With git, to pull in changes from another branch but not commit them, you'd think running 'git merge --no-commit ' would do the trick. Well, you'd be wrong because "merge" is an overloaded word. It's specific meaning in git is to combine histories, not working files. If there were no commits on the original branch, then it just fast-forwards and --no-commit still commits - a bit counterintuitive. Solution is to run 'git merge --no-ff --no-commit '.
I wanted to test the speed difference of a Wordpress blog before and after WP-Supercache was set up. Here's how I did it, using just bash, curl, sed and awk:
Basically, time how long it takes to get the website about 100 times, and get the average. For my non-cached site on a low end Linode VPS this took between 0.8 and 1.3 seconds. After setting up WP-Supercache, the average time dropped to under 0.5 seconds -- a 40-60% improvement in pure speed. So, WP-Supercache works!
Bonus: I ran the same script to load the page 100 times while running top on the server and eyeballing the numbers. The results were impressive - with no caching there were 2-5 apache processes running, each taking about 5% memory and 8% cpu. That is a recipe for disaster and the site would probably last about 2 seconds before dying under the Digg effect. With WP-Supercache fully enabled, there was only a single apache process running, consuming 5% memory and virtually no cpu.
If you're interested in more detail about how the command line stuff works, here is a brief explanation. The first loop runs 100 times, piping the output of time to a text file. The curl command itself (the one being timed) is executed in a subshell and its output is discarded). The second loop is identical, just saving to a different file to record the new speeds.
The last line is meant to be saved as an executable file so it is convenient for reuse. It takes two arguments, a file and the type of time you want the average of (from the time command; you want "real"). So the script prints out the file with cat, separates the actual time value nicely with a search-and-replace regular expression with sed, then uses awk to find the average by looking at the correct times (real/user/sys), counting those lines and summing the total, and dumping the average when complete.
It took me longer to write this brief post than it did to come up with that script. There must be commercial or free benchmarking software for websites that exists, but this worked for me and didn't require any wasted time searching and evaluating other options.
Comments [0]
I always used to insert date/time into vim by running
:r!date
which inserts the full date in its own line. Well and good, but if you want to insert on the current line it would take you a few keystrokes. And if you want just the time inserted...well text expansion is a perfect candidate for this. The following code lets you type 'ptime' in insert mode and it gets replaced with the current time (e.g. 3:28 PM). Cool.
Comments [0]
Comments [0]
Comments [0]
Bill Gates didn’t crack down much on Chinese piracy because they were a developing country. they much rather would have people pirate their software than other peoples’ software, and Gates believed that they would pay later because of the then-majority of software being Microsoft. and—they did pay later and is now a huge market for Microsoft
(via YCombinator's 2009 Startup School presentation from Chris Anderson, summary from http://journal.markbao.com/2009/10/startup-school-2009-summary/)Comments [0]
I was testing a near-completion Drupal website today, and noticed that I could not log in with Internet Explorer (8). Firefox worked, Chrome worked, Safari worked...so what was the problem? Turns out there was an underscore in the domain name, and that gave IE fits. I changed the subdomain to use a dash instead, and the login worked.
Comments [0]
Sometimes I see an ad that I actually want to look at, but lose it because the ad spot refreshed and put something new in its place. Ad spots should really have forward/back buttons. Google does this with Web Clips in Gmail, and the text block ads have this as well (not sure if they all do, or just sometimes, but I have seen it).
The benefits are obvious - if a user clicks a forward or back button, then their engagement is pretty strong and the impressions per page skyrocket. Also, those buttons could really turbocharge analytics for "best performing ad" in a sector.Comments [0]
A week or two ago, I removed TechCrunch and some similar sites from my feed reader. I'm a voracious consumer of RSS feeds, but recently decided there was a serious need to fix the signal-to-noise ratio. Since I started reading TechCrunch, it changed from a stream of interesting tech business news to a fountain of sensationalist headlines. Some of the content is great, but the noise is too much. My theory was that even if I never directly read TechCrunch, the best of the bunch will get to me anyway through Twitter or Digg. So far that theory has proved unequivocally correct.
Other sites to get the axe were New York Times and (tragically) Wired Top Stories. I tried filtering the feeds through Postrank to view only the higher post-ranked entries, but "popular" as defined by Postrank doesn't necessarily capture "interesting". I would also guess that it might end up selecting the silly, sensationalist, and controversial articles that drove me away from TechCrunch in the first place, because by their nature of being controversial, there would be lots of activity on the page. Postrank is a neat idea though, and I've played around with my own algorithm to accomplish a similar goal; maybe I will release it at some point.Comments [0]
Gmail's web interface has suffered some downtime recently, no doubt sparking panic among many users. However, the IMAP connection has stayed up, meaning that if you use a 3rd party email program (like Mail.app or Outlook), then you could have used Gmail without interruption. You may not even have noticed the downtime if it wasn't for the thousands of tweets lamenting the outage!
Comments [0]
Comments [0]