to infringe any
216 | patents or other property right claims or to contest validity of any
217 | such claims; this section has the sole purpose of protecting the
218 | integrity of the free software distribution system, which is
219 | implemented by public license practices. Many people have made
220 | generous contributions to the wide range of software distributed
221 | through that system in reliance on consistent application of that
222 | system; it is up to the author/donor to decide if he or she is willing
223 | to distribute software through any other system and a licensee cannot
224 | impose that choice.
225 |
226 | This section is intended to make thoroughly clear what is believed to
227 | be a consequence of the rest of this License.
228 |
229 | 8. If the distribution and/or use of the Program is restricted in
230 | certain countries either by patents or by copyrighted interfaces, the
231 | original copyright holder who places the Program under this License
232 | may add an explicit geographical distribution limitation excluding
233 | those countries, so that distribution is permitted only in or among
234 | countries not thus excluded. In such case, this License incorporates
235 | the limitation as if written in the body of this License.
236 |
237 | 9. The Free Software Foundation may publish revised and/or new versions
238 | of the General Public License from time to time. Such new versions will
239 | be similar in spirit to the present version, but may differ in detail to
240 | address new problems or concerns.
241 |
242 | Each version is given a distinguishing version number. If the Program
243 | specifies a version number of this License which applies to it and "any
244 | later version", you have the option of following the terms and conditions
245 | either of that version or of any later version published by the Free
246 | Software Foundation. If the Program does not specify a version number of
247 | this License, you may choose any version ever published by the Free Software
248 | Foundation.
249 |
250 | 10. If you wish to incorporate parts of the Program into other free
251 | programs whose distribution conditions are different, write to the author
252 | to ask for permission. For software which is copyrighted by the Free
253 | Software Foundation, write to the Free Software Foundation; we sometimes
254 | make exceptions for this. Our decision will be guided by the two goals
255 | of preserving the free status of all derivatives of our free software and
256 | of promoting the sharing and reuse of software generally.
257 |
258 | NO WARRANTY
259 |
260 | 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
261 | FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
262 | OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
263 | PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
264 | OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
265 | MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
266 | TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
267 | PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
268 | REPAIR OR CORRECTION.
269 |
270 | 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
271 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
272 | REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
273 | INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
274 | OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
275 | TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
276 | YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
277 | PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
278 | POSSIBILITY OF SUCH DAMAGES.
279 |
280 | END OF TERMS AND CONDITIONS
281 |
282 | How to Apply These Terms to Your New Programs
283 |
284 | If you develop a new program, and you want it to be of the greatest
285 | possible use to the public, the best way to achieve this is to make it
286 | free software which everyone can redistribute and change under these terms.
287 |
288 | To do so, attach the following notices to the program. It is safest
289 | to attach them to the start of each source file to most effectively
290 | convey the exclusion of warranty; and each file should have at least
291 | the "copyright" line and a pointer to where the full notice is found.
292 |
293 | {description}
294 | Copyright (C) {year} {fullname}
295 |
296 | This program is free software; you can redistribute it and/or modify
297 | it under the terms of the GNU General Public License as published by
298 | the Free Software Foundation; either version 2 of the License, or
299 | (at your option) any later version.
300 |
301 | This program is distributed in the hope that it will be useful,
302 | but WITHOUT ANY WARRANTY; without even the implied warranty of
303 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
304 | GNU General Public License for more details.
305 |
306 | You should have received a copy of the GNU General Public License along
307 | with this program; if not, write to the Free Software Foundation, Inc.,
308 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
309 |
310 | Also add information on how to contact you by electronic and paper mail.
311 |
312 | If the program is interactive, make it output a short notice like this
313 | when it starts in an interactive mode:
314 |
315 | Gnomovision version 69, Copyright (C) year name of author
316 | Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
317 | This is free software, and you are welcome to redistribute it
318 | under certain conditions; type `show c' for details.
319 |
320 | The hypothetical commands `show w' and `show c' should show the appropriate
321 | parts of the General Public License. Of course, the commands you use may
322 | be called something other than `show w' and `show c'; they could even be
323 | mouse-clicks or menu items--whatever suits your program.
324 |
325 | You should also get your employer (if you work as a programmer) or your
326 | school, if any, to sign a "copyright disclaimer" for the program, if
327 | necessary. Here is a sample; alter the names:
328 |
329 | Yoyodyne, Inc., hereby disclaims all copyright interest in the program
330 | `Gnomovision' (which makes passes at compilers) written by James Hacker.
331 |
332 | {signature of Ty Coon}, 1 April 1989
333 | Ty Coon, President of Vice
334 |
335 | This General Public License does not permit incorporating your program into
336 | proprietary programs. If your program is a subroutine library, you may
337 | consider it more useful to permit linking proprietary applications with the
338 | library. If this is what you want to do, use the GNU Lesser General
339 | Public License instead of this License.
340 |
341 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # fast\_sample
2 |
3 | fast\_sample - A script for rapidly sampling a proportion of lines from a file
4 |
5 | # SYNOPSIS
6 |
7 | fast\_sample \[options\] \[file ...\]
8 |
9 | Options:
10 | -proportion|p The proportion of lines to sample (e.g. .5 for half.)
11 | -number|n The number of lines to sample.
12 | -header|h Always print the header for every file
13 | -seed|s The random seed, for reproducibility
14 | -man|m The full man page
15 |
16 | # OPTIONS
17 |
18 | - **-proportion|p**
19 |
20 | This is a floating point number between 0 and 1 which determines the proportion
21 | of lines to sample from the input files.
22 |
23 | - **-number|n**
24 |
25 | This is an integer of the number of lines to be sampled. Sampling a specific
26 | number of lines instead of a constant "coin-flip" proportion is implemented
27 | using resevoir sampling. Wikipedia has a great explanation of [reservoir sampling](http://en.wikipedia.org/wiki/Reservoir_sampling),
28 | but I used the code available [here](http://data-analytics-tools.blogspot.com/2009/09/reservoir-sampling-algorithm-in-perl.html).
29 |
30 | - **-header|h**
31 |
32 | This is a boolean flag, if it exists then the first line of every file will be
33 | printed. This is useful for when you want to keep the header of a CSV file.
34 |
35 | - **-seed|s**
36 |
37 | This is an integer flag, it is the seed passed to the random number generator
38 | which determines which lines are sampled. If you want to make your research
39 | reproducible, make sure to specify a seed, and the same lines will always
40 | be selected.
41 |
42 | - **-man|m**
43 |
44 | The man page for fast\_sample.
45 |
46 | # DESCRIPTION
47 |
48 | **fast\_sample** is a program that allows you to work with a subset of your
49 | data. Sometimes you have a super large file, and you wish you could just
50 | work with 5% of the data. **fast\_sample** let's you do this simply. It also
51 | allows you to sample files reproducibly by simply specifying a random
52 | number seed. **fast\_sample** currently supports line-by-line textual
53 | formats such as CSV, and the DBF format.
54 |
55 | # DEPENDENCIES
56 |
57 | **fast\_sample** attempts to be as smart as possible about requiring 3rd
58 | party modules. If you are just going to use it for just sampling out of
59 | text files (line by line format such as CSV) it should work without the
60 | addition of any 3rd party modules, and any perl (I think going back as
61 | far as 5.6) will work.
62 |
63 | However, if you want to sample binary formats (currently only dbf is
64 | supported), you will unfortunately need to install two modules. In order
65 | to sample DBFs you need [XBase](https://metacpan.org/pod/XBase) for parsing dbf files, and [Text::CSV](https://metacpan.org/pod/Text::CSV)
66 | in order to have "correct" CSV file generation. I could have hand-coded
67 | a chintzy CSV generator, but it would be wrong and would handle weird
68 | stuff incorrectly (like embedded newlines.)
69 |
70 | If you need help installing Perl modules (because you want to use the
71 | dbf file capabilities of fast\_sample), check out the "how do I install
72 | perl modules" documentation available [here](http://www.cpan.org/modules/INSTALL.html).
73 |
74 | # INSTALLATION
75 |
76 | ## Mac and Linux
77 |
78 | **fast\_sample** is very lightweight and requires no 3rd party packages
79 | installed other than a default Perl installation. Perl comes installed
80 | on OSX and Linux, so for both of those, simply clone the repository
81 | and you should be able to execute it at the command line. If you
82 | want to have it available just for your user, copy the **fast\_sample**
83 | script to your ~/bin as follows:
84 |
85 | cp fast_sample ~/bin
86 |
87 | if you want it available for all users in the system, copy it to your
88 | system /usr/local/bin using the following command:
89 |
90 | sudo cp fast_sample /usr/local/bin
91 |
92 | Make sure that the directory you copy the script to is in your search
93 | path. So to summarize, a full installation would look as follows:
94 |
95 | # first we clone the repo
96 | % git clone https://github.com/earino/fast_sample.git
97 | Cloning into 'fast_sample'...
98 | remote: Counting objects: 31, done.
99 | remote: Compressing objects: 100% (26/26), done.
100 | remote: Total 31 (delta 14), reused 18 (delta 5), pack-reused 0
101 | Unpacking objects: 100% (31/31), done.
102 | Checking connectivity... done.
103 |
104 | # then we go into the newly cloned directory
105 | % cd fast_sample
106 |
107 | # then we copy the script to our /usr/local/bin
108 | % sudo cp fast_sample /usr/local/bin
109 | Password:
110 | %
111 |
112 | ## Windows
113 |
114 | I have not installed a perl script on windows in a very long time, so
115 | I unfortunately do not know how to do this. If you want to use
116 | **fast\_sample** on Windows, drop me a note and I'll figure out how to
117 | get this done :-)
118 |
119 | # PERFORMANCE
120 |
121 | ## Text Files
122 |
123 | **fast\_sample** attempts to be as fast as possible. Sampling should be
124 | effortless even when dealing with huge files.
125 |
126 | $ ls -alh big.csv
127 | -rw-r--r-- 1 earino staff 3.1G Feb 7 09:15 big.csv
128 | $ time fast_sample -h -p .001 big.csv > sampled.csv
129 |
130 | real 0m5.949s
131 | user 0m5.209s
132 | sys 0m0.694s
133 | $ wc -l big.csv
134 | 12174947 big.csv
135 | $ wc -l sampled.csv
136 | 12277 sampled.csv
137 |
138 | When given an integer count via the -n flag, the system executes the
139 | resevoir sampling algorithm for fair sampling across a stream. I don't
140 | believe this is needed for the "coinflip" percentage of lines approach,
141 | but someone better than me at statistics can probably chime in if I'm
142 | wrong. Either way, the reservoir sampling is also relatively fast:
143 |
144 | $ time ./fast_sample -h -n 3 big.csv > /dev/null
145 |
146 | real 0m8.615s
147 | user 0m7.928s
148 | sys 0m0.685s
149 |
150 | ## DBF Files
151 |
152 | **fast\_sample** has to be clever about DBF files, they are clearly not a
153 | particularly fast format for linear access, so a simple coinflip approach
154 | did not work. Current performance seems pretty acceptable. 12 seconds to
155 | sample .001 of a nearly 2 gigabyte dbf file with over 38 million rows.
156 |
157 | $ ls -alh rp19682011.dbf
158 | -rw-r--r--@ 1 earino staff 1.9G Oct 16 14:29 /Users/earino/Downloads/rp19682011.dbf
159 | $ time ./fast_sample -p .001 -h ~/Downloads/rp19682011.dbf > /dev/null
160 |
161 | real 0m12.004s
162 | user 0m3.707s
163 | sys 0m1.267s
164 |
165 | # AUTHOR
166 |
167 | The home for **fast\_sample** is on github at https://github.com/earino/fast_sample
168 |
169 | Eduardo Arino de la Rubia
170 |
171 | # ACKNOWLEDGEMENTS
172 |
173 | Reservoir sampling code from [Program-o-Babble](http://data-analytics-tools.blogspot.com/2009/09/reservoir-sampling-algorithm-in-perl.html).
174 | Motivation to implement reservoir sampling in the first place provided by [Neal Fultz](https://github.com/nfultz).
175 |
--------------------------------------------------------------------------------
/fast_sample:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env perl
2 |
3 | use warnings;
4 | use strict;
5 |
6 | use Getopt::Long;
7 | use Pod::Usage;
8 | use Data::Dumper;
9 |
10 | my $proportion=1;
11 | my $header=0;
12 | my $seed;
13 | my $man;
14 | my $number;
15 |
16 | GetOptions("proportion=f", \$proportion,
17 | "header", \$header,
18 | "seed=s", \$seed,
19 | 'man' => \$man,
20 | 'number=i' => \$number)
21 | or pod2usage(2);
22 |
23 | pod2usage(-exitval => 0, -verbose => 2) if $man;
24 |
25 | srand($seed) if $seed;
26 |
27 | die "Proportion must be between 0 and 1" if $proportion > 1 || $proportion < 0;
28 |
29 | for my $input_file (@ARGV) {
30 | die "ERROR: $input_file doesn't exist.\n" unless -e $input_file;
31 | die "ERROR: $input_file can not be read.\n" unless -r $input_file;
32 |
33 | if (-T $input_file) {
34 | open(INFILE, "<", $input_file)
35 | or die "ERROR: $input_file can not be opened, reason: $!\n";
36 |
37 | # this branch does the 'stream' coinflip sampling instead of the
38 | # resevoir sampling
39 | unless(defined $number) {
40 | while () {
41 | print && next if ($. == 1 && $header);
42 | print $_ if rand() < $proportion;
43 | }
44 | }
45 | # this branch handles reservoir sampling
46 | else {
47 | my @sample = ();
48 |
49 | while () {
50 | print && next if ($. == 1 && $header);
51 | if ($. <= $number) {
52 | $sample[$.-1] = $_;
53 | }
54 | elsif (($. > $number) && (rand() < $number/$.)) {
55 | my $replace = int(rand(@sample));
56 | $sample[$replace] = $_;
57 | }
58 | }
59 | print foreach (@sample);
60 | }
61 |
62 | close(INFILE);
63 | }
64 | else {
65 | if ($input_file =~ /\.dbf$/i) {
66 | my $rc = eval
67 | {
68 | require XBase;
69 | 1;
70 | };
71 |
72 | die "ERROR: Parsing dbf files requires installation of the perl XBase module.\n"
73 | unless $rc;
74 |
75 | $rc = eval
76 | {
77 | require Text::CSV;
78 | my $csv = Text::CSV->new ( { binary => 1 } ) # should set binary attribute.
79 | or die "Cannot use CSV: ".Text::CSV->error_diag ();
80 |
81 | 1;
82 | };
83 |
84 | die "ERROR: Parsing dbf files requires installation of the perl Text::CSV module for correctness.\n"
85 | unless $rc;
86 |
87 | my $csv = Text::CSV->new ( { binary => 1 } ) # should set binary attribute.
88 | or die "Cannot use CSV: ".Text::CSV->error_diag ();
89 |
90 | my $table = new XBase $input_file or die XBase->errstr;
91 | if ($header) {
92 | my $status = $csv->combine($table->field_names);
93 | my $line = $csv->string();
94 | print "$line\n";
95 | }
96 |
97 | my $total_records = $table->last_record;
98 | my $desired_records = defined $number ? $number : int($total_records * $proportion);
99 | my @record_indexes = ();
100 | if (! defined $number && $proportion == 1) {
101 | @record_indexes = (0 .. $total_records)
102 | }
103 | else {
104 | while ($desired_records--) {
105 | push(@record_indexes, int(rand($total_records)));
106 | }
107 | @record_indexes = sort {$a <=> $b} @record_indexes;
108 | }
109 |
110 | for my $i (@record_indexes) {
111 | my @data = $table->get_record($i);
112 | my $is_deleted = shift(@data); #remove the deleted indicator
113 | next if $is_deleted;
114 |
115 | my $status = $csv->combine(@data);
116 | my $line = $csv->string();
117 | print "$line\n";
118 | }
119 | }
120 | else {
121 | die "ERROR: $input_file is an unknown format.\n fast_command only knows how to parse text and dbf files.\n"
122 | }
123 | }
124 | }
125 |
126 | __END__
127 | =head1 fast_sample
128 |
129 | fast_sample - A script for rapidly sampling a proportion of lines from a file
130 |
131 | =head1 SYNOPSIS
132 |
133 | fast_sample [options] [file ...]
134 |
135 | Options:
136 | -proportion|p The proportion of lines to sample (e.g. .5 for half.)
137 | -number|n The number of lines to sample.
138 | -header|h Always print the header for every file
139 | -seed|s The random seed, for reproducibility
140 | -man|m The full man page
141 |
142 | =head1 OPTIONS
143 |
144 | =over 8
145 |
146 | =item B<-proportion|p>
147 |
148 | This is a floating point number between 0 and 1 which determines the proportion
149 | of lines to sample from the input files.
150 |
151 | =item B<-number|n>
152 |
153 | This is an integer of the number of lines to be sampled. Sampling a specific
154 | number of lines instead of a constant "coin-flip" proportion is implemented
155 | using resevoir sampling. Wikipedia has a great explanation of L,
156 | but I used the code available L.
157 |
158 |
159 | =item B<-header|h>
160 |
161 | This is a boolean flag, if it exists then the first line of every file will be
162 | printed. This is useful for when you want to keep the header of a CSV file.
163 |
164 | =item B<-seed|s>
165 |
166 | This is an integer flag, it is the seed passed to the random number generator
167 | which determines which lines are sampled. If you want to make your research
168 | reproducible, make sure to specify a seed, and the same lines will always
169 | be selected.
170 |
171 | =item B<-man|m>
172 |
173 | The man page for fast_sample.
174 |
175 | =back
176 |
177 | =head1 DESCRIPTION
178 |
179 | B is a program that allows you to work with a subset of your
180 | data. Sometimes you have a super large file, and you wish you could just
181 | work with 5% of the data. B let's you do this simply. It also
182 | allows you to sample files reproducibly by simply specifying a random
183 | number seed. B currently supports line-by-line textual
184 | formats such as CSV, and the DBF format.
185 |
186 | =head1 DEPENDENCIES
187 |
188 | B attempts to be as smart as possible about requiring 3rd
189 | party modules. If you are just going to use it for just sampling out of
190 | text files (line by line format such as CSV) it should work without the
191 | addition of any 3rd party modules, and any perl (I think going back as
192 | far as 5.6) will work.
193 |
194 | However, if you want to sample binary formats (currently only dbf is
195 | supported), you will unfortunately need to install two modules. In order
196 | to sample DBFs you need L for parsing dbf files, and L
197 | in order to have "correct" CSV file generation. I could have hand-coded
198 | a chintzy CSV generator, but it would be wrong and would handle weird
199 | stuff incorrectly (like embedded newlines.)
200 |
201 | If you need help installing Perl modules (because you want to use the
202 | dbf file capabilities of fast_sample), check out the "how do I install
203 | perl modules" documentation available L.
204 |
205 | =head1 INSTALLATION
206 |
207 | =head2 Mac and Linux
208 |
209 | B is very lightweight and requires no 3rd party packages
210 | installed other than a default Perl installation. Perl comes installed
211 | on OSX and Linux, so for both of those, simply clone the repository
212 | and you should be able to execute it at the command line. If you
213 | want to have it available just for your user, copy the B
214 | script to your ~/bin as follows:
215 |
216 | cp fast_sample ~/bin
217 |
218 | if you want it available for all users in the system, copy it to your
219 | system /usr/local/bin using the following command:
220 |
221 | sudo cp fast_sample /usr/local/bin
222 |
223 | Make sure that the directory you copy the script to is in your search
224 | path. So to summarize, a full installation would look as follows:
225 |
226 | # first we clone the repo
227 | % git clone https://github.com/earino/fast_sample.git
228 | Cloning into 'fast_sample'...
229 | remote: Counting objects: 31, done.
230 | remote: Compressing objects: 100% (26/26), done.
231 | remote: Total 31 (delta 14), reused 18 (delta 5), pack-reused 0
232 | Unpacking objects: 100% (31/31), done.
233 | Checking connectivity... done.
234 |
235 | # then we go into the newly cloned directory
236 | % cd fast_sample
237 |
238 | # then we copy the script to our /usr/local/bin
239 | % sudo cp fast_sample /usr/local/bin
240 | Password:
241 | %
242 |
243 | =head2 Windows
244 |
245 | I have not installed a perl script on windows in a very long time, so
246 | I unfortunately do not know how to do this. If you want to use
247 | B on Windows, drop me a note and I'll figure out how to
248 | get this done :-)
249 |
250 | =head1 PERFORMANCE
251 |
252 | =head2 Text Files
253 |
254 | B attempts to be as fast as possible. Sampling should be
255 | effortless even when dealing with huge files.
256 |
257 | $ ls -alh big.csv
258 | -rw-r--r-- 1 earino staff 3.1G Feb 7 09:15 big.csv
259 | $ time fast_sample -h -p .001 big.csv > sampled.csv
260 |
261 | real 0m5.949s
262 | user 0m5.209s
263 | sys 0m0.694s
264 | $ wc -l big.csv
265 | 12174947 big.csv
266 | $ wc -l sampled.csv
267 | 12277 sampled.csv
268 |
269 | When given an integer count via the -n flag, the system executes the
270 | resevoir sampling algorithm for fair sampling across a stream. I don't
271 | believe this is needed for the "coinflip" percentage of lines approach,
272 | but someone better than me at statistics can probably chime in if I'm
273 | wrong. Either way, the reservoir sampling is also relatively fast:
274 |
275 | $ time ./fast_sample -h -n 3 big.csv > /dev/null
276 |
277 | real 0m8.615s
278 | user 0m7.928s
279 | sys 0m0.685s
280 |
281 | =head2 DBF Files
282 |
283 | B has to be clever about DBF files, they are clearly not a
284 | particularly fast format for linear access, so a simple coinflip approach
285 | did not work. Current performance seems pretty acceptable. 12 seconds to
286 | sample .001 of a nearly 2 gigabyte dbf file with over 38 million rows.
287 |
288 | $ ls -alh rp19682011.dbf
289 | -rw-r--r--@ 1 earino staff 1.9G Oct 16 14:29 /Users/earino/Downloads/rp19682011.dbf
290 | $ time ./fast_sample -p .001 -h ~/Downloads/rp19682011.dbf > /dev/null
291 |
292 | real 0m12.004s
293 | user 0m3.707s
294 | sys 0m1.267s
295 |
296 | =head1 AUTHOR
297 |
298 | The home for B is on github at https://github.com/earino/fast_sample
299 |
300 | Eduardo Arino de la Rubia
301 |
302 | =head1 ACKNOWLEDGEMENTS
303 |
304 | Reservoir sampling code from L.
305 | Motivation to implement reservoir sampling in the first place provided by L.
306 |
--------------------------------------------------------------------------------