It's over a year since I have my G1. I'm still loving it mostly, but there are signs that I need a new phone whenever a new one coming.

One of the biggest annoyance I have is the probably the CPU, or so I believed. The power of Android is I can run multiple programs at the same time and quickly switch among them. But in reality, G1 is getting sluggish when I have Twitdroid running and listen to podcast/music. Even with no program in the foreground, the Home screen with widgets (I have 9 running across 3 screens) already showing slow response from time to time. The same when I run Google Maps, surf the web, and in general multi-tasking. I do not think this is out of memory issue, since system monitor reports memory usage is under 80%. So, I think a faster CPU is needed, a Snapdragon is a requirement for me and let's hope it's not a battery eater.

Battery life is the next major short-coming on G1. Granted, I run a lot of things that keep the phone active, but it looks like making a phone call or two eat up quite a look of battery. In a typical day, I'm able to keep my G1 away from charger if I only listen to music and read tweets lightly while on MUNI. But soon I talk on the phone, I have to charge the phone otherwise I won't last til leaving work.

In addition, I have yet to have T-mobiles HSPA 7.2(Mbps) coverage, which my G1 can get at. But with T-mobile deploying HSPA+ (@21Mbps) next year, there is no reason not to have more downlink/uplink bandwidth.

I never realize this pitfall in PHP related to foreach until I revisit the manual when prepare for PHP certification:

$somearray = array('1', 2', '3', '4');

foreach ($somearray as $value) {

  $total += $value;

  $value += 10;

}

When dumping the contents of $somearray after the loop, the values are not changed..... I ran into this in a few occasions while working on NucleusCMS, but didn't realize until I read doc again in PHP5:

... On each loop, the value of the current element is assigned to $value and the internal array pointer is advanced by one (so on the next loop, you'll be looking at the next element). ...

The key here is "assigned". When the code changes $value, it actually got overwritten and discarded in the next iteration. The loop works properly for $total or if I only read from $value.

And yes, the pitfall was pointed out in the comments.... but who reads it? Certainly I didn't.

Because of this problem, there is a second form of foreach which provides the key as well. It turns out the second form is designed to fix just this pitfall:

$somearray = array('1', 2', '3', '4');

foreach ($somearray as $key => $value) {

  $total += $value;

  $somearray[$key] += 10;

}

Now, the contents in the array are changed after the loop.

Someone must have thought of a better way to handle this since PHP4. There is now a solution involving reference in PHP5:

$somearray = array('1', 2', '3', '4');

foreach ($somearray as &$value) {

  $total += $value;

  $value += 10;

}

The result is identical to the second form of foreach loop.

Sometimes, it's just not clear of the rational behind a language feature at first, like in this case (the second form of foreach). But once you understand the reason, it's much appreciated, especially when an elegant solution comes along (the reference in foreach).

Cool, eh?

My first project at L::G is built with CakePHP. It's been almost 3 months since I started using it. Here's some of my observations.

It's really a piece of cake!
I started with the tutorial. A few hours later, I gain enough confident and start working on the real project. It's just a simple tasks to add CRUD, but I already know how to do it after following the tutorial and look at code already in the real project.

Once I get a hang of MVC, it's a pleasure to work with; Define model that tie to database, access data with find(). Logic/algorithm goes to controller. Pass output to view via variables for output, which almost like NucluesCMS's skin.

It makes easy the things that matter the most
CakePHP has a very powerful mean to access data with containable. One find() often can scoop up all the related data and ready for action.

There are support to many common tasks; pagination, authentication, ACL, and etc. These build-in functions provide great help to web application developer.

Data submitted by user could contain data that harmful to the application, therefore, checking these data are an important task. However, CakePHP's Sanitize core library makes sanitizing in-coming data a snap.

Scafffolding is a powerful function in Rails application. CakePHP provides the same function and it's as easy as setting a var $scaffold in controller to get it to work.

The debugger function provides a great way to inspect the controller execution, It saves me a lot of echo to dump internal variables (but still have to set and pass them to view) .

After all, you are still a web application developer
It might be a disadvantage, but it does not bother me that there is no installation script. In order to get a CakePHP project up and running, there are only a few changes to make after unzip the package to webroot.

Along the same line, there is no tool to auto generate model/controller/view code to start with. But this generally mean I have to cut and paste some more code.

On the other hand, CakePHP's handle to model and database mapping is rather raw and bare. To create and add a model to my project, I have to manually create the database table. It feels like a car with manual transmission, I get to do all the things.

Of course, nothing is perfect
Through out the course of the project, I've ran into a few bumps.

It is not a big issue, but CakePHP imposed a strong naming convention for model/controller/view naming/database table/database columns, there are implicit behaviors associate with the convention. It trips me a few times at the beginning, but I get around it after I get use to it.

The documentation, according to a comment, is a great improvement. I found it's right to the point and easy to find information. However, what's good for documentation if important information is missing, or you need to read all the comments to learn a pitfall or realize doc? I can see there are still a lot of rooms to improve.

Meanwhile, the new and improved data validation still not mature enough (as of 1.2.5). i.e. range rule is non-inclusive. For a range check of 0-10, the code is actually 'rule' => array('range', -1, 11)...

Containable only work for second level, I have a few cases that need a third level containable with conditions. In those cases, I ended up have to do extra find() to get the needed data.

Performance, performance, performance. It's scary to see from the debugger so many extra SQL queries to get the job done.

There are still unanswered questions...
Even with some digging around, some questions still remain a mystery to me and I'm sure some re-factoring may come out because of these:
- What if I need transaction/rollback? Model data validation for associated models help, but is it enough? I'm having problem with cross-model data validation i.e. A question must has 2-5 answers - How do I enforce that?
- How to design architecture for functions that not quite fit into MVC model? i.e. SMS function is the prime example at hand.

Just installed Fedora 12 on my laptop, glad to see Eclipse updated to 3.5.1. But as usual, I have to jump through hops to get Android and App Engine SDK installed again. Sign.

First, I installed required Eclipse plugins available from Fedora (from GNOME System -> Admin -> Add/Remove Software):
eclipse-swt-3.5.1-4.fc12.i686
eclipse-platform-3.5.1-4.fc12.i686
eclipse-subclipse-1.6.5-1.fc12.noarch
eclipse-linuxprofilingframework-0.3.0-2.fc12.i686
eclipse-emf-2.5.0-4.fc12.noarch
eclipse-jdt-3.5.1-4.fc12.i686
eclipse-subclipse-graph-1.6.5-1.fc12.noarch
eclipse-emf-xsd-2.5.0-4.fc12.noarch
eclipse-rse-3.1-2.fc12.noarch
eclipse-svnkit-1.3.0-1.fc12.noarch
eclipse-phpeclipse-1.2.1-5.fc12.noarch
eclipse-cdt-6.0.0-10.fc12.i686
eclipse-rcp-3.5.1-4.fc12.i686
eclipse-gef-3.5.1-2.fc12.noarch
eclipse-callgraph-0.0.1-3.fc12.i686
eclipse-dtp-1.7.0-5.fc12.noarch

I installed a few extra for C/C++, PHP development and various things.

Then, install the SDKs and missing plugins from Eclipse's update manager (Help -> Install New Software):
Web Tools Platform - http://download.eclipse.org/webtools/updates/ (might have to skip Dali components....)
Google ADT plugin - https://dl-ssl.google.com/android/eclipse/
Google App engine plugin - http://dl.google.com/eclipse/plugin/3.5

After some waiting and a few restart. The two SDKs should be installed and working.

I've been thinking about setting up this new blog for a while.

I become aware it's not the best to mix all my topics in a single blog a while back. I would like to use this blog as the focal point to my tech-related interests, to separate from my more private/casual blog.

I moved all my tech posts to this blog so they got better visibility.

[update]The site is almost ready, now let's get the tag cloud working![/update]

The Google Reader team just unlashed a bunch of new features. The most interesting one is the ability to send an item to other social networks like twitter, Facebook, and etc.

I've been trying to find a good way to share interesting items I read on Google Reader since I switched from Gregarius. I have a blogroll in my blog, which is sync to delicious. I used to have Friendfeed consolidate my blogroll and stream it to Facebook.



With the new Sent to function, I'm able to directly send an item to Twitter. Then I read and share items on Google Reader to share with people following me, and have Friendfeed pick it up and stream to Facebook.

But the most interesting part is user can customize and send to 3rd party service not listed. I setup Google Reader so I can share an item and add to blogroll separately (which I use as a link dump). This way, I can selectively bookmark share items that I also want to keep on my blogroll.

Google's Chrome OS is still a while from release. However, I'm already curious how its system architecture likes.

Chrome OS is designed to be a light-weight OS for the net today, where applications are hosted on the cloud and access via wireless or wireline broadband connection. It is initially targeted at netbooks. What's this translate into plain English: slower CPU, small screen/keyboard/mouse, and storage on-line.

Google aims to make Chrome OS fast and lightweight, to allow system start up in seconds. It has a minimal user interface, and let the application take care of other aspects of the user experience. Most interestingly, it has a security architecture redesigned to prevent virus and malware.

At the moment, not much is known about Chrome OS's software architecture, except it is "Chrome running within a new windowing system on top of a Linux kernel". But there are comment that Android and Chrome OS might "merge even closer" in the future. This make sense because there are many overlap functionalities between the two platforms. So, let me jump ahead and assume that these OSs going to be merged in the future. If I were to make them common in the future, I might well start to design them similarly or even reuse as much as possible.

Here's an architecture view of Android:


From the view, I'd think the Linux kernel, libraries, and application framework can reuse on Chrome OS. The Android run-time is dropped, obviously. On the application layer, only one application - Chrome is running.

Meanwhile, some tweaking on the application framework is needed to address how Chrome to access hardware; LAN/Wifi/broadband network device, build-in camera, printer, and etc. One solution I can think of is to add a web server-like service to provide local device access via HTTP. Afterall, we already configure WIFI router via web interface and there are WIFI web camera to stream video; We are just moving these functions to access local devices. In addition, Chrome already used Gears for off-line operations.

Viola, Chrome OS we got here.

Further reading:
10 Things We're Dying to Know About Chrome OS

A quick search reveals how well bing.com works: