Compile you? I don't even know you!

XAMPP on OSX, WordPress and Permalinks

2012-02-05T14:04:00.000-03:00

Finally I got Permalinks in Wordpress to work on my Wordpress install on local. Whew! Thanks to this post.

For me the answer was:

Make sure mod_rewrite is installed. Using <?php phpinfo(); ?> is a good way to find out, but you probably just need to comment it out in one of the Apache config files.

Then for your directory of interest you have to use something like this:

Options Indexes FollowSymLinks ExecCGI Includes

AllowOverride All

Order allow,deny

Allow from all

</Directory>

MSB and LSB with MIDI (NRPN)

2012-02-02T17:34:00.001-03:00

If you know nothing about programming yet find yourself programming for a living (as I do), you might have to deal with some of this bit-shifting bullshit that you should've learned in University. I was drinking coffee and reading Foucault, so I did not learn any of this stuff. My skills as a PUA are probably better, however. But I digress...

I'm about to implement NRPN in MIDI Designer, which led me to this wonderful article. It's pretty easy. Kind of.

Normal MIDI Messages

In the normal case (not NRPNs), MIDI messages typically have a channel, command, data1 (the cc number or the "note") and data2 (the value). The first two are combined into status. So the whole thing looks like this:

int status = ((command & 0xF0) | (channel & 0x0F));

Byte theBytes[3] = {status & 0xFF, data1 & 0xFF, data2 & 0xFF};

NRPNs - How to Get MSB and LSB?

NRPNs are like this, but the data2 is being used in two different 3-length byte arrays to get more values in there. So instead of going from 0-127, they go from 0-16129 (127 squared).

Now this creates a problem where you need to get the MSB and the LSB. After talking to my father on the phone today, I realized that what's so messed up about this is that they're not 8-bit bytes. They're "7-bit bytes" (they go from 0-127). So anyway, you have a value and you want to get the MSB and LSB.

static int msb(int value) {

return (value) >> 7;

}

static int lsb(int value) {

return value & 0x7f;

}

static int makeValue(int msb, int lsb) {

return ((msb << 7)) + lsb;

}

Now, you could do this the normal way, too, like this:

int msb = (float)value / 128.0f;

int lsb = value - msb * 128.0f;

which also works, but people will laugh at you.

Conclusion

By now most of you still have no idea what I'm talking about, and this article hasn't helped you. But if you were about to implement NRPNs in a MIDI program, this article might help.

Perfect Storm for Software Development

2012-01-24T17:39:00.000-03:00

Convo with Friend of Mine, Head of Development at Medium-Sized Company

Lots of big companies with low quality iOS apps

which strikes me as odd

CTO

not surprising at all

development doesn't scale well

in fact scales horribly

not add on top of that mixed directions and signals from project and product managers

introverted opinionated team

company politics

etc.. and you have a perfect storm

Rename Screenshots OSX

2012-01-03T11:46:00.001-03:00

I'll have to figure out how to get this into an automatic folder action somehow, but for now this works:

#!/usr/bin/env ruby
Dir.chdir("Desktop") # works in 1.8.7, later use Dir.home

prefix = "Screen shot-"

dir = Dir.glob("#{prefix}*.*")

dir.each do |file|

time_date = File.new(file).mtime.strftime("%Y-%m-%d-%H%M-%S")

new_name = "#{prefix}#{time_date}"

if (file.split(".").count > 1)

extension = file.split(".").last

new_name = "#{new_name}.#{extension}"

end

`mv '#{file}' '#{new_name}'`

end

puts "Renamed #{dir.count} files with prefix #{prefix}"

Multiple Build Targets, Same Name

2011-12-19T17:14:00.001-03:00

Have you ever wanted to have multiple build targets like Angry Birds? They use the different icons and the same name, "Angry Birds," below each of them (mostly). The icons says stuff like, "Free."

Anyway, here we go... Xcode 4.x, an already-working target:

Duplicate your target
Now you will have a target with a funny new name and a new plist. So
Rename the target
Rename the plist
Adjust the target settings in the Project to point to the renamed plist. If you don't do this, Xcode will pretend it's fine, but it's not.
Adjust target so that it's got a unique Product Name. If you use the same one more than once, you will rot in hell forever.
Now go to the plist that your target is pointing to and adjust its CFBundleIdentifier to be something unique.
In that same plist, adjust your CFBundleDisplayName to be the display name you want, e.g., "Angry Birds"
Xcode will also have created a new "Scheme" for this new target. You can rename the scheme and it will still point to the target, so all is good.
Bring a bunch of Icon.png files in, and associate each one with the correct target.

Now I have to figure out if this is a good idea. However, don't do this just to get some different arguments in. For that, you can just use schemes, as I mention here: http://compileyouidontevenknowyou.blogspot.com/2011/09

Keywords

Xcode, iOS, iPad, Objective-C, Multiple Build Targets

Adjust Hue and Saturation of a UIImage

2011-12-03T02:52:00.000-03:00

How to adjust Hue and Saturation of a UIImage in Objective-C. This is a method for a UIImageView subclass, but you could easily remix in another context.

- (void) changeToHue:(float)hue saturation:(float)saturation {

UIGraphicsBeginImageContext(self.bounds.size);

CGContextRef context = UIGraphicsGetCurrentContext();

UIView *hueBlend = [[UIView alloc] initWithFrame:self.bounds];

hueBlend.backgroundColor = [UIColor colorWithHue:hue saturation:saturation brightness:1 alpha:1];

CGContextDrawImage(context, self.bounds, self.image.CGImage);

CGContextSetBlendMode(context, kCGBlendModeHue);

[hueBlend.layer renderInContext:context];

self.image = UIGraphicsGetImageFromCurrentImageContext();

UIGraphicsEndImageContext();

}

From my answer here.

Running Two Notational Velocities at Once

2011-10-24T18:16:00.000-03:00

Do this:

Copy your Notational Velocity application (so it's called "Notational Velocity copy" or whatever)
Adjust the key in the info.plist of the package so that the CFBundleIdentifier points someplace different (e.g., net.notational.velocity2)
Launch the NV copy and adjust the notes directory to also be something different

That's it! Now you've got two Notational Velocities running side by side. You can even give them different hot keys and different fonts so you can tell them apart.

I think the same thing would work with nvAlt, which by the way will sync your tags to simplenote, which is a big deal.

Check out the mini-discussion with heavy-weight Superuser @Lri and a

cleaner solution using defaults write and ditto.

Ensure in Main Thread

2011-10-19T20:13:00.001-03:00

Multithreaded programming is a must in iOS. If you schedule an NSTimer and yet it doesn't fire, it might be because the thread that launched it is now dead. If it's related to UI, you'll need to get to the main thread eventually, anyway. Use this

#define ensureInMainThread(); if ([NSThread currentThread] != [NSThread mainThread]) { [self performSelectorOnMainThread:_cmd withObject:nil waitUntilDone:NO]; return; }

It can only be used from methods that accept (and return) no parameters.

Now you can have pretty code like this

- (void) msgReceived {

ensureInMainThread();

Beware that using code like this might indicate that you are completely lost in a sea of threads. If you have no idea why you need this method, you should go back to your code and figure it out.

Localize Your NSString

2011-10-12T16:57:00.001-03:00

I probably won't use this code since I'm pretty sure it's a terrible idea. But since I put it together today, I'll share it. It's an NSString category that overwrite just the stringWithFormat: method, and just replaces the text with the text from your local language bundle.

I thought it would be useful for localizing an application, but it will probably result in madness and, eventually, blindness. It's a category "extension" for NSString.

#import "NSString+Localization.h"

@implementation NSString (Localization)

+ (NSString*)stringWithFormat:(NSString *)text, ...

{

va_list args;

va_start(args, text);

text = NSLocalizedString(text, nil);

NSString *retVal = [[NSString alloc] initWithFormat:text arguments:args];

return retVal;

}

Perhaps more sane than overriding the stringWithFormat: method would be to create one. Any help on the name would be appreciated:

#import "NSString+Localization.h"

@implementation NSString (Localization)

+ (NSString*)twiceLocalizedStringWithFormat:(NSString *)text, ...

{

va_list args;

va_start(args, text);

text = NSLocalizedString(text, nil);

NSString *retVal = [[NSString alloc] initWithFormat:text locale: [[NSUserDefaults standardUserDefaults] dictionaryRepresentation] arguments:args];

return retVal;

}

NOTE: This code is ARC friendly. If you are not ARCing, autorelease the retVal in both examples.

UIViewAutoresizing Has Limits

2011-09-18T22:44:00.001-03:00

If you're going to try to resize stuff seriously -- all the way down, then all the way up -- UIViewAutoresizing sucks. Why? All the information it uses comes from the current state. Over resizes to smaller sizes it will lose precision. This means that a "flexible sides" subview that was 180 pixels from the left side could end up at 18 pixels from the left side. So you'll need to store your subviews' sizes. You get something cheesy like this:

#pragma mark All the storing position stuff since UIViewAutoSize sucks

-(NSMutableDictionary *)positions {

if (!positions)

self.positions = [NSMutableDictionary dictionary];

return positions;

}

- (void) repositionElements {

for (NSValue *key in [self.positions allKeys]) {

[self reposition:[key nonretainedObjectValue]];

}

- (void) storePositionsForElements {

for (UIView *view in self.subviews) {

if (CGRectIsEmpty(view.frame))

return;

[self storeFirstPosition:view];

}

-(void)storeFirstPosition:(UIView *)element {

NSValue *key = [NSValue valueWithNonretainedObject:element];

[self.positions setObject:[NSValue valueWithCGRect:element.frame] forKey:key];

}

- (void)reposition:(UIView *)element {

NSValue *key = [NSValue valueWithNonretainedObject:element];

CGRect firstFrame = ((NSValue*)[self.positions objectForKey:key]).CGRectValue;

float newX = firstFrame.origin.x / firstBounds.size.width * self.bounds.size.width;

float newY = firstFrame.origin.y / firstBounds.size.height * self.bounds.size.height;

float newWidth = firstFrame.size.width / firstBounds.size.width * self.bounds.size.width;

float newHeight = firstFrame.size.height / firstBounds.size.height * self.bounds.size.height;

element.frame = CGRectMake(newX, newY, newWidth, newHeight);

}

This would allow you to store everything in an NSMutableArray called positions (make sure it's not nil, for God's sake). You would call storePositionsForElements in your initWithCoder (so it stores the NIB's positions), and then call repositionElements from your setFrame override.

If categories let you store stuff, I would monkey patch UIView and get a much better solution.

Lock Aspect Ratio to Max

2011-09-16T16:59:00.000-03:00

This is pretty much exactly how UIImageView works when you use "Aspect Fill." If you use this in the enclosing view's setFrame, you can let the UIImageView freewheel (scale to fit), and resize other things according to your new size:

-(void)setFrame:(CGRect)frameIn {

if (!ignoreAspectRatio) {

float aspectRatio = self.frame.size.width / self.frame.size.height;

frameIn.size.width = MIN(frameIn.size.width, frameIn.size.height * aspectRatio);

frameIn.size.height = frameIn.size.width/aspectRatio;

}

[super setFrame:frameIn];

}

This kind of programming is not my strong suit, but with a few minutes goofing around in Excel, it was easy to figure out. And it works.

Parameters into iOS Apps, XCode 4.x

2011-09-05T13:25:00.000-03:00

I needed to get arguments and version numbers and so forth into an iOS app.
The arguments come in via the Scheme. You can have one scheme per target, or one scheme per multiple targets. I'm using arguments in the format paramOne=35.5 paramTwo=release

Then you can pick up the arguments like this:

Version numbers, differently, come in via the target. You get a "version" number and a "build" number.

Then you can pick up the build and version with something like this

Yet another way, from my colleague Jose at Thingee:

Regular Expressions for Those Nasty Autorelease Statements in ARC Migration

2011-08-28T14:33:00.000-03:00

Maybe I'm missing something and the "migrate to ARC" tool in the XCode 4.2 Beta replaces your autorelease statements, but that's not what I've found. Please correct me if I'm wrong.

Here's a simple regex replace in Textmate for changing lines like:

[[[Thing alloc] initWithWhatever:whoever] autorelease];

[[Thing alloc] initWithWhatever:whoever];

The regex is

(\[\[\[)([^\]]*\][^\]]*\])(\ autorelease\]);

and the replacement string is:

[[$2;

Then I needed to fix these lines

for (Thinger *thing in [thingers copy] autorelease])

and I used regex:

(\[)([^\]]*\][^\]]*)(\ autorelease\])

and the replacement string is:

Backing up an MS SQL 2008 Database with Metadata Intact

2011-05-17T16:18:00.005-03:00

In a strange time machine recently which has taken me back to ASP (which still feels higher GL than reference counting in Objective-C, but whatever), I'm dealing with a shopping cart vendor on that platform. I explain that I have the entire system (website) copied over, but two functions aren't working. When I describe the symptoms, he writes, "Based on your comments I am about 99% sure the issue was the database transfer. The primary keys were probably dropped." So he writes, check the docs, which say, essentially, get a DBA. Ironically, they were the same docs I was following when I decided not to call a DBA. Primary keys!? Oh yeah, right.

So how do you transfer a Microsoft SQL 2008 DB with Enterprise Manager and the kind of permissions they give you on a shared cheapo server somewhere like Winhost?

There are many false paths. First, do not do what I did, which drops the primary keys and any relevant meta-information about the DB:

Do not try to backup or mirror, as both of these are dead ends due to permissions issues:

You have to script it. But do NOT use this!

Instead you have to use:

It's mostly intuitive, but use these options:

And then you have a query which you can execute as if you had typed it by hand.

Update: You should set "If not exists" to Yes and "indexes" to "Yes" too.

Rails 2.3x and Gem Hell

2011-05-05T20:10:00.000-03:00

Today I hit up against two of my favorite problems in Rails 2.3x, or maybe both at once. This is a play by play with no additional info. It's mostly here for me, but if someone else finds it useful, great. The reason that I need to document most of this is that my brilliant coworker, Sean, fed me a lot of the lines here.

First error: the Rails install was 2.3.8 and on deploy to the server the message was:

uninitialized constant ActiveSupport::Dependencies::Mutex

Which led me to this post on SO. So my choices were two: either upgrade the Rails installation or downgrade my rubygems system.

I opted for upgrading Rails, of course.

sudo gem install rails --version=2.3.11

I updated the environment.rb, deleted the vendorized version of Rails and tried to vendorize the new one:

rake rails:freeze:gems

Could not vendor with problem:

undefined method `version_requirements' for #<Gem::Dependency:0x10055b0d0>

To fix this, I needed to follow the instructions here, which were basically

gem install rubygems-update -v='1.4.2'

gem uninstall rubygems-update -v='1.5.0'

update_rubygems

However, at this point I had already screwed up my RubyGems by doing a

sudo gem update --system

and I ended up at 1.8.0 with a whole bunch of deprecation warnings and intermediate versions of RubyGems. So first I had to uninstall all the unwanted versions. So I listed them:

gem list | grep rubygems

and then proceded to run gem uninstall rubygems-update -v='1.x.x' on each except the last one. Then I followed the three lines above.

This allowed me to vendorize the right Rails version and I was cooking, just 35 minutes later. Well, really 70 person-minutes later, at least. Thanks Sean!

Wimpy OSX! Nano for everything!

2011-03-31T17:01:00.000-03:00

So one of the things I never asked was how to use nano as the default editor on OSX. I'd always run piecemeal commands like this

git config --global core.editor "nano"

but it makes a lot more sense, in most cases, to use

export EDITOR=nano

unless you want to use vi for some things and nano for others.

Autocomplete with BBEdit

2011-02-23T09:33:00.001-03:00

This is my nightmare of autocomplete: Visual Studio was 100x better than this 10 years ago. Autocomplete in BBEdit on OSX fails utterly to pick up relevant cues from context.

More Boilerplate Code for Obj-C Done In Ruby

2011-02-15T20:11:00.001-03:00

# excuse this terrible code, but sometimes you want a solution now

# put your text in between the TEXT markers

text = <<-TEXT

@property (nonatomic, retain) IBOutlet UIWindow *window;

@property (nonatomic, retain) IBOutlet UIViewController *viewController;

@property (nonatomic, retain) IBOutlet CentralNavController *navController;

@property (nonatomic, retain) IBOutlet SplashController *splash;

TEXT

puts "// Generated by #{$0.split('/').last}"

text.lines.each do |line|

variable_name = line.split("*")[1].chomp(";\n");

puts "self.#{variable_name} = nil;"

end

Simplest Example: Hierarchical Plist in iOS and Rails

2011-02-02T17:32:00.001-03:00

These examples are based directly on this article, Reading a Plist into an NSArray.

The Plist

<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">

<dict>

<key>Region 1</key>

<array>

<string>Billy Ray</string>

</array>

<key>Region 2</key>

<array>

<string>Frankie Sanders</string>

</array>

</dict>

</plist>

and the code

NSString *path = [[NSBundle mainBundle] pathForResource:@"regions" ofType:@"plist"];

NSDictionary *regions = [[[NSDictionary alloc] initWithContentsOfFile:path] autorelease];

for (NSString *region in regions) {

NSArray *array = [regions objectForKey:region];

for (NSString *representative in array) {

NSLog(@"rep=%@, region=%@", representative, region);

}

If you're dealing with a plist, the file is read in one giant gulp, so if you have memory problems, you'll have to use CoreData, and I don't think XML can be the backing store.

So what's the simplest code to read this same plist in Rails?

require 'test_helper'
class ReadPlistTest < ActiveSupport::TestCase
  test "recover names of regions and names of reps" do
    require "rexml/document"
    file = File.new("../data/regions.plist") # we're run from test, otherwise it'd be /data/regions.plist
    doc  = REXML::Document.new file
    region_names = []

    doc.elements.each("plist/dict/key") { |element| region_names << element.text }

    regions = {}
    i = 0    # can't use each_with_index
    doc.elements.each("plist/dict/array") do |element|
      region = []
      regions[region_names[i]] = region
      element.each_element("string") { |thing| region << thing.text }
      i+=1
    end

    puts regions.count.to_s
    puts y(regions)

  end
end

DRY Way to Read CSV File Into ActiveRecord Model

2011-02-02T14:47:00.003-03:00

I've been looking for a DRY way to do this for a while.

require "csv"

file_path = "data/lunar-lander-1520turbo.csv"

# define columns, symbols MUST be column names in module

names = {}

names[:lander_name] = names.length

names[:wing] = names.length

names[:wheel] = names.length

names[:wheel_base] = 95

start = true

CSV.open(file_path, 'r') do |row|

if start # skip first row

start = false

else

puts "Creating #{row[names[:lander_name]]}"

m = MobileDevice.new

names.each_key { |name| m[name] = row[names[name]] }

m.save

end

Me on IDEs, Again

2011-01-30T15:00:00.000-03:00

Me on IDEs yet again.

What Does Your Obj-C Code Look Like?

2011-01-17T15:02:00.003-03:00

With a little help from Ruby, mine looks like this

example_model = ExhibitContributionSponsorship.new
field_names = example_model.attribute_names
field_names.collect! do |field_name|
  splitOut = field_name.split("_");
  splitOut.collect! { |piece| piece==splitOut.first ? piece : piece.capitalize }
  splitOut.join
end

field_names.each { |field| puts "@property (nonatomic retain) NSString *#{field};" }
synthesize = "@synthesize "
synthesize << field_names.join(", ")
puts synthesize

field_names.each { |field| puts "@property (nonatomic retain) IBOutlet UITextField *#{field}Control;" }
synthesize = "@synthesize "
synthesize << field_names.collect.join(", ")
puts synthesize

I used to use Excel, but Ruby is cooler.

Why ObjectiveResource Sucks

2011-01-11T11:40:00.001-03:00

I could be wrong about this, so feel free to school me. Until then...

I've been screwing around with ObjectiveResource for a bit and comparing it as an option to ASIHTTP until I realized that even I can tell how bad ObjectiveResource is, so it must really suck. Why?

Overall architectural choice: Categories Only - these guys are Rubyists and they code like it. Ruby is an interpreted (read "slow") language which is only run in environments where it's fast enough (and with modern computers, those are many). Obj-C (especially on iPhone) ain't Ruby, so the choice to use categories instead of inheritance for your objects, while cute, costs the library dearly. Categories cannot, by definition, hold state, so you have to build the entire state every time you need something. This results in very inefficient code. As an example, put an NSLog in any of the methods of the NSObject+PropertySupport and watch 'er spin for even the simplest Web Services call. "No premature optimization" is not a get-out-of-jail pass.
Overall architectural choice: Focus On Crud - I actually enjoy a good Scaffold now and again, having built major admin portions of sites that rely on map.resources (now in Rails 3, just resources), but you'll need more than create-read-update-delete if your app is actually going to be more than a few milliseconds away from your server. Then you'd have to think about how to aggregate, and how to call other methods. There is a method called: - (BOOL)createRemoteWithParameters:(NSDictionary *)parameters andResponse:(NSError **)aError; I wonder what that does. Which leads to:
No Documentation At All - I mean the kind where you look at the .h and it tells you what methods actually do. But I guess if you're a Rubyist you consider all code to be self-documenting. Isn't true in Ruby, and it isn't true here. There is a five-minute walkthrough and a jumpstart.
No Concept of Async - While I would love to run my stuff in performInBackgroundThread, that's not really how serious business is done. Check out ASIHTTP to see how to actually do async correctly, and provide examples of it.
No Concept of Failure - that's right. If your Rails site is down, ObjectiveResource doesn't throw any Exceptions. In fact, it returns answers just like those you expect, only with incorrect values. It's like that employee you don't want to have, who just pretends that there are no problems while they make a small problem into a large one.

I could probably go on, but I have to finish a CRUD layer for ASIHTTP (or maybe just do some HelloWorlds and start running). I'm sure I'll be stealing lots of code from ObjectiveResource anyway.

Update: The thought of doubling my code on the client and server side was just too much to take, so I'm taking another look at ObjectiveResource. Coming back to a project after a pause often provides perspective. At least #5 is wrong. Here's the code to get it to work:

NSError *error = nil;

NSArray *posts = [Post findAllRemoteWithResponse:&error];

if (error)

NSLog(@"yes this is an error");

else {

NSLog(@"NO Error! Groovy");

}

ObjectiveResource Bug Fix

2011-01-11T11:07:00.003-03:00

Needing to do some Rails-iOS bridging, I've been checking out ObjectiveResource. At first I thought it was dated and also checked out the wonderfully documented and loved ASIHTTP request, which does indeed rock. But really, why parse when you can freeload?

The only problems I've found with ObjectiveResource is

How terribly the Google Groups forum works for solving problems. It's also notorious how few comments there are in the code.
It's pathetically synchronous, or perhaps there are just no docs for that.
If I bring down the Rails server it immediately returns wrong answers, which is just so disheartening I don't know why I didn't give up on it hours ago.
The following problem, which I spent over an hour fixing just now. To be honest, I'm not sure why the fix works exactly, but it's the right fix.

The problem is that ObjectiveResource blows up under iOS 4.x. I actually forget where the call is made to cause it to blow up, but the blowup begins by infinitely recursing and calling a method on NSObject+PropertySupport.h called propertyNamesAndTypes. The fascinating thing about the problem is that the line that causes the blowup is

[propertyNames setObject:[self getPropertyType:type] forKey:propName];

and it's actually the calling of setObject that causes the recursion to spin. I still have no idea why this happens, but here's the fix. Change this line

while (currentClass != nil) {

to this

while (currentClass != nil && [NSObject class]!= currentClass) {

and that's it. Why does this only happen if you set the object on the propertyNames mutable dictionary? I don't know. I'm not really here to solve problems, just make them go away.

THroNG's latest music video: 100% improvised in realtime like a mugging or a marriage proposal.

Default Init in iOS (Obj-C)

2011-01-06T12:58:00.002-03:00

The standard advice is to use a "designated initializer" which ends up with me having to think, which is always bad. Here is a pattern that never fails, and you could put it in a UIView category (UIView+SharedInit) and be done with this forever.

- (void) sharedInit {

}

- (id) init

{

self = [super init];

if (self != nil) {

[self sharedInit];

}

return self;

}

-(id) initWithCoder:(NSCoder *)aDecoder {

self = [super initWithCoder:aDecoder];

if (self) {

[self sharedInit];

}

return self;

}

- (id)initWithFrame:(CGRect)frame {

self = [super initWithFrame:frame];

if (self) {

[self sharedInit];

}

return self;

}

This is for a UIView subclass only, of course, but you could do the same thing for other classes as well.

Singleton in Interface Builder?

2011-01-05T20:13:00.001-03:00

Here's an answer to my question on StackOverflow: can you use a Singleton in Interface Builder?

Why would you want to do this? I know, Singleton is bad bad bad, but here it is.

THroNG's latest music video: 100% improvised in realtime like a mugging or a marriage proposal.

Change the Colors of a UIImage Using a UIColor

2010-12-13T20:26:00.002-03:00

Lots of googling led me here. Then it was pretty easy to mash this into that. When you combine this with the UIButton images provided by Apple, you can do cool stuff.

- (UIImage *) changeColorForImage:(UIImage *)image toColor:(UIColor*)color {

UIGraphicsBeginImageContext(image.size);

CGRect contextRect;

contextRect.origin.x = 0.0f;

contextRect.origin.y = 0.0f;

contextRect.size = [image size];

// Retrieve source image and begin image context

CGSize itemImageSize = [image size];

CGPoint itemImagePosition;

itemImagePosition.x = ceilf((contextRect.size.width - itemImageSize.width) / 2);

itemImagePosition.y = ceilf((contextRect.size.height - itemImageSize.height) );

UIGraphicsBeginImageContext(contextRect.size);

CGContextRef c = UIGraphicsGetCurrentContext();

// Setup shadow

// Setup transparency layer and clip to mask

CGContextBeginTransparencyLayer(c, NULL);

CGContextScaleCTM(c, 1.0, -1.0);

CGContextClipToMask(c, CGRectMake(itemImagePosition.x, -itemImagePosition.y, itemImageSize.width, -itemImageSize.height), [image CGImage]);

// Fill and end the transparency layer

const float* colors = CGColorGetComponents( color.CGColor );

CGContextSetRGBFillColor(c, colors[0], colors[1], colors[2], .75);

contextRect.size.height = -contextRect.size.height;

contextRect.size.height -= 15;

CGContextFillRect(c, contextRect);

CGContextEndTransparencyLayer(c);

UIImage *img = UIGraphicsGetImageFromCurrentImageContext();

UIGraphicsEndImageContext();

return img;

}

Now that I know this simple code, I can even create idiotic methods like:

- (UIColor*) oppositeColor:(UIColor*)color {

const float* colors = CGColorGetComponents(color.CGColor);

return [UIColor colorWithRed:1-colors[0] green:1-colors[1] blue:1-colors[2] alpha:1];

}

More Stupid Comments on Programming: "This Code Sucks!"

2010-12-08T23:59:00.001-03:00

Wordpress and jQuery: Where's the Dollar?

2010-12-08T13:22:00.001-03:00

The first thing I did in Wordpress was add jQuery, using the normal tag:

wp_enqueue_script( 'jquery' );

right above wp_head();

But then $ was not recognized? Why? Because Wordpress loads in noConflict mode. To put jQuery back for a particular page, use:

// set the _$ which jQuery uses for backup

window._$ = $;

// set the $ to jQuery

window.$ = jQuery;

Another option is to put this above the wp_enqueue_script, which will load jQuery in normal conflicty mode.

wp_deregister_script( 'jquery' );

wp_register_script( 'jquery', 'http://ajax.googleapis.com/ajax/libs/jquery/1.4.4/jquery.min.js', false, '' );

See the meandering Stackoverflow thread here.

Arbitrary Code Plugin for Wordpress

2010-12-07T17:02:00.000-03:00

My favorite silly plugin for Wordpress. Sometimes you don't care about what Wordpress can do for you. you just want to do some PHP.

Ruby Scripts to Get Keys As Array for Objective-C

2010-11-28T19:34:00.001-03:00

It's a common situation to have a line like this

@synthesize control, command, channel, data1, min, max;

which you want to make into one of these

NSArray *arrayOfKeys = [NSArray arrayWithObjects:@"control", @"command", @"channel", @"data1", @"min", @"max", nil];

While I could look for the Objective-C way to do it, another way is to cut my losses and do it with Ruby. Just fire up irb and paste this in

keys = "control, command, channel, data1, min, max"

keys = keys.split(", ")

text = "NSArray *arrayOfKeys = [NSArray arrayWithObjects:"

keys.each {|key| text << %Q!@"#{key}", ! }

text << "nil];"

puts text

Funky Errors with Unit Tests on iOS

2010-11-28T13:23:00.000-03:00

These are just two links for my own reference. If you see this message

Command /bin/sh failed with exit code 1

go here. Though the answer is that you need to get your Unit Tests set up.

If you see this message:

An internal error occurred when handling command output: -[XCBuildLogCommandInvocationSectionRecorder endMarker]:

go here.

Switch Ruby and Rails Versions

2010-11-26T09:56:00.000-03:00

Once you've done the one-time setup of rvm, some ruby versions, and Rails:

$rails -v

Rails 3.0.1

$rails _2.3.10_ -v

Rails 2.3.10

$ruby -v

ruby 1.8.7 (2009-06-12 patchlevel 174) [universal-darwin10.0]

$rvm -use 1.9.1

$ruby -v

ruby 1.9.1p378 (2010-01-10 revision 26273) [i386-darwin10.4.0]

$rvm -use system

$ruby -v

ruby 1.8.7 (2009-06-12 patchlevel 174) [universal-darwin10.0]

Once the Rails app is created, you manipulate the Rails version from the environment.rb.

Installing Rails 3 on OSX

2010-11-11T15:08:00.002-03:00

Most see this link. However, lets's skip the rvm stuff:

sudo gem install tzinfo builder memcache-client rack rack-test rack-mount erubis mail text-format thor bundler i18n

sudo gem install rails

which gave me this message

File not found: lib

The answer was here:

cd /Library/Ruby/Gems/1.8/gems/rails-3.0.1

mkdir lib

Also, I should mention that I never knew that

gem server

is the answer to all documentation on Ruby (thanks Sean). Another good answer is railsapi.com, including the 2.3.8 documentation and the 3.0.1 documentation.

I'll have to run through a Rails 3 book soon, I see, mostly to learn the stuff I should've known even in Rails 2.

Note: After writing this it turns out that I still couldn't connect to MySql from XAMPP on my OSX box. I ended up using this command

env ARCHFLAGS="-arch x86_64" sudo gem install mysql -- --with-mysql config=/usr/local/mysql/bin/mysql_config

but I have no idea why I have a mysql_config there. Perhaps it's from an earlier botched install of MySql? If all of this doesn't work, you can always run through the real MySql directions here.

From there, there was a simple two-step process:

1. Add

gem 'mysql'

to the gemfile.

2. Change the development section of the database.yml to say

development:
  adapter: mysql
  encoding: utf8
  database: test-rails3
  username: root
  password: 512ilwae2iwo
  host: localhost
  socket: /Applications/XAMPP/xamppfiles/var/mysql/mysql.sock

subbing the appropriate values as necessary.

Then use rails server and you're up and running.

Fascinating Style Guide

2010-10-24T02:59:00.001-03:00

This is a style guide on Objective-C, but it makes a case for self-descriptive code by basing it on a fascinating fact: "[Since] you do far more reading of code than writing, invest the time to type an extra eight characters." Of course you don't need to save characters typing (especially if you've learned to touch type)! While I'm here I should link to the controversial and insulting rant by Steve Yegge, who I didn't know about until today.

Windows Permissions Problems: ASP.NET Site Won't Serve

2010-10-04T02:28:00.000-03:00

I'll readily admit that I don't know much about Windows admin, but today -- suddenly, it seemed, but I'm sure I haven't touched my Windows VirtualBox in months -- I couldn't get an ASP.NET site to work at all. The error in the Event Viewer was completely vague, but I knew it was a permissions problem. In the end, I found a directory that did work, and copied the permissions from there. They have to look like this, or you're sunk (the CREATOR OWNER and SYSTEM being pretty pivotal).

Object-oriented Coding

2010-09-22T03:10:00.000-03:00

Object-oriented coding is easy, in some sense.

First you use functions. Repeated code disappears and the problem of "what do I call THIS variable?" disappears. Redefine the context, and suddenly giving nice, coherent names is easy. That's pre-OO but always the first step. No methods, no OO.

Then you start to generalize. When you generalize well, things start to fall into line. Messages are being called on "this" or "self" instead of that other thing. And code gets deleted from subclasses.

Then at some point you look at the monster base or superclass or the whole hierarchy and you say, "this cannot be good." So you start to delegate some of its work to other classes. If you do it right, they have a small bridge between them (and not a "Spaghetti Junction"). Then you start to work on the new structures that develop.

And as you add code, you continue to do this, and crystalline or fractal-ish structures form. And then you say, "my God, I've got a total mess. Any code could be anywhere." But of course that didn't happen, because you've got people yelling at you from every direction to do MVC or some super-structure to organize. But isn't it obvious, really, that you need to organize your stuff? Coding is 10% brilliant solutions and 90% organization.

And THAT the fizz-buzz problem does not test.

Problems With Unit Testing, XCode 3.2.4 for iOS

2010-09-19T18:35:00.000-03:00

I was beating my head against this for a while, but I eventually found someone else's solution. Here's the error

An internal error occurred when handling command output: -[XCBuildLogCommandInvocationSection setTestsPassedString:]: unrecognized selector sent to instance 0x200ed2d80

An internal error occurred when handling command output: -[XCBuildLogCommandInvocationSectionRecorder endMarker]: unrecognized selector sent to instance 0x200efecc0

I arrived at by following Apple's docs on Unit Testing.

The answer was here (http://stackoverflow.com/questions/3516745/3741466#3741466) and basically says to change the Run Script stage of the target from

"${SYSTEM_DEVELOPER_DIR}/Tools/RunUnitTests"

to

"${SYSTEM_DEVELOPER_DIR}/Tools/RunUnitTests" 1> /tmp/RunUnitTests.out

iPhone Apps with Companion Desktop Apps

2010-09-17T22:27:00.000-03:00

Yes, they are possible. While some early reports indicated that you cannot require a companion application, these turn out to be bogus (or premature, really): it's actually not a problem. Whew!

Sometimes, Ruby is Beautiful

2010-09-01T03:02:00.001-03:00

I hate \concise languages, since they seem so data-use averse (i.e., cheap), but sometimes Ruby is so cool

@config = YAML.load(File.open("#{RAILS_ROOT}/config/my_thinger.yml"))[RAILS_ENV].symbolize_keys

The String interpolation here is totally unnecessary, of course, but idiomatic.

Make a Universal App From an iPhone App

2010-08-16T01:55:00.002-03:00

I was trying to figure out what makes a Universal application tick, because I wanted the easiest way to switch mine over. Here it is:

In the build settings for the project, switch Targeted Device Family to iPhone/iPad
Optional: if you want to use different initial NIBs for the two devices, add this to your application's plist file:

<key>NSMainNibFile~ipad</key>

<string>MainWindow_iPad</string>

Then you can adjust anything you want differentially using

if (UI_USER_INTERFACE_IDIOM() == UIUserInterfaceIdiomPad)

It might be worth noting that right now the iPad can only do up to 3.2 at this time.

What about the default.png? See here for how to include images for the iPad: it's basically manual.

Autoresize, programmatically

2010-08-04T12:03:00.002-03:00

There are two ways to do this. One is the super-custom-I-don't-know-about-autoresizing way. Before I delete it (since it could be useful later, I wanted to document it here) It looks like this in a UIView subclass.

Init Method

didRotate method

Of course, this makes no sense at all when you can just do this (or do in IB, but that's not my point here):

The important advantage of the second one is that -- I think -- it actually readjusts DURING the rotation as well. That could be wrong.

Why iPhone Apps Don't Rotate With Keyboard

2010-08-01T15:10:00.000-03:00

I've always wondered, as a user, why most (but not all) iPhone apps cannot rotate when the keyboard is showing. The reason is that you generally need a UIKeyboardDidShowNotification to figure out how big the keyboard is. You cannot get one of these when you want. You usually get them only when the keyboard is being shown or hidden. I'm not saying it's impossible to get the keyboard height without this notification, but it might require calling private SDK methods.

Simple Serialization (Well, Archiving) on iOS: Filenames

2010-07-28T11:13:00.003-03:00

I sometimes forget that the simulator is not a real device, though it sure does get you there about 99% of the time. So anyway, the filenames that I presented in this post do not work. But after looking around for 10 seconds, I found the answer:

NSString *docsPath = [NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES) objectAtIndex:0];
NSString *filename = [docsPath stringByAppendingPathComponent:@"somewhereInAppSandbox"];

Then you can just use

[NSKeyedArchiver archiveRootObject:array toFile:filename];

and

[NSKeyedUnarchiver unarchiveObjectWithFile:filename];

as I suggested in the other post. Obviously for anything serious you would use CoreData or Sqlite.

NSKeyedArchiver: Encoding and Decoding for Fun and Profit

2010-07-26T22:43:00.005-03:00

More on Serializing and deserializing, which -- if you're ever on a job interview -- is called "archiving and unarchiving." Serialization is a superset, or something... I'm a bit loose with the terminology these days.

What if you want to get a serializable piece and then just reconstruct your object with it yourself? Could you save that piece to disk and recover it and do this too? I needed this for an app I'm doing, so the answer is yes. It works.

Simple object .h

@interface Person : NSObject <NSCoding>{
NSString *name;
}

@property (nonatomic, retain) NSString *name;

and .m

@synthesize name;

- (void)encodeWithCoder:(NSCoder *)aCoder {
[super encodeWithCoder:aCoder];
[aCoder encodeObject:self.name forKey:@"name"];
}
- (id)initWithCoder:(NSCoder *)aDecoder {
if ((self = [super initWithCoder:aDecoder])) {
self.name = [aDecoder decodeObjectForKey:@"name"];
}
return self;
}

and now the code that does everything

Person *person = [[[Person alloc] init] autorelease];

person.name = @"bob";

NSData *dater = [NSKeyedArchiver archivedDataWithRootObject:person];

BOOL result = [NSKeyedArchiver archiveRootObject:dater 

                toFile:@"somewhereInAppSandbox"];

NSLog(@"How was saving? %@", result ? @"yes" : @"no" );

dater = [NSKeyedUnarchiver unarchiveObjectWithFile:@"somewhereInAppSandbox"];

person = [NSKeyedUnarchiver unarchiveObjectWithData:dater];

NSLog(@"Are we there? %@", person.name);

Since we call the super methods for both init and encode, this works perfectly well even if the initial serialized object was a NIB.

The point of this entire exercise is to allow you to reuse your encodeWithCoder and initWithCoder methods to, for instance, make deep clones. This is very important because it will keep:

Your code DRY and therefore
You from going insane

Note: The filenames as presented won't work on an iOS device. See my later post here for just the filenames bit.

Serializing for iPhone, Part II: HelloWorlds With NSCoder

2010-07-24T19:31:00.001-03:00

So all this NSUserDefaults stuff is cute, but it's probably the most annoying option. It does handle some of the code for you, but you end up writing it in other places. So let's start again, this time using the NSCoder paradigm 100%. This time we have two objects:

Inner which holds a CGPoint

#import <Foundation/Foundation.h>

@interface Inner : NSObject {

CGPoint point;
}

@property CGPoint point;

@end

and Outer which holds an Inner

#import <Foundation/Foundation.h>
#import "Inner.h"

@interface Outer : NSObject {
Inner *inner;

}

@property (nonatomic, retain) Inner *inner;

@end

That's it! Now figure it out. I do not want to inform the client class about Outer's structure.

First, without any clue as to how this will work, I implement NSCoding for Outer like so (after making the changes to the .h):

- (void)encodeWithCoder:(NSCoder *)aCoder {

[aCoder encodeObject:self.inner forKey:@"inner"];
}

- (id)initWithCoder:(NSCoder *)aDecoder {
self.inner = [aDecoder decodeObjectForKey:@"inner"];

return self;
}

and for Inner

- (void)encodeWithCoder:(NSCoder *)aCoder

[aCoder encodeObject:NSStringFromCGPoint(self.point) forKey:@"point"];

}

- (id)initWithCoder:(NSCoder *)aDecoder {

self.point = CGPointFromString([aDecoder decodeObjectForKey:@"point"]);

return self;
}

Now let's just serialize and them, or something.

Outer *outer = [[[Outer alloc] init] autorelease];
outer.inner = [[[Inner alloc] init] autorelease];

outer.inner.point = CGPointMake(525, 525);

BOOL result = [NSKeyedArchiver archiveRootObject:outer toFile:@"somewhereInAppSandbox"];
NSLog(@"how did that go %@", result ? @"yes" : @"no" );

And deserialize them?

Outer *outer = [NSKeyedUnarchiver unarchiveObjectWithFile:@"somewhereInAppSandbox"];

NSLog(@"moment of truth %@", NSStringFromCGPoint(outer.inner.point));

Wow, this is actually pretty cool. I mean, not as cool as C#'s automatic XML serialization, but I'll take it.

Now... what happens if we house the Outer in an NSArray?

Outer *outer;

NSArray *array = [NSKeyedUnarchiver unarchiveObjectWithFile:@"somewhereInAppSandbox.anArray"];
if (array != nil) {

outer = [array objectAtIndex:0];

NSLog(@"moment of truth %@", NSStringFromCGPoint(outer.inner.point));

}

outer = [[[Outer alloc] init] autorelease];

outer.inner = [[[Inner alloc] init] autorelease];
outer.inner.point = CGPointMake(525, 525);

array = [[[NSArray alloc] initWithObjects:outer, nil] autorelease];

BOOL result = [NSKeyedArchiver archiveRootObject:array toFile:@"somewhereInAppSandbox.anArray"];

NSLog(@"how did that go %@", result ? @"yes" : @"no" );

Awesome. Now I can have a drink. So what was the point? This way we get an object graph, and each object -- via conforming to the NSCoding protocol -- gets to define how it wants to serialize. And standard property-set objects know how to do their own stuff anyway. Nice!

HelloWorld: Saving Application State for iPhone Apps

2010-07-24T18:01:00.001-03:00

So now I urgently need to know how to persist the data for my iPhone app. Otherwise, every time the user exits, they have to start from zero. Let's forget about how much mucking my app will need to be able to reconstruct state at any moment. First, the HelloWorld.

So before anything, I Google the problem and the first information points to the NSUserDefaults class. Apparently, this class can do it all, though if you need more complex models you might use CoreData. And it looks like all I'll need to do is to have my objects conform to the NSCoding protocol if they're in an NSArray. Nice. I might be able to do that. Let's get started.

But on the off-chance that you can just call writeToFile in arrays, let's try that first:

NSArray *fred = [NSArray arrayWithContentsOfFile:@"myArray"];

NSLog(@"is he null %@", [fred description]);

So that worked perfectly! It's null. So now let's try:

NSArray *bob = [[[NSArray alloc] initWithObjects:@"one", @"two", @"three", nil] autorelease];

[bob writeToFile:@"myArray" atomically:YES];

NSLog(@"that seemed okay %d", [bob count]);

And now I run it a third time and it's:

2010-07-24 16:06:56.217 TestingPersistence001[4442:207] is he null (

one,

two,

three

)

2010-07-24 16:06:56.219 TestingPersistence001[4442:207] that seemed okay 3

Which basically means that if we wanted to, we could run with that for now!

So let's go a bit further because that only took about 5 minutes so far, and see if we can get it into this NSUserDefaultsClass. And then we have to see if we can somehow serialize a thing that is holding a CGPoint (for instance).

So first we try to retrieve the defaults, so we can make sure we're not getting faked out (as we did above):

NSUserDefaults *defaults = [NSUserDefaults standardUserDefaults];

NSArray *bob = [defaults arrayForKey:@"bob"];

if (bob == nil)

NSLog(@"sorry it's nil");

Now I have no idea how to persist this thing. Hmmm... So first let's set the array. But wait! There's no setArray method. Whatever. It must use setObject and sort 'em out.

bob = [[[NSArray alloc] initWithObjects:@"one", @"two", @"three", nil] autorelease];

[defaults setObject:bob forKey:@"bob"];

NSLog(@"made it");

Okay, so putting these two together and running twice didn't work. It doesn't persist automatically. How does it, then? Let's try the synchronize method. Good, that worked. So this is the final HelloWorld code:

NSUserDefaults *defaults = [NSUserDefaults standardUserDefaults];

NSArray *bob = [defaults arrayForKey:@"bob"];

if (bob == nil)

NSLog(@"sorry it's nil");

else

NSLog(@"bob has %d things", [bob count]);

if (bob == nil) {

bob = [[[NSArray alloc] initWithObjects:@"one", @"two", @"three", nil] autorelease];

[defaults setObject:bob forKey:@"bob"];

[defaults synchronize];

}

NSLog(@"made it");

Great. Now we only have to make a custom object, put a CGFloat on it and we're done with most things.

Note: I'm oversimplifying and there's a bunch of stuff about domains on the save, and a bunch of stuff about property lists on the setObject. But let's finish this up first.

TheObject

#import <Foundation/Foundation.h>

@interface TheObject : NSObject {

CGPoint point;

}

@property (nonatomic, assign) CGPoint point;

@end

with standard .m file (with @synthesize!)

So we run this

TheObject *object = [[[TheObject alloc] init] autorelease];

object.point = CGPointMake(1024, 768);

[defaults setObject:object forKey:@"theObject"];

and the console says,

-[NSUserDefaults setObject:forKey:]: Attempt to insert non-property value '<TheObject: 0x59351c0>' of class 'TheObject'.

That was fast and wrong! Sooo... how can we make it a property value. Before reading anything, let's make it conform to NSCoding. And that didn't work. At all. And reading about it, it has nothing to do with anything. This thing has to be a property list. I can't take it anymore. Brute force!

On TheObject class

- (NSDictionary*) getSerializablePiece {

NSArray *keys = [[[NSArray alloc] initWithObjects:@"point", nil] autorelease];

NSArray *objects = [[[NSArray alloc] initWithObjects:NSStringFromCGPoint(self.point), nil] autorelease];

return [[[NSDictionary alloc] initWithObjects:objects forKeys:keys] autorelease];

}

And in the testing main

TheObject *object = [defaults objectForKey:@"theObj"];

if (object != nil) {

NSDictionary *dic = (NSDictionary*)object;

NSLog(@"what %@", [dic objectForKey:@"point"]);

}

object = [[[TheObject alloc] init] autorelease];

object.point = CGPointMake(1024, 768);

[defaults setObject:[object getSerializablePiece] forKey:@"theObj"];

[defaults synchronize];

man is that ugly. But it does work. Now that we're done with the HelloWorld, it's time to sweep the floor and figure out how to use this stuff for real(s).

Conclusion: you can use NSUserDefaults, but it's a bit particular about what it will serialize.

Blocks in Objective-C

2010-07-17T13:35:00.001-03:00

Unfortunately, Objective-C is dominated mostly by geniuses (who laugh at any possible ambiguity between "heap" and "stack") and newbies to programming ("I have an IBOutlet, but the screen is blank!?"). Thanks to iPhone development, there are increasing numbers of simple HelloWorld types of examples of things, but...

Block Basics

I'm not going to talk about whether the variables are on the heap or stack (which apparently is not that important in Obj-C, and many C developers misunderstand it anyway), nor about the problems with retain counts and other technical details about blocks. This topic is best covered by people who can really take it apart... maybe. Maybe I'll get to it myself. In any case, here's the definitive source.

But here are some examples for the lost. The syntax really really really sucks (compared to Ruby blocks, which are beautiful, but I digress).

Simplest Example (void in, void out) of Passing a Block to a Function

typedef void (^simplestBlock)(void);

@synthesize window;

- (void)applicationDidFinishLaunching:(NSNotification *)aNotification {

simplestBlock b = ^(void) { NSLog(@"inside simple block"); };

[self blah:b];

}

- (void)blah:(simplestBlock)b {

b();

}

@end

Simplest Example, NSString* in

typedef void (^simplestBlock)(NSString *);

@synthesize window;

- (void)applicationDidFinishLaunching:(NSNotification *)aNotification {

simplestBlock b = ^(NSString* thing) { NSLog(@"got your string %@", thing); };

[self blah:b];

}

- (void)blah:(simplestBlock)b {

b(@"got one");

}

@end

Simplest Example, NSString* in and out

typedef NSString* (^simplestBlock)(NSString *);

@synthesize window;

- (void)applicationDidFinishLaunching:(NSNotification *)aNotification {

simplestBlock b = ^NSString* (NSString* thing) { return [NSString stringWithFormat:@"processed: %@", thing]; };

[self blah:b];

}

- (void)blah:(simplestBlock)b {

NSLog(@"here goes nothing: %@", b(@"stirng sent to block"));

}

@end

Simplest Example In and Out, No typedef

The typedef is recommended by Apple: "You can also create types for blocks—doing so is generally considered to be best practice when you use a block with a given signature in multiple places:" But what does it look like without them?

@synthesize window;

- (void)applicationDidFinishLaunching:(NSNotification *)aNotification {

NSString* (^b) (NSString *) = ^NSString* (NSString* thing) { return [NSString stringWithFormat:@"processed: %@", thing]; };

[self blah:b];

}

- (void)blah:(NSString* (^)(NSString* thing))b {

NSLog(@"here goes nothing: %@", b(@"stirng sent to block"));

}

@end

I cannot seem to increase the retain count of the block, so perhaps you don't need to release it?

Later: You don't need to release it nor retain it. However, under ARC you should always use copy if you're going to store it.

Animation in iOS 4.0

2010-07-13T02:44:00.001-03:00

Well, I just got schooled on Core Animation. Apparently, iOS 4.0 changed everything, and now they want you to use blocks. Matz would be so happy, although it's a C thing (like so much of Ruby). Anyway, here's how I'm using blocks until told better (I could, at least, invert the logic so the BOOL is called "animate"):

- (void) makePhrases:(BOOL)inAnimation {

if (!inAnimation) {

[UIView animateWithDuration:3

animations:^{ [self makePhrases:YES]; }

completion:^(BOOL finished){ [self zoomIn:NO]; }];

return;

}

CGSize spaceSize = [@" " sizeWithFont:tView.font];

float leftMargin = 8;

float topMargin = 8;

I'm using this inAnimation parameter everywhere, until I can figure out why it's a bad idea. Blocks! Small world, eh?

Your First Extension With Radiant

2010-07-11T22:05:00.001-03:00

So I'm trying out Radiant, and one of my first questions is, what if I wanted to actually get a Rails view in there as content? Without a doubt I'll need to make a Radiant extension, so I'll be following the steps laid out here.

So you do a script/generate extension whatever and you get a full directory replete with views and even a routes.rb and everything. Which seems to be a normal Rails extension. So I just did a HelloWorld, but they did an example with scaffold. Either way works. The interesting thing is that your routes end up available in Rails just like normal. So:

map.connect 'helloWorld', :controller => 'root', :action => 'hello'

refers to the real url helloWorld, even though it's in its own mini routes.rb. So it's all normal and 100% Rails.

Only one thing in the instructions was wrong: the routes.rb, they say, "In the extension directory (/sample/vendor/extensions/queen), edit the queen_extension.rb file, to define the routes to your extension, replace:" but this should be in the routes.rb, actually (perhaps it's changed?).

And then you... follow the instructions and finish, and it works perfectly. Not much of a blog entry, I know, but perhaps my optimism about Radiant has gotten the better of me.

Custom UIActionSheet, Those Damn Blue Buttons and Simple Animation

2010-07-08T12:17:00.005-03:00

Custom UIActionSheet

CustomUIActionSheet? Sometimes blogs are annoying and the process is hard to follow. But here it's just a matter of grabbing the code in GitHub and making sure to call the viewWillAppear method. Great great great thank you to Andrew Homeyer!

He's got a view sliding up inside a background view. For what I need to do (right now), I ended up doing this:

self.tweakPage.view.frame = self.view.window.frame;
[self.view.window addSubview:self.tweakPage.view];
[self.tweakPage viewWillAppear:YES];

which is to say, you can attach it to the UIWindow at the top of your view hierarchy and then it really is -- for all intents and purposes -- modal.

You can also get into his NIB and put in a close button instead of the one he's got, and get rid of the UINavigationBar... actually you can do whatever you want. The example code is (like) perfect!

Those Cool Blue Buttons

What about the stupid blue buttons that appear in real iPhone apps? This is directly from the Apress book. First, download the icons from here:

http://developer.apple.com/iphone/library/samplecode/UICatalog/index.html

and then in your ViewDidLoad method do something like this

UIImage *image = [UIImage imageNamed:@"blueButton.png"];
UIImage *stretchedImage = [image stretchableImageWithLeftCapWidth:12 topCapHeight:0];
[self.closeButton setBackgroundImage:stretchedImage forState:UIControlStateNormal];

Then you can mess around with the button's type. I just got a great result setting the type to custom (in IB). Anyway... the stretchable part is the stuff you wouldn't figure out without opening the book.

Tweens on the iPhone

They're tweens. iPhone devs might not know this, but that's what we called them in Flash.

Animation on iPhone is pretty simple, or at least the simple kind is. This is 99% from the CustomUIActionSheet code above.

[UIView beginAnimations:@"someName" context:nil];

// Set delegate and selector to remove from superview when animation completes

[UIView setAnimationDelegate:self];

[UIView setAnimationDidStopSelector:@selector(animationDidStop:finished:context:)];

// move stuff by setting frames to the end positions

[UIView commitAnimations];

Then the animationDidStop method looks like this

- (void)animationDidStop:(NSString *)animationID

finished:(NSNumber *)finished context:(void *)context {

if ([animationID isEqualToString:@"someName"]) {

// do something here

}

}

NOTE: I just noticed that this is behind the times. See this question on SO.

Load UIView From Nib

2010-07-06T23:08:00.001-03:00

This was a hassle to track down and I actually ended up opening a book. The Internet is not very helpful sometimes. Books still rule, sometimes.

- (void)viewDidLoad {

[super viewDidLoad];

NSArray *bundle = [[NSBundle mainBundle] loadNibNamed:@"Blah" owner:self options:nil];

Blah *blah;

for (id object in bundle) {

if ([object isKindOfClass:[Blah class]])

blah = (Blah *)object;

}

[self.view addSubview: blah];

}

This assumes a Xib named 'Blah' with its view set to a UIView subclass named 'Blah' (which also exists).

Skype Call Problems

2010-07-01T16:36:00.001-03:00

I found this from a few years ago...

Random Colors in Objective-C

2010-06-21T19:14:00.003-03:00

My main focus in coding always has been doing Hello Worlds. Putting a bunch of them together to actually write a well-organized program that does something is trivial, really. HelloWorlds, on the other hand, are a bitch.

So anyway... I've been going through the different ways to get multiple pages in CocoaTouch (UIScrollView, UINavigationViewController, UITabViewController etc.). Then you can put one inside the other in lots of cool combinations. But when you're creating views programmatically, it's really helpful to have random colors so your HelloWorlds look cool. To do that, you need this ridiculously simple method:

- (UIColor *) getRandomColor {

return [UIColor colorWithRed:(random()%100)/(float)100 green:(random()%100)/(float)100 blue:(random()%100)/(float)100 alpha:1];

}

If you can simplify the math, you should. [Later] Here's a better version, based on the answer here.

- (UIColor *) getRandomColor {

CGFloat hue = ( arc4random() % 256 / 256.0 ); // 0.0 to 1.0

CGFloat saturation = ( arc4random() % 128 / 256.0 ) + 0.5; // 0.5 to 1.0, away from white

CGFloat brightness = ( arc4random() % 128 / 256.0 ) + 0.5;

return [UIColor colorWithHue:hue saturation:saturation brightness:brightness alpha:1];

}

World's Simplest Example: UINavigationController

2010-06-20T21:03:00.001-03:00

This is a Simplest Example of how to use a UINavigationController (based on my question here). The idea is that the example is 100% self contained. Warning: if you actually code this way (big monolithic classes) your code will suck. In this example, the UIViewControllers should actually be subclasses that contain the behavior and data they need to do their work. But that's another problem.

 - (void)viewDidLoad {
     [super viewDidLoad];
 
    UIViewController *one = [[[UIViewController alloc] init] autorelease];
 
    [one.view setBackgroundColor:[UIColor yellowColor]];
     [one setTitle:@"One"];
 
    navController = [[UINavigationController alloc] initWithRootViewController:one];
     // here 's the key to the whole thing: we're adding the navController's view to the 
     // self.view, NOT the one.view! So one would be the home page of the app (or something)
     [self.view addSubview:navController.view];
 
    one = [[[UIViewController alloc] init] autorelease];
 
    [one.view setBackgroundColor:[UIColor blueColor]];
     [one setTitle:@"Two"];
 
    // subsequent views get pushed, pulled, prodded, etc.
     [navController pushViewController:one animated:YES];
 }

Twitter Over Capacity

2010-06-04T10:53:00.001-03:00

parse_git_branch, still

2010-06-04T10:48:00.001-03:00

I'm still working with the parse_git_branch script that goes against the plumbing instead of the porcelain to get it right (well, I haven't done any of the actual work). I'm beginning to think that it's better to use what I had before (I'll redo it in Ruby sometime, probably): if you compare using git status (inflexible) to writing your own git status (crazy!), the former has some clear advantages: git status is a relatively complex program that works.

Using YAML to Move ActiveRecord Objects Between Rails Installations

2010-05-21T12:00:00.000-03:00

So I was trying to get some ActiveRecord objects from the server to my local box. This is not the right or cleverest way to do it, but it works.

On the server I just did

Then I moved to the file to local via ssh and fired up a local Rails console.

I read the file in

and then I realized that I had these weird YAML objects and no time to actually read any documentation on how they work. But thanks to spelunking with inspect, this worked:

where Thinger is my ActiveRecord obj. Now I'm locked and loaded.

Why did Ruby get Rails but Python got a whole bunch of things...

2010-05-19T13:23:00.001-03:00

My question about the history of Ruby/Rails and Python/Django got a bit of attention of late, so I finally got around to answering it. Ah, the beauty of talking to oneself in public! The question:

This is a(n) historical question, not a comparison-between-languages question:

This article from 2005 talks about the lack of a single, central framework for Python. For Ruby, this framework is clearly Rails. Why, historically speaking, did this happen for Ruby but not for Python? (or did it happen, and that framework is Django?)

Also, the hypothetical questions: would Python be more popular if it had one, good framework? Would Ruby be less popular if it had no central framework?

My notes:

Check out this article on why we'll never see Python-on-Rails. The author gives some of the basic reasons why Python has never had and will never have a central framework. I might add, myself, that Java doesn't have one either, and for the same reasons.

According to the author, Rails is strictly tied to its "implementation," which is Ruby. Rails was adopted by many developers and Ruby was just part of it. Rails works perfectly on Ruby (or Ruby copy-cats like Groovy), but more importantly, as many other answers say, Rails led the way to Ruby adoption.

This is why Rails-for-Python won't work, or at least what people have been focusing on with Rails isn't correct. It's not about the implementation or the quality of the framework, it's about the pattern of adoption. It's about putting the framework up front, and the implementation in the back -- even if this wasn't the Rails developers intentions (though maybe they are clever and this was their intention).

Basically, you can't get a bunch of language-loving folk to gather around a single framework. On the Java side, while Spring is well-loved, it's no Rails in terms of popularity in the Java community. In a mature community developers have their own ideas about what metaphors work and don't work in a framework. This is why Rails leads to Ruby and not the other way around (typically, mostly, not in all cases).

Are IE6 and IE7 Standalone Browsers The Real Thing?

2010-05-18T18:07:00.001-03:00

Today I finally got around to making Windows XP images to have all the versions of IE running for real. And then I compared them to the "standalone" versions and found no difference. Ever. Even on the strangest IE6 rendering.

So I asked for evidence of the "you can't have more than one real IE running at a time" position on SO. As expected, the question got little interest and no new data points. What did I expect? Hearsay is cheap, and evidence is hard to come by. The standalone IEs (not that IETester thing) are perfect, in my experience, and when they're not, it's that I got confused in my testing. I haven't been able to find any differences.

IE6 in MultipleIEs install.
IE7 Standalone.
Then install IE8 for real (as in download it from Microsoft).

Three IEs on one XP box isn't bad.

And browsershots ain't bad either. Litmus, on the other hand, seems like a cheesy simulator, in my two days of experience with it... but that might be wrong. Anyway, I don't care what the site looks like on a bunch of browsers. It's a UI: it's a question of look, feel and behavior. That's something screenshots can't tell you. That's why I need these browsers to take up as few virtual machines as possible, hence the initial question.

GIT Branch in Zsh (or Bash)

2010-05-17T15:24:00.004-03:00

It's been a long damn day and a half, but here's the final code for my .zshrc:

For Bash, here's my PS1 (albeit, the important part is \$(parse_git_branch) , but I don't want to go shopping for a Bash prompt for now):Big thanks to the people who answered my questions here and here. Now I'm finally ready for zsh!

Note: So much has been learned from this one, though. Even GIT has a plumbing vs. porcelain (logic/presentation) split. The new command, thanks to Chris Johnsen, teaches us all that what felt wrong all this time--parsing git status---really was. Using the plumbing commands gives us long-term stability. This is particularly important if you don't ever want to do shell scripting again :)

To be fair (to myself), parsing GIT status is just one call, whereas the new script is three. But until we have speed problems, we don't worry about it. Better yet, Chris' script runs on Bash and Zshell without modification!

Oh yeah, the only other change I made was to sub

ls -G

for

ls --color

from the zsh scripts I got from Zsh: The Last Shell You'll Ever Need.

Notes on Converting an HTML Comp

2010-05-16T14:27:00.001-03:00

I learned HTML on the street. I've been a corporate trainer in Java and XML, so I know all the basic stuff in those fields by heart, but I don't think I've ever read an entire book on HTML/CSS. On the other hand, maybe I have, but it was probably before I understood anything much.

This weekend was massive. I've been commissioned to do a tiny site on Rails (overkill, but good to have around). We've got the PDF comps and the resources half in Photoshop and half exported to GIFs, JPGs and whatever. The site is very simple, so the Rails part took about 10 minutes to finish the entire structure, including the paging stuff and my famous ie6? and ie7? helper methods, which I try never to use.

So what have I learned?

CSS Reset - Based on my lukewarm experience the last time I did a full site, this time I followed the CSS-reset route. I chose this radical CSS reset, which was probably not the best idea. A more opinionated CSS-reset might've worked better. But the CSS-reset, plus using the strict HTML doctype, resulted in a minimum of cross-browser problems.
HTML5: We didn't use any HTML5 on this project. But when I was asked to change the doc-type to the one for HTML5, and the site didn't budge, so the future is sound.
Cross-Browser Problems - I tested every single thing I did in IE6, IE7 and Chrome on Mac. This is totally annoying to say the least, but it kept me sane throughout (IE6 really makes one think that Microsoft was hoping to kill the Internet as a concept).
Design is a science, kind of: I usually kind of eyeball the comps and then build the thing. I measure later. Getting good measurements using PixelStick in the first place even while working on more interesting problems would have been brilliant (so I didn't do it this time, either).
One of the biggest mistakes I made was not making a serious attempt to figure out that the PDF comp needed to be zoomed to 73% (why?) to match the assets I was given. But measuring it worked. Design is empirical. Mostly, so...
Use a test harness to figure stuff out. A test.html (like a test.fla, or a HelloWorldSocketConnection or whatever) is invaluable, especially when compared to poking around in a complex situation to figure things out. HelloWorlds are our friend at all phases of a project.
Default position of elements is not static: it's relative. This took a while to debug, since I was debugging my knowledge itself :)... If the comps are really exact, just make the outer DIV relative, and then place everything inside with absolute positioning. This works solidly, unlike what I had thought previously. But more importantly...
Designers do design much faster than me, but I can still do design. I learned Fireworks a bit back, and while I can barely do anything beyond slicing in Photoshop, in Fireworks I can create assets as needed (though sometimes I have to go spelunking to find how stuff works). The thing about asset creation is that if you need an asset, you should try to measure what you do know. I only created one asset here: a rounded rectangular box with an inner and outer glow. The outer glow needed to go to transparent, and the inner glow needed to go over white (PNG8!). So I eye-balled the rounded corners, but the rest was data-based. I knew I needed it to have a certain size and use certain colors. The amusing thing was that when I finished I thought, "damn, that inner glow looks like crap, let me check the comps." The comps looked the same: anybody can do glows and drop shadows with modern tools.
Margins and padding are very different, though in certain browsers they kind of work the same. The point is mostly that negative margins are usually a mess (at least in IE), whereas negative padding is pretty cool. This is a topic for study, but nothing simple like this will help.
Z-index, negative margins and absolute position: z-index does not work with static positioning (I think), but with absolute positioning it works well. While in some cases I was able to get a negative top value to work, in others I couldn't get it below zero. In those cases a negative margin in addition to the position worked nicely. I have no definitive point here, but surely in 10 minutes of goofing around one could figure out what the real situation is.
Measuring stuff in em's is incredible. Though sometimes you want to measure in pixels, of course.
Font sizes like +1 and -1 are relative to the current font size. I wish I had known this a while back.
Next time, no more screwing around: I will do the CSS as an ERB. There is no reason to not be able to define variables in your stylesheet. Otherwise how do you keep colors consistent?

The next time I do one of these corporate designs, I'll probably do 90% of it as absolutely positioned.

Working with GIT was a dream, especially now that I know about soft and hard reset.

Now I can get back to my iPhone/Objective-C learning, which has taken a turn for the better. The other day I was actually writing code...

Why Software Development is Hard

2010-04-06T06:36:00.001-03:00

Software development is difficult to do well because the developer must constantly weigh:

Invention vs. reuse and integration time (build or buy, essentially)
Generalization/abstraction for reuse vs. the one-off job (hack): if the generalization never pays off, you've wasted time. Same goes with externalizing configurations, using inheritance, etc.
Building test instruments vs. building the actual software. No test instruments and the software takes forever to build. Too much testing and you never build the software.

Striking the right balance given a specific situation is difficult and requires experience and flexibility. I don't mean years of experience, I mean hours.

shell_here: Spawning a New Terminal Window in OSX

2010-04-04T12:40:00.001-03:00

My lack of desire to learn bash scripting has resulted in today's script.

This script will take your current terminal window and spawn a new one with your entire environment copied. paste_env will also, as a side effect, leave a file called /pasted_env.

shell_here

#!/bin/sh

env | pbcopy

osascript <<END

tell app "Terminal" to do script "$(paste_env)"

END

paste_env

#!/usr/bin/env ruby
`pbpaste`.each { |var| puts var }
puts 'clear'

This opens up a previously locked door for Ruby scripting: affecting the current shell.

I ended up bothering people here and here and here to do this.

So easy, you just call your Java code from Groovy!

2010-03-19T12:29:00.001-03:00

This is total crap. There are several cases where it just doesn't work, but my favorite is this runtime exception (well, we're using Groovy, what did you expect) that looks like this, "org.codehaus.groovy.runtime.GStringImpl cannot be cast to java.lang.String"... right right. So what should I do? I guess I'll have to go in and hack up my Java code with .toString() methods?