Autonomously crawling through DICE job postings
Posted: April 13, 2012 Filed under: Operating Systems, Programming Languages | Tags: Bash, Mac OS X, Mathematica, Perl Leave a comment »I am currently working on a book that requires me to search through thousands of job advertisements. For the last couple of days I have been looking at the various websites, collecting data and looking for patterns in employment listings. Even if you are not working on a book, I am sure at some point in time you will be looking for a new job online. I love searching for jobs, and if you don’t love it too… you are probably doing it wrong.
First of all don’t manually search for jobs! It is a waste of time and it will drive you insane. Instead use a scripting language, such as PERL, that mines website databases and outlines the best matches. In fact I wrote a post a few days ago about mining employment postings on craigslist. If you are new to this entire field of data-mining, I recommend the book “mining the social web” by Russell… Nice chap…. Met him at Harvard Square a couple of years ago.
Here I outline the steps I took to extract all job postings from DICE. First of all you have to know how everything is stored in the database. Make any random search on the initial screen (e.g. embedded).
"Embedded" search on dice.com
This particular search generated the following very-long URL… so long that I had to include spaces:
Since this particular search detected 1689 job postings, we just have to change NUM_PER_PAGE=30 from 30 to 1689, in order to see every single job post on a single page. Save that file into your hard-disk in the HTML format. For completeness, here is the file with all 1689 postings I just downloaded. The following PERL script parses the contents of this file and looks for the associated URL for each job posting.
|
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 |
#neat function that helps catching scripting errors like undefined variables... use strict; #use the generated HTML as an input argument my $input_file=$ARGV[0]; #parse through every single line in the file open (MYFILE, "<$input_file") or die $!; while (<MYFILE>) { chomp; #each job posting starts with a well defined pattern... #search for <div><a href="/jobsearch/servlet/Jo if (/<div><a href=\"\/jobsearch\/servlet\/Jo/) { #...the following line: #print "$_\n"; #...will print for example: #<div><a href="/jobsearch/servlet/JobSearch?op=302&dockey=xml/7/0/70b58be12a872e8268939a525389b927@endecaindex&source=19&FREE_TEXT=embedded&rating=99">ASIC/FPGA Verification Engineer</a></div> #get the job title my $job_title; my @tmp_data; @tmp_data=split(/">|<\/a><\/div>/,$_); $job_title=$tmp_data[1]; #...the following line: #print $job_title . "\n"; #...will print for example: #ASIC/FPGA Verification Engineer @tmp_data=split(/">|div><a href="\//,$_); #...the following line: #print $tmp_data[1] . "\n"; #...will print for example: #jobsearch/servlet/JobSearch?op=302&dockey=xml/7/0/70b58be12a872e8268939a525389b927@endecaindex&source=19&FREE_TEXT=embedded&rating=99 #but that is not the correct URL... instead we want something like: #http://seeker.dice.com/jobsearch/servlet/JobSearch?op=302&dockey=xml/b/4/b4ba4b9ed60baf2cf7a3397f336e451e@endecaindex&source=19&FREE_TEXT=embedded&rating=0 #in essence, we must replace all & with & $tmp_data[1] =~ s/\&/\&/g; $tmp_data[1] =~ s/\/jobsearch\/servlet//g; print "http://seeker.dice.com/" . $tmp_data[1] . "\n"; } } close (MYFILE); |
Save the file (e.g. dice.pl) and execute it with the following command:
This will store every single URL, one per line, in the file embedded_url.txt. Once again… for completeness, here is my generated file.
The next step is to download every single job posting onto a separate file. Since, I am a macsox user, I need to download a the contents of each of the URLs from the web via the OS X command line. This is easy accomplished with the following bash script:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 |
FILENAME=embedded_url.txt NUMBERLINES=`wc -l < ${FILENAME}` echo ${NUMBERLINES} > xxx.tmp PNUMBERLINES=`perl -n -e '@splitline=split(/ /,$_); $splitline[1]=~s/ //g; print $splitline[0] ."\n"; ' xxx.tmp` rm -f xxx.tmp echo ${PNUMBERLINES} RUN=0 until [ ${RUN} -eq ${PNUMBERLINES} ] do RUN=$(( $RUN + 1 )) LCONTENTS=`sed -n $RUN'p' ${FILENAME}` #echo "line # ${RUN} with contents : ${LCONTENTS}" wget "${LCONTENTS}" done |
On the same directory as the output of the previous PERL script (e.g. embedded_url.txt), save this bash scrip (e.g. download_all_jobs.sh) and execute it with the following commands:
./download_all_jobs.sh embedded_url.txt
The compressed outcome of this last step is a file of 27 MBs.
Now that I have all this data, I need to extract the skills required for each advertised position. So, I placed all the compressed files in the sub-directory dice_jobs and ran the following script:
|
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 |
#!/bin/bash JOBS_DIRECTORY=dice_jobs ls ${JOBS_DIRECTORY}/* > all_jobs.txt #count the number of lines on each file impcount=`wc -l < all_jobs.txt` echo ${impcount} > xxx.tmp NIMPS=`perl -n -e '@splitline=split(/ /,$_); $splitline[1]=~s/ //g; print $splitline[0] ."\n"; ' xxx.tmp`; rm -f xxx.tmp #rename jobs postings into something more readable RUN=1 until [ $RUN -gt ${NIMPS} ] do FNAME=`sed -n $RUN'p' all_jobs.txt` mv ${FNAME} ${JOBS_DIRECTORY}/${RUN}.txt RUN=$(( $RUN + 1 )) done #extract the job title and necessary skills rm -f parsed_job_data.txt RUN=1 until [ $RUN -gt ${NIMPS} ] do FNAME=`sed -n $RUN'p' all_jobs.txt` perl extract_data.pl ${JOBS_DIRECTORY}/${RUN}.txt >> parsed_job_data.txt RUN=$(( $RUN + 1 )) done rm -f all_jobs.txt echo "...parsed job data is @ parsed_job_data.txt" |
The skill extraction is actually done on the following PERL script (extract_data.pl).
|
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 |
use strict; #use the generated HTML as an input argument my $input_file=$ARGV[0]; my $next_line_is_job_title=0; my $jobTitle=""; my $next_line_is_area_code=0; my $areaCode=""; my $next_line_is_skills=0; my $skills=""; my $next_line_is_company_name=0; my $companyName=""; #parse through every single line in the file open (MYFILE, "<$input_file") or die $!; while (<MYFILE>) { #chomp; if ($next_line_is_skills==1) { $skills=$_; $skills =~ s/<dd>|<dt>|<\/dt>|<\/dd>| |\t|\n//g; $skills =~ s/\/assets\/images\/detail\/default\/highlite.gif//g; $skills =~ s/<b style="background:url\(//g; $skills =~ s/\); font-weight: bold;">|<\/b>//g; #print "skills=" . $skills . "\n"; $next_line_is_skills=0; } if ($next_line_is_company_name==1) { $companyName=$_; $companyName =~ s/<dd>|<dt>|<\/dt>|<\/dd>| |\t|\n//g; my @splitString=split(/>|</,$companyName); $companyName = $splitString[2]; $next_line_is_company_name=0; #print "companyName = " . $companyName . "\n"; } if ($next_line_is_area_code==1) { $areaCode=$_; $areaCode =~ s/<dd>|<\/dd>|<dt>|<\/dt>| |\t|\n//g; $next_line_is_area_code=0; #print "areaCode=" . $areaCode . "\n"; } if (/<h1 id="jobTitle">/) { my @split_tmp=split(/>|</,$_); $jobTitle=@split_tmp[2]; #print "jobTitle=" . $jobTitle . "\n"; } if (/<dt>Skills:<\/dt>|<dd>Skills<\/dd>/) { $next_line_is_skills=1; } if (/<dt>Company:|<dd>Company/) { $next_line_is_company_name=1; } if (/<dt>Area Code:|<dd>Area Code/) { $next_line_is_area_code=1; } } close (MYFILE); print "$input_file\t$companyName\t$jobTitle\t$areaCode\t$skills\n"; |
I then feed the extracted data into a mathematica script; a (readable) pdf version of the Mathematica script is here, and the source is here. In this script, I combined all found skills, ignored skills that were required in less than 30 distinct advertisement (e.g. COBOL and Pascal). Below is the resulting piechart.
Most requested skills in embedded computing jobs.
As expected the most sought after skills in “embedded computing” jobs are C,C++ and Linux. Java, mysql and kernel development is also very strong in demand these days. Surprisingly I saw lots of mobile computing and networking skill requests. However the most surprisingly requested skills is databases (mysql)!
Finally, I am aware that I could have done everything on this post on a single PERL script. However, writing a post about a single script would get tedious very quickly. I also wanted to save the outcome of every single step in my hard-disk so I could perform some additional data tests, without having to connect to dice.com each time.
Automated craigslist job search with Perl and Bash
Posted: March 30, 2012 Filed under: Operating Systems, Programming Languages | Tags: Bash, Linux, Mac OS X, Perl Leave a comment »Most of my students are in the job market and after suggesting them websites where they could look for jobs, I took a peak at craigslist. I like craigslist; its a simple, bare-bones website with pure text. However, the search functionality is a bit awkward, and it is hard to find a good match between the candidate skills and a particular job posting. If you are seriously looking at every single “filtered” post, it may still take you over an hour to look for the best skill-to-job matches. So I created two scripts, one in Perl and the other one in Bash, that scavenge all the job postings for skill matches, and create a new webpage with all the appropriate positions and matched skills in a ready to click link. Data mining at its best!
There are some “limitations” of these scripts. First of all they were only tested in macOSX and Linux, however I am sure you can convert them quite easily to Windows. Secondly, I’ve focused all the craigslist searches around New England. You may add other craigslist locations quite easily by following the instructions on the perl script.
This automated job search requires two files: search_jobs.sh and craigslist.pl. Both can be found below, or at my github repository. To run the code place both files in the same directory, and edit the search_jobs.sh, shown below, with a text editor (after emacs, my second favorite text editor is TextWrangler). In this file modify the appropriate keywords that are being assigned to the variable SEARCH_SKILLS. Currently the search skills are the standard qualifications for an engineering graduate.
|
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 |
#!/bin/bash ################################################################################# #Functions (do not modify anything here) ################################################################################# SEARCH_SKILLS="" function search_jobs { SEARCH_NAME=${1} perl craigslist.pl ${SEARCH_SKILLS}> raw_data.txt #create the header of an html file echo "<html><title>Job Search Results</title><body>" > job_data.html #sort the entire file contents and make sure the best matches are on top sort -t! -n -r -k3 raw_data.txt >> job_data.html #clean up the file perl -p -i -e "s/!/\ \ \ \ \ \ /g" job_data.html #terminate the html file echo "</body></html>" >> job_data.html mv job_data.html ${SEARCH_NAME}.html rm -f raw_data.txt } ################################################################################# #You may modify your skills below ################################################################################# SEARCH_SKILLS="embedded, circuit, transistor, VLSI, firmware, RTOS, kernel, MacOSX, JTAG, oscilloscope, HDL, FPGA, Arduino, MSP430, OMAP3540, micro-controllers, microcontrollers, SVN, programmer, Perl, linux, Mathematica, LabVIEW, schematics, Verilog, VHDL" SEARCH_NAME="engineering" search_jobs ${SEARCH_NAME} SEARCH_SKILLS="quantitative, mathematica, finance, programmer, developer, high-frequency, fpga, microcontroller" SEARCH_NAME="finance" search_jobs ${SEARCH_NAME} |
This script runs with the following command line:
./search_jobs.sh
Running the scripts on a macosx terminal window.
After it is done executing it will create two files engineering.html and finance.html, where the candidate can see his best job matches.
Two html files are created with the best job matches
Generated HTML file with the best job matches
Below is the Perl script that parses the craigslist job postings.
|
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 |
#This script fetches the last 2 days new job postings from craigslist that match #a specific criteria and reports the URLs that correspond to that match. #The search criteria comes from the input arguments. The cragislist sites #are hardwired to the New England area. You may change them by manually #altering the variables in Section #3. # #Version 0.2 30/march/2012 #Author: Nuno Alves # ############################################################################# #Section #1 - load libraries ############################################################################# use strict; use POSIX; use LWP::Simple; ############################################################################# #Section #2 - input arguments are your skillsets ############################################################################# my $num_args = $#ARGV + 1; if ($num_args == 1) { print "You must add some skills as arguments\n"; exit; } ############################################################################# #Section #3 - defining variables ############################################################################# #what cragislist sites my @search_site=("http://boston.craigslist.org","http://nh.craigslist.org","http://maine.craigslist.org","http://burlington.craigslist.org","http://westernmass.craigslist.org","http://worcester.craigslist.org"); #type what positions you are looking for (egr = engineering, sof = software) my @positions=("egr","sof","bus","acc"); #this array contains the arguments which are your resume skills my @skills=@ARGV ; ############################################################################# #Section #4 - debug code ############################################################################# #instead of work on every single URL, setting $debug=1, will just scan #two webpages my $debug=0; my @debug_urls=("http://boston.craigslist.org/gbs/egr/2902012136.html","http://boston.craigslist.org/bmw/egr/2929181526.html","http://boston.craigslist.org/gbs/egr/2926742528.html"); ############################################################################# #Section #5 - subroutines for collecting craigslist data ############################################################################# sub collect_job_posting_http { my $url=$_[0]; my $content = get $url; #print $content . "\n"; my @splitcontents=split(/<h4 class=\"ban\"/,$content); my $size_splitcontents=@splitcontents; my @url_data=(); for (my $i=1 ; $i<$size_splitcontents ; $i++) { #just want the last 2 days of postings if ($i<3) { #print "============\n\n\n"; #print $splitcontents[$i] . "\n"; #get all the posting urls for this particular day my @postingdata=split(/<p><a href=\"|\">/,$splitcontents[$i]); for (my $j=0; $j<@postingdata ; $j++) { #print ">>[$j]>>" . $postingdata[$j] . "<<<\n"; if ($postingdata[$j]=~m/^http/) { push(@url_data,$postingdata[$j]); } } } } return(@url_data); } sub extract_date { my @url_data=$_[0]; my @date_data=split(/Date: 2012-|EDT<br>/,$url_data[0]); return("2012-" . $date_data[1]); } ############################################################################# #Section #6 - main program: collecting http data for each job posting ############################################################################# my @urls=(); if ($debug == 0) { for (my $k=0;$k<@search_site;$k++) { for (my $z=0;$z<@positions;$z++) { my $base_url=$search_site[$k]."/".$positions[$z]; my @tmp_data=collect_job_posting_http $base_url; push(@urls,@tmp_data); } } } else { @urls=@debug_urls; } #foreach (@urls) #{ # print $_ . "\n"; #} ############################################################################# #Section #7 - check if each posting matches at least one skill ############################################################################# my @matched_skills=(); my @skill_type=(); my @post_date=(); for (my $i=0 ; $i<@urls ; $i++) { my $url=$urls[$i]; my $content = get $url; my $counter=0; my $date; # print $url . "\n"; # print $content . "\n"; my $skill_type_desc=""; for (my $k=0; $k<@skills ; $k++) { if ($content =~ m/$skills[$k]/i) { $counter++; $skill_type_desc = $skill_type_desc . $skills[$k] . " "; } } push(@matched_skills,$counter); push(@skill_type,$skill_type_desc); push(@post_date,extract_date($content)); } ############################################################################# #Section #8 - print results to the screen ############################################################################# for (my $i=0; $i < @matched_skills ; $i++) { if ($matched_skills[$i]>0) { print "<li><a href=\"$urls[$i]\">site #$i\<\/a\>" . "!" . $post_date[$i] . "!" . $matched_skills[$i] . "!" . $skill_type[$i] . "\n"; } } |
Using PERL to select a random lines from a file
Posted: March 10, 2012 Filed under: Programming Languages | Tags: Perl Leave a comment »This is a neat PERL script, which takes in a text file as an input, and prints a specified number unique random lines from that file. In order for you to run this script, save the following code into a file and run it from the command line with the argument is the file name and the second argument is the number of random lines you want to print.
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
#!/usr/bin/perl die "Usage: $0 , where N is the number of lines to pick\n" if @ARGV < 1; $N = shift @ARGV; @pick = (); while (<>) { if (@pick < $N) { push @pick, $_; ($r1, $r2) = (rand(@pick), rand(@pick)); ($pick[$r1], $pick[$r2]) = ($pick[$r2], $pick[$r1]); } else { rand($.) <= $N and $pick[rand(@pick)] = $_; } } print @pick; |
