shell

Sometimes I wonder why I ever bothered to compile wget for OS X. I keep finding clever new things that curl can do — even some that make me wonder why I bother with Perl and Ruby's Mechanize module/gem.

For instance: today I found myself needing to export a friend's Typepad blog. He's got a lot of content, and I wanted to pull it down on a speedy connection and then gzip it before pulling it onto my laptop over my sluggish DSL (Typepad unhelpfully delivers its exported content in an uncompressed non-ML format).

Typepad provides a URL for the export, but it only works when you've authenticated through their website. Curl makes working around this pretty simple, though, as I found out from this Ask Metafilter thread. Here's how it works:

curl -k -d "__mode=redir&next=http://www.typepad.com/t/app&username=user%27s+name&password=their%20password" -c cookie.txt https://www.typepad.com/t/app

You simply use -k to turn off SSL requirements, -d to pass a series of key/value form pairs (properly url-encoded) to the URL to which the form posts, and -c to output the returned cookie to a file.

Then you pass in the session-containing cookie file, allowing you to stay authenticated on subsequent requests:

curl -b cookie.txt -o complete-export.txt "http://www.typepad.com/t/app/weblog/post?__mode=export&blog_id=YOUR_BLOG_ID"

Naturally, you'll want to replace the URL in that last request with the one corresponding to your own export URL.

It all worked perfectly — well, up until the point at which I discovered that Typepad exports seem to inevitably fail once they top 100 megs. Seriously guys: file compression. Try it, you'll like it.

Hi I'm Ethan, the new guy. This is my first post, but beyond that brief bio and the unequivocal denial of any claims that I don't change my shirt on a daily basis, this post isn't going to be about me.

It's going to be about the totally off the hook tricks I leaned about the "previous command" shell shortcuts I learned today from this post and how they totally rocked my world a second ago.

So here's the situation: I downloaded a file and couldn't remember whether it was a zip or a tarball, so I ran this command:

   tar -xzf ~/downloads/google-sitemap-generator.3.0b7.zip
 

Why did I try to unzip a tarball, you ask? Because I'm lazy, and when autocomplete says "press enter", I ask "how fast"? Needless to say, the result was some mumbo-jumbo that I'm sure must have had ample amounts of bit-level impatience buried somewhere underneath it.

Remembering the tricks from that post I tried this:

   unzip !^
 

I was thinking that maybe the first argument (referenced by !^) would be recognized as the argument after the flags. But no go, it resolved to unzip -xzf

Now I had a bit of a dilemma: I couldn't use any more of the cool tricks because they all referred to the command prior to the last...unless. What if all the argument referencing commands could be prefixed with a pattern!?! Sure enough, this gem worked like a charm...gem...:

   unzip !tar:2
 

Which basically reads unzip {the second argument of the last command matching the glob-string 'tar'}

To my total, complete astonishment it actually worked. Brilliant.