├── .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 | --------------------------------------------------------------------------------