Final Update to This Post For those not following along at home, I finally took Twitter Stats to the next level and released a webified version over at TweetStats.com. I, somewhat unfortunately, had to go with TweetStats as twitterstats.com was already taken. :( I made a post about it here and you can see an example of my stats on the site on the graphs page.
So I’ve been a user on Twitter for a little over a year, but it wasn’t until recently when I hit 2000 tweets that I wanted to see what my Twitter history looked like over that period. Ever being the statistics nerd, I pulled down all of my tweets and using a combination of curl, sed, grep, Excel, and Numbers, managed to generate some nice graphs.
Being the automation weenie that I am, I eventually hacked together a perl script that did everything except paste the data into Numbers.Although I won’t post it here (because I think the Twitterocracy would have a cow with how it’s implemented), you can DM me your email and I’ll send you the code and instructions. See below - bugs be damned, I’ve made it publicly available.
Basically, the script pulls down all your tweets and stores them in an csv file. It then runs some statistics on the csv file and then copies the resulting stats to the OS X clipboard to paste into each table within Numbers. If run with a pre-existing tweets csv file, the script will calculate the different between your current status count and only download the pages necessary, thus saving the Twitter servers from some bandwidth. ;)
For the record, here are mine. :-D
Update Thanks to kosmar for pointing out that I can actually do this entire thing without your password. Head => Wall. I’ll be updating it accordingly and hopefully even making a webservice out of it soon. :)
In addition to not needing your password, the script should also adjust the times for your tweets to whatever the system time is where you run the script.
Another update: I’ve also posted the script on my site and you can download it here: twitter_stats.zip. Feel free to contact me with any questions via twitter or web.dpc at dcortesi . com.
Many people have noticed a large after-lunch spike around 2pm. At least for me, this was due to Twitter being down most of the morning one day and then tweeting like crazy when it came back online.
BUG FOUND AND SQUASHED
There was a small bug that cropped up after I switched the script to not require your password. It accounts for the odd “[Tuesday|January|2pm] Peak” that people were seeing. This bug has now been fixed and an updated script is available. My apologies.
Unfortunately, if want the most accurate tweets, you will have to rm your csv file and run the script again.
Date::Calc aka failure on line 13 Some of you (on Tiger?) may be missing the Date::Calc module that I use to figure out weekdays. Although I tried to use as few perl modules as possible, this one was essential. Use the following command (thanks to a couple twitter peeps for the reference) to install:
1
| |
and keep hitting ‘y’. ;)
Final (hopefully) Update on this page as it’s getting messy.
For those of you not on OS X with Numbers, there are a few options: @bck webified my code (w00t): Twitter Stats @mmc decided to use gnuplot: Twitter Stats in SVG Using GNUPlot @cbarrett modified it to utilize the Google Chart API: Twitter Stats with GChart @kejadlen reverse engineered my original script to Ruby: Twitter Stats in Ruby
I still want to write my own webified version (Google Chart aesthetics leave a little bit to be desired…), but I have yet to settle on an option that I like.