The first one, “git” just runs git with whatever command you provide to it (clone, pull, etc) along with any arguments you pass to it. Clone:

And pull:

(Other git command will likely work, these are just the ones I needed)

The second one, “git-clone-pull” uses the first one to clone a repository then pull from it. This effectively clones the repository if it hasn’t already been cloned, otherwise it pulls. However, since ant is fairly limited in what sorts of conditional execution you can perform, it just does both (clone will fail if it’s already been cloned, and pull will always be executed, even immediately after a the initial clone). Obviously not ideal, but it works, and I couldn’t figure out a better way without writing actual code.

There is plenty of room for improvement, but I suspect a proper Ant task written in Java is the right way to go.

Luckily my current RSS reader, NetNewsWire, has AppleScript built in, so I whipped up this script that prompts for the number of days you want to keep as unread, and marks the rest as read.

But wouldn’t it be better if it automagically filled in the username and password for you? I thought so, so I wrote a few lines of code in the form of a bookmarklet and a JSONP web service to do this.

BugMeNot doesn’t provide an API so I had to do a little screen scraping with Hpricot. They also try to obfuscate the usernames and passwords by shifting the characters by some offset calculated from a “key” then Base64 encoding the string, and prepending 4 characters. Luckily their obfuscation was no match for a single line of Ruby:

The bookmarklet makes the request via an injected <script> tag. When it’s callback gets called it finds the most likely input elements for the username and password and fills them in with the result.

The Rails app consists of a single action that makes a request to bugmenot.com for the specified site, extracts and decodes the usernames and passwords, and picks the one with the highest rating. It then returns the result as JSON wrapped in a function callback (i.e. JSONP)

I’m not going to post the location of the live JSONP web service since BugMeNot limits the number of requests you can make, but the code is available on GitHub.

I wanted to hook all the browser’s asynchronous JavaScript “entry points” : events, timers, asynchronous XMLHttpRequests, script tags, and “javascript:” URLs. This article deals with the first two, events and timers.

To do this, I figured I could somehow “wrap” all the callback functions passed to setTimeout(), setInterval(), and addEventListener(). By “wrapped” I mean another function that we specify calls the original function, rather than the original function getting called directly. This allows us do whatever we want right before and after calling the original function, including manipulating the arguments and return value, logging to the console, calling other functions, putting it in a try/catch, etc.

Here’s the implementation of “callbackWrap”, the function that does all the work:

To use it, you need function that takes another function as a parameter and returns a wrapped version of that function which does whatever you want it to. In this case it just prints out “whoooaaaaa:” followed by the “extra” parameter:

Finally, to actually use this, we call callbackWrap with the object that contains the function we want to wrap, the name of the function property, the index of the callback argument to the function (0 for setTimeout/setInterval, 1 for addEventListener), a wrapper “factory” function, and optionally an extra argument that’s made available (via the closure) to the wrapped function. The extra parameter can be used for any data you need to pass to the wrapper function, or it can be ignored.

Here’s how this would be used on setTimeout, setInterval, and addEventListener on window, Element.prototype and Document.prototype (so it applies to all Elements and Documents):

Here’s a live example. (edit: the button example seems to be broken in Firefox)

The first thing callbackWrap does is save the original function (window.setTimeout, for this example) in a local variable, “original”, since we’ll need it later. We then replace the original with a new function. When this new function is called (the new window.setTimeout), we first call the “wrapperMaker” function (”logWrapper” in the example), passing it the callback function, which is the argument at the argumentIndex position, and the optional “extra” argument. That function returns a new function which replaces the original callback argument, before we then call the original function (saved in the local variable “original”) with the same arguments (except the callback is now wrapped).

So really this blog post should have been called “a function that wraps other functions such that all callback functions passed to it are wrapped by yet another function”. There’s two levels of wrapping going on: the original function is wrapped, such that it wraps every callback given to it. If you’re confused I don’t blame you.

Why would I want to do this, you ask? There are a couple scenarios I had in mind, and I’m sure you can think of others.

First, I wanted to catch all uncaught exceptions for a debugging tool I’m working on. Sure, you can assign an error handler to the “onerror” property of the window object, but this only gives you the error name, file name, and line number – I wanted the full exception object to be caught by a try/catch.

Here’s a wrapper that catches all uncaught exceptions and alerts them:

Second, in Cappuccino we mimic Cocoa’s concept of a “run loop”, but due to the way browsers work the implementation is a bit different. A consequence of this was that we couldn’t automatically call “[[CPRunLoop currentRunLoop] performSelectors];” at the end of each run loop, which is required for various pieces of Cappuccino. If you’ve ever come across a situation where you perform some action but the UI doesn’t update until you move the mouse, it’s probably because performSelectors isn’t being called. This isn’t common, since Cappuccino encapsulates events, XMLHttpRequests, and now timers with CPEvent, CPURLConnection, and CPTimer, respectively. These Cappuccino classes handle calling performSelectors for you. However, if you wanted to integrate with a third party library or use setTimeout, XMLHttpRequest, etc directly, currently you would need to call performSelectors manually. With this new trick we can call it automagically.

Here’s the wrapper that does exactly that (note the inline Objective-J call to “[[CPRunLoop currentRunLoop] performSelectors]”):

One minor feature I didn’t demonstrate was that callbackWrap returns the original function that it wrapped. If you need the original for any reason you can save it to some other variable.

callbackWrap could use a few improvements. Currently this breaks on cases where you pass a string instead of a function. A simple check for the argument type, and appropriate handling would solve this (either skipping the wrapping, or compiling the the string with a “new Function()” call). But using strings instead of functions is considered poor practice anyway.

I wanted something more “real-time” and closer to the Quartz Debug utility included with the Mac OS X developer tools (which essentially provides this feature for all of OS X), so I present my modified bookmarklet and code:

Rather than recording each event and displaying them when you click, this version immediately disables the MozAfterPaint event listener (to avoid the recursion issue), shows the translucent red divs, waits 100 ms, removes the rectangles, and re-enables the MozAfterPaint event listener.

It will miss some events during the 100 ms flash, but overall it seems to work pretty well. You could modify it to re-enable the event listener between adding and removing the divs, but I’m not sure it’s worth the effort.

Well, I never got very far, but thankfully Patrick Geiller apparently had the same idea and actually executed it: He announced JSCocoa today. It looks like it’s a solid bridge, about up to par with PyObjC and RubyCocoa.

While the included GUI interface for trying out JSCocoa is nice, I prefer command line interfaces for my languages, so I ripped out the few lines of code from my original bridge and plugged in JSCocoa.

Code and build instructions on GitHub.

It’s very bare bones at the moment: it will either read one or more file names from the command line arguments, or if no arguments are supplied it will present a no-frills REPL. Obviously line-editing, etc would be one of the next steps, but for now it works nicely with rlwrap.

Ultimately I failed to recover most of the censored text (except “to”), so I had to cheat a little. The following video is the program running on a very similar image I created. This proves it works in ideal conditions, but needs some improvement to work in less than ideal cases.

(and no, as far as I know my friend’s project has nothing to do with eating monkeys)

Applying a filter like Photoshop’s “mosaic” filter obscures the original data, but doesn’t remove it entirely. If we can reconstruct an image with known text that looks very similar to original image, then we can be pretty sure the original text is the same as our known text. This is very similar in principle to brute-force cracking a password hash. For a more detailed explanation see this article.

Photoshop was an obvious choice since I needed to recreate the exact same fancy styling as the original image, then apply the exact same mosaic filter. I figured I would have to write a script that tells Photoshop to generate images, then use an external tool to actually compare them to the original.

It turns out that Photoshop CS3 has all the features necessary to pull the whole thing off without any other programs or tools. The most important feature is the JavaScript scripting environment built into Photoshop, which is far more powerful than the AppleScript environment (and a much nicer language, in my opinion).

CS3 added two other features that are critical to this task: Smart Filters, and Measurements. Smart Filters lets you edit a layer (namely the text with effects applied) after you apply a filter that would have previously require rasterization. This lets us apply the censoring filter to our styled text, and later change the text without having to manually reapply the filter. The “measurements” feature lets you record various statistics about an image or portion of an image: in our case we’ll want the “average gray value” of the “difference” between the original and generated images.

First we need to prepare the environment. Open the original image in Photoshop, and attempt to replicate the original un-censored text as closely as possible (you need some uncensored text as a reference). Place your text layer on top of the original and toggle between normal and “difference” blending modes to see how you close you are. Ideally everything will be black in “difference” mode. It’s very important to precisely match the font, size, spacing, color, effects like drop shadows or outlines, and even the background. If these are off even by a little bit it will throw things off. I ended up having to cheat because I couldn’t match the slick styling of the original text with my lame Photoshop design skills.

Once the text matches and is lined up perfectly, select the layer then choose “Convert for Smart Filters” from the “Filter” menu. Now select the censored portion of the text and apply the same filter used on the original image, again matching it as closely as possible. For the mosaic filter, you can line up the “grid” by adjusting the origin and size of the selection (yeah, it’s a pain).

Finally, make sure your layer is on top of the original, and the blending mode on your layer is set to “difference”. Double-click the Smart Object layer to open it’s source document, and adjust the variables listed at the top of the JavaScript to match the names and layers. Also, in the menu “Analysis”: “Select Data Points”: “Custom…” make sure only “Gray Value (Mean)” is checked.

Code

Rather than attempting to explain it in detail here, just read the code and comments. Here’s a quick summary:

1. Start with the first character. Try setting it to each of the possibilities (a through z, and a space), and record the difference score between the original image and generated image. Only look at the first half of the current character (since the second half will be influenced by the *next* character).
2. Sort the results. Lower scores are better (less different)
3. Now try each of the top 3 characters along with every possibility for the *next* character. This time record score for the whole width of the current character since we’re checking the next character as well.
4. Pick the best choice, either the best permutation out of all 81 combinations (3 best * 27 possible), or out of the 3 averages for each best.
5. Repeat for the next character until done.

The raw code and sample Photoshop file are available on GitHub.

Issues

This problem is particularly tricky for proportional fonts, since if you get any character wrong and it’s width is different than the actual character, then all subsequent characters will be misaligned, causing more incorrect guesses, compounding the problem even more, and so on. I’m not sure how to deal with this, other than improving the overall matching quality. Ideally we would test every possible combination for the entire string, but that would require 27^n tests, where n is the number of unknown characters. This is obviously not feasible.

With the simplistic method of iterating over each position and trying each possible character, it turned out that almost every single “guess” was for the letters “m” or “w”. This was because for positions where the original was narrower characters, the “m” would “bleed” over into the next position, improving the score regardless of how well it actually matched the current character. To get around this, we only look at the difference for the first half of the character’s position.

Since looking at the first half of the character removes some valuable information, we then do a second pass using the top several guesses from the first pass, this time looking at the full width of the current character along with each of the possible next characters (27 tests + 3 runs times 27 tests results in 108 tests per character).

Further improvements could definitely be made, but I’ve already spent several hours too many on this.

The current algorithm runs at about 3 characters per minute. The overhead of Photoshop saving the Smart Object document on every individual test case is significant. If this were a special purpose program manipulating images directly it would likely be much faster. The tradeoff, of course, is you have all of Photoshop’s flexibility at your disposal for matching the original document’s font, size, style, spacing, and censoring effects, which is very important. For small amounts of text speed isn’t a problem.

Conclusion

While my original goal of recovering the censored text on my friend’s page was never achieved, the project was a success. It works well on my test image, and I learned about 3 obscure but cool and useful features of Photoshop!

Oh, and that’s why ██████████ uses black ink to ██████ their ██████!

“git bisect” is a tool that facilitates a binary search of changes to your git repository to help find where a bug was introduced. You can walk through the process manually using “git bisect {good,bad,skip}“, or if you can write a script that automates checking for the bug, you can use “git bisect run scriptname” to have git do all the work for you.

The script should return 0 if the bug does not exist, and some other number (except 125) if the bug does exist.

In my case, Cappuccino’s “steam” build tool was failing, so I wrote a simple script that would test it by trying to run “steam” on Foundation:

I provided “git bisect start” with a known bad commit and a known good commit, then ran “git bisect run”:

git bisect start c1e882ace1dd29aea98d9247db304fe5d5077df7 d6c0f8802a2fd3a07e14418de7744ae04ae4499e
git bisect run ../test.sh

Sure enough, a few minutes later “git bisect” reported exactly which commit caused the problem:

01177a7e0237b1bd026cf0c4ca923fced8536772 is first bad commit
commit 01177a7e0237b1bd026cf0c4ca923fced8536772
Author: name removed to protect the not-so-innocent

Date:   Tue Sep 30 17:29:56 2008 -0700

   Fix keys conflict in CPDictionary
   
   [#77 state:resolved]

:040000 040000 bf4ceafbf439aa54790fc57a2a78dec8283abadf b269f85f3f4ed4d08e4c8261ee409e9d88255b1d M    AppKit
:040000 040000 f9c1fdb3c02c7c899285c8cb8123b29e99a17e46 48cc5aefde707119fe956dd882dcd8f8182e7012 M    Foundation
:040000 040000 b780743d17ddb0ebda188703725e9dabf23fd458 c72750dabd99ca42d9144f3074c68c42b727d028 M    Objective-J
bisect run success

So, I wrote this simple little bookmarklet which modifies the embed code for any YouTube videos to allow fullscreen. It could also easily be made into a user script for GreaseMonkey, etc, to perform the modifications automatically.

Bookmarklet

Source:

Try it out on this page (I’m too lazy to actually make it fullscreen).

Note that currently it doesn’t actually check to make sure it’s a YouTube video that it’s modifying, so it might stomp all over other types of embeds.