Kevin's random thoughts - tech

Rails assumptions

2008-11-13T20:02:00Z

Second in what's now a series of "list" posts. First here.

Standard assumptions in the Rails community:

Working in a small start-up consultancy
Mac laptop for development
TextMate for editing
git for version control
MySQL for database backend
Deployment on a hosted VPS
"Application" means "Web application"
Any Web site you could want to develop is a Web application
Java background

As before, add your own in the comments.

2008 Web design distinctives

2008-11-12T17:09:00Z

First in what's now a series of "list" posts. Next here.

Gradients
Georgia for headings
Large-surface-area navigation tabs
Design on a grid
Rich but totally random stock photos
Subtly patterned backgrounds
Everything in ems

Add others in comments. Preferrably with links to examples.

Subversion to Mercurial

2008-11-03T17:04:00Z

I've been working on migrating at least some of my work projects from svn to hg. Most of my projects are relatively simple and not very branchy, so converting them was a matter of doing:

$ hg convert svn://my-repo.example.com/svn/project/trunk project

Some of my bigger projects, though, made the convert extension choke on my many branches. For these I resorted to hgsvn, which converts one branch at a time. For the benefit of others, here's documentation of my process. [Ed. note: now cleaned up a bit. Let me know if it's unclear.]

First, create a clean directory to hold all of the converted trees. Then use hgsvn to convert the trunk:

~ $ mkdir project && cd project
~/project $ hgimportsvn svn://my-repo.example.com/svn/project/trunk
...
~/project $ (cd trunk && hgpullsvn)
...

Now do the same to convert each branch:

~/project $ mkdir branch-1 && cd branch-1
~/project $ hgimportsvn svn://my-repo.example.com/svn/project/branches/branch-1
...
~/project $ (cd branch-1 && hgpullsvn)
...

Once you've got all the branches converted, you can merge them into one big named-branch repository (if you roll that way). Since hgsvn preserves the ancestry of your branches from where they split from trunk, it's a sensible operation to merge them all together. This is optional though; if you want to keep each branch as a separate directory, you can.

~/project $ hg clone trunk whole-project
~/project $ cd whole-project
~/project/whole-project $ hg pull ../branch-1
...

Repeat the pull for each branch. Note that the logs might appear a bit out-of-order afterward compared to the svn log. The logs for each branch will be brought in as a chunk, not interleaved with other branches' logs in commit-date order.

Too much cleverness

2008-09-09T18:01:00Z

DataMapper is, in general, pretty cool, but like most Ruby developers lately, its developers seem to have seen all the "syntactic" fun you can have with the language and just taken it too far ("syntactic" in quotes because Ruby's syntax is actually fixed). Here's how DataMapper expects conditions to queries:

exhibitions = Exhibition.all(:run_time.gt => 2, :run_time.lt => 5)
# => SQL conditions: 'run_time > 1 AND run_time < 5'

And it does the same thing for ORDER BY clauses:

@zoos_by_tiger_count = Zoo.all(:order => [:tiger_count.desc])
# in SQL =>  select * from zoos ORDER BY tiger_count DESC

They actually defined methods on the Symbol class called lt, gt, like, not, in, etc. This is monkey patching almost at its worst. What use, in any Ruby code outside of an argument hash to DataMapper model, is being able to say :foobar.like(:a_duck) ?

One example of good monkey patching was (I think) introduced by Rails: the method Object#blank?, defined (basically) thus:

class Object
  def blank?
    nil? or (respond_to?(:empty?) and empty?)
  end
end

So it returns true on '' (an empty String), [] (an empty Array), and nil, and false on just about anything else (including objects that don't have an empty? method). (The Rails implementation might also return true if you call false.empty?, as a special case. This might make sense; I haven't thought the semantics through fully.) Object#blank? is a well-behaved monkey because it doesn't modify existing, expected behavior, and it's globally useful (throughout the Ruby language). It's something that could be added to Ruby's core without breaking things.

DataMapper's Symbol extensions probably wouldn't break other libraries, but the semantics are senseless. If a Symbol had a meaningful 'less-than' comparison to another symbol, it would just implement <=> and mix in Comparable. What the line

exhibitions = Exhibition.all(:run_time.gt => 2, :run_time.lt => 5)

really means is that the column called 'run_time', which isn't a Ruby quantity at all, should be between 2 and 5. A better way to say it would be:

exhibitions = Exhibition.all(:run_time => [:>, 2], :run_time => [:<, 5])

or just

exhibitions = Exhibition.all("run_time BETWEEN 2 AND 5")

which is basically how ActiveRecord does it. Using a global monkey patch to Symbol to implement a syntactic translation of Ruby to SQL just muddies the semantics of Ruby without providing any extra clarity.

It's a back door (through method definitions on symbols) into the kind of real syntax translations you can write as macros in Lisp (which don't rely on the function invocation mechanism). You could define a macro conditions to be used as an argument to the finder method, which could then be invoked like this:

(find-all-exhibitions (conditions (run-time > 2)
                                  (run-time < 5)))

The difference is that the syntax you define only exists within the conditions macro.

Doing the math, cont'd

2008-08-21T19:36:00Z

Not sure if John Gruber noted this jackassery or not. Writing on ars technica, Joel Hruska manages to stop just short of a Jackass-of-the-Week award, but nonetheless veers deep into irrational anti-Artie-MacStrawman-ism.

Hruska's analysis of the fact that Apple's average selling prices are higher than the rest of the market starts off well enough:

Desktop PC average selling prices (ASPs) have scarcely changed in the past two years, while laptop ASPs have fallen 21 percent, probably in part thanks to rising demand and steeper competition among vendors. Mac desktop ASPs, on the other hand, have actually gone up by 7.7 percent, while laptop ASPs have dropped by just 3.8 percent. Apple, in other words, is bucking conventional wisdom, and selling computers that are significantly more expensive than the average unit shipped by the likes of Dell, HP, or Lenovo.

...and turns back from the brink for a moment of insight...

This would normally be the point at which any Mac vs. PC discussion veers off into a heated debate over hardware specs, manufacturer build quality, and acerbic commentary on the more dubious characters in one's family history. I'm rather explicitly not going there. Instead, I'd like to draw attention to a different facet of the situation. ... Apple is clearly picking up steam without significantly reducing its ASPs. It's even happening while PC ASPs, particularly notebook ASPs, are dropping.

The current situation implies that Apple may be drawing high-end notebook sales that would have otherwise gone to Dell, HP, Lenovo, or the other major players in this space.

...before veering off into warrantless-conclusion-land:

Apple has clearly created a value proposition that customers are willing to pay a substantial premium in order to possess. [emphasis added]

I think he's right that Apple is drawing higher-end (not just top-end, but mid-range too) notebook sales away from other makers, which makes his conclusion all the more senseless. As has been demonstrated already, comparably equipped Macs sell in the same price range as their Windows-preinstalled counterparts. If it's true that Apple's sales are coming from customers who would have otherwise bought $1000+ Dell or hp laptops, those buying Macs are not paying a premium.

No, really, everything is an expression

2008-08-19T22:26:00Z

I just had a realization about Ruby that others coming from C, Perl, Javascript, PHP, etc. may not have quite grasped yet: In Ruby, since everything is an expression, semicolons are allowed within parentheses. That means parentheses effectively create a statement block (though not a new scope, and not a block in the Ruby sense of a first-class procedure object).

>> ("foo"; 42; :coolness)
=> :coolness

The above evaluates the values "foo" (a String), 42 (a Fixnum), and :final (a Symbol) in that order, and returns the last one. In Lisp, this would be a progn (Common Lisp) or begin (Scheme).

I've made use of this behavior before, I think, without really realizing the implications of it. In Perl, for example, you can use a bare block to group statements together. But you can't even apply a statement modifier to such a grouping. The following is a syntax error:

{             
  print "foo\n";
  exit(254);
} unless open(FOO, "foo.txt";

...leading to such idioms as:

print "foo\n" && exit(254) unless open(FOO, "foo.txt");

In Ruby, since everything is an expression, you can use statement modifiers on a parenthesized expression, like so:

>> (f = open("hello.c"); f.read) if File.exist? "hello.c"
=> # contents of hello.c

You could even use a compound expression as a conditional test (in a statement modifier!):

>> "non-empty file" if (f = open("hello.c"); !f.eof?)
=> "non-empty file"

Doing the math

2008-08-18T18:13:00Z

Someone out there did an extensive price comparison between the MacBook and comparably-equipped laptops from Dell, hp, and Sony (h/t daring fireball). I thought I'd post here a similar comparison I did between comparably-equipped desktop machines from Apple and Dell back in 2002 (originally posted to a slashdot comment). These numbers were drawn directly from the respective companies' online stores at the time.

Dell Dimension 8200

Pentium 4 1.7GHz
256MB RAM
40GB HD (no 60GB available)
3yr ltd warranty/service plan
Windows XP Home
15in flat panel LCD (analog connector)
64MB Geforce2 MX
DVD-R/CD-RW drive
Dell Jukebox, Image Expert 2000, Standard Dell Movie Studio bundle
Microsoft Works 2002
Price: $2,209

Apple iMac SuperDrive

G4 800MHz
256MB RAM
60GB HD (no 40GB available)
3yr AppleCare Protection Plan [apple.com]
Mac OS X and OS 9
15in flat panel LCD (digitally connected)
32MB Geforce2 MX
DVD-R/CD-RW SuperDrive
iTunes, iPhoto, iMovie, iDVD
AppleWorks 6.2 Price: $1,948

Difference: $261

Apache's trailing slash and proxying to RESTful apps

2008-04-09T19:52:00Z

I have a somewhat odd case with the Web app I just deployed today (no, not Rails, it's Merb!). I have a RESTful Documents controller which needs to receive POST requests on /documents. I also have a documents directory in the public document root. The whole thing is hosted behind Apache with mod_proxy_balancer, and mod_rewrite to let Apache serve static files (all as described on the Mongrel site).

So when I would POST to /documents, Apache in its (normally quite useful) wisdom decided to rewrite it to /documents/ and (worse) to change the request to a GET. This is because of mod_dir, which provides the DirectoryIndex directive and the (less well-known) DirectorySlash directive, which defaults to On. So to fix it, I just set DirectorySlash to Off for the documents directory, and bingo, it works.

Now, if I needed directory indexes on the subdirectories of the public/documents/ directory, I'd have to work around, but this is fine for my current setup.

textutil: OS X command-line tool of the week

2008-01-15T21:23:00Z

$ textutil -convert rtf 'Business Plan.doc'
$ ls
Business Plan.doc  Business Plan.rtf

omg d00d! Command-line conversion between proprietary and open-source formats!

$ textutil -convert doc 'Live Free Or Die.odt'
$ ls
Live Free Or Die.doc  Live Free Or Die.odt
$ file Live\ Free\ Or\ Die.doc
Live Free Or Die.doc: Microsoft Office Document

man textutil for more details. Other supported formats (besides the demonstrated RTF, MS Word, and OpenOffice) include plain text and HTML. It can also display document metadata (including of Word documents) and convert documents from one text encoding (UTF-8, ASCII, KOI-8, ...) to another. Whee!

I hated DRM before it was cool

2007-11-21T20:55:00Z

From some random ZDNet-employed tech pundit:

Apple goes to great lengths — usually through its digital rights management technologies (what I call C.R.A.P.) — to tightly control the relationship between its software, its hardware, and, in the case of the iPhone, the relationship of both to carriers and the Internet (God forbid you should attempt to acquire new audio online for your iPhone — music, ringtones, etc — through anything but the iTunes Music Store).

Just the latest in a long string of Apple hatred (and now litigation) based on their "FairPlay" DRM system. Apple can't be exonerated of their complicity in the foisting of DRM on us consumers, but let's review the facts.

Apple does not force you to use DRM-crippled music either in iTunes or on the iPod. In fact both products were capable of using the de facto industry standard (though unfortunately patent-encumbered) MP3 format before FairPlay existed in the market—i.e., before the opening of the iTunes Music Store.
Apple themselves sell non-DRM-crippled tracks through their own music store.
Apple is neither the only, nor the worst company to use DRM.

I happily use an iPod and iTunes (and a Mac). I have never paid money for a single DRM-encumbered track from the iTunes Store (though I do own a few such tracks via gift cards and Pepsi caps). I've been an eMusic subscriber since their plans included unlimited downloads. I find it unconscionable that the RIAA could sue a family into oblivion for simply depriving them of potential (that means neither real nor certain) revenue to which they (and the courts) think they have a right.

But why is it suddenly cool to hate Apple for using DRM (based on utter fabrications), and not actually support ditching it entirely?

And if you hated Bush before it was cool, see here.

Mac command line niceties

2007-11-09T22:42:00Z

I've made some additions to my .bashrc today based on this article. First, the usage:

Then I press any key (Return, Esc, Space, whatever), and...

I can also do this:

That is, I can open a finder window in the current directory just by typing open—no need to explicitly specify '.' as the argument.

Here's the code:

function ql () {
  (qlmanage -p "$@" > /dev/null 2>&1 &
    local qlmanage_pid=$!
    read -sn 1
    kill ${qlmanage_pid}) > /dev/null 2>&1
}

function open () {
  if [ -z "$*" ]; then
    /usr/bin/open .
  else
    /usr/bin/open "$@"
  fi
}

Geeks, nerds, and dorks

2007-10-30T20:56:00Z

If you've ever wondered what the difference is between a geek, a nerd, and a dork, or if you never realized there was one, here it is:

The Mac is made for dorks.
Linux is made for geeks.
Windows is made for nerds.

Further notes on Apple's recent releases

2007-09-12T15:19:00Z

On iTunes 7.4:

Having to pay twice for a song just to excerpt it for use as a ringtone is wrong. I'll keep using Audacity (and AudioHijack where necessary) to excerpt songs to put on my phone as MP3s.

I don't own an iPhone, so this won't affect me anyway, but it's perfectly within the bounds of fair use to excerpt a song for use as a ringtone. It's hard to tell if Apple was forced to do this by the music industry, or if they're just hedging their bets, but in either case, I'm disappointed.

~~The open question, though, is whether it will be free to make a ringtone from your non-DRM-crippled (ripped from CD to MP3 or other formats) music.~~ It looks like it's not only not free, but actually impossible to create ringtones from your non-DRM-encumbered music.

On the new third-generation nano:

The screen pixel density (204ppi) is even higher than that of the iPhone. I expected it to be the same, but they've leapfrogged it.

We might see iPhones with the higher pixel density soon, but it will depend on how well the touch capability can be integrated with it. OS X (as shipped on the iPhone, and with the upcoming release of Leopard on Macs) is supposedly resolution-independent, so that makes it easier to change software to understand a different pixel density.
The game Steve demo'd is essentially 3D Breakout—a nod to his early days in the computing market.

Graphic resizing

2007-09-12T15:17:00Z

This is neato! Content-aware graphic resizing (up and down) by finding "seams" within the image.

WANT

2007-09-06T15:11:00Z

So I noticed Software Update pop up this morning with an iTunes update:

With iTunes 7.4, sync your favorite music and more with the new iPod nano (third generation), iPod classic, and iPod touch…

Bwuh? Apple released the much-rumored iPhone-like iPod?

http://www.apple.com/ipodtouch/

YAY! WANT! IS CAN TOUCH IPOD NOW PLZTHX