Android Ant builder

Android build process is quite complicated
First of all it is based on ant script together with sdk tools and java apps aligned to ant.

Lets take a closer look to it
to create ant script manually just type:

cmd <enter>
if you don't have Android SDK folder in system PATH environmental variable - add it
set path=%path%;X:\android\android-sdk-windows-1.5_r3\tools

now you can create new or update existent eclipse project
(I assume that you like me create project in eclipse first and then, for some reason, you need to build your fancy application from command line

set current dir to your android project dir

cd c:\apps\myapp <enter>
android update project  <enter>

now you have build.xml in your projects root folder

for now you can open them in eclipse but you cannot build (execute) ant script from eclipse
The problem is in java component that connect you build.xml ant script with platform dependent android ant script that located in SDK

for example for API level 8 it will be android_rules_r3.xml.

File contains a lot of comments about every section
but I'll explain some of them just for those who  who know nothing about ant engine :)

Ant stuff
I will explain only things that I need to build android project manually

every project contains
<project name="myapps" default="<default target>" basedir=<path to your android project> 
...
project>
nice to see basedir here because you are able to move you Ant script

main concept of ant is targets (like "release", "debug", "release-signed", ets.)
every target could have dependency on other targets
and targets could be public and private (hidden)

for example simple target:
<target name="help">
<echo>This project has extended helpecho>
target> 

complex target
<target name="release" depends="resources, java_compiler, zip, sign, renameing">

target> 

you can specify properties ( they are playing the same function as variables in other languages )
<property name="app.file.name" value="${build.prop.name}" />
let me explain this
you can read/ write app.file.name using ${app.file.name} sintax
in sample you define property and set value for it equal to value of  build.prop.name property

you could override this prop value in build.properties file
lets call it first level of customization

you also  able to define properties in build.properties and then load to project
<property file="build.properties"/>

Combination of both <property name.../> and <property file=.../> creates a list of properties that could be redefined in a separate settings file 
Example:

<property name="one" value="ha-ha"/>
<property file="build.properties"/>

now we can move to targets and other xml elements

<taskdef name="setup" classname="com.android.ant.SetupTask"
classpathref="android.antlibs"/>

Defines a java class with a starting "main" point in com.android.ant.SetupTask.
reference android.antlibs - a link to path(s) where SetupTask could be found.
All android elements (tags) could use (execute) this application
using <setup>setup> tag 

few more samples
<taskdef name="apkbuilder"
classname="com.android.ant.ApkBuilderTask"
classpathref="android.antlibs" />

<taskdef name="xpath"
classname="com.android.ant.XPathTask"
classpathref="android.antlibs" />
usage:

<xpath input="AndroidManifest.xml" expression="/manifest/@package"
output="manifest.package" />


<aaptexec executable="${aapt}"
    command="package"
    verbose="${verbose}"
    manifest="AndroidManifest.xml"
    androidjar="${android.jar}"
    rfolder="${gen.absolute.dir}">
aaptexec>   
if you command sequence that you have to use in different targets you could write a macro
definition:

<macrodef name="dex-helper">
   <element name="external-libs" optional="yes" />
   <element name="extra-parameters" optional="yes" />
   <sequential> 
      <echo>Converting compiled files and external libraries into ${intermediate.dex.file}...echo>
   sequential>
 macrodef>usage: 
<dex-helper /> 
or 
<dex-helper>
   <extra-parameters>
     <arg value="--no-locals" />
   extra-parameters>
   <external-libs>
     <fileset file="${emma.dir}/emma_device.jar" />
   external-libs>
 dex-helper/>

I hope you've got the idea.

Now about problem I met

for android platform 2.2 (API level 8) 
if you open build script that contains external targets (targets that were defined in android_rules_r3.xml or other external files)
in eclipse. you've get ant compilation errors (unknown target)

if you run the same script (build.xml) from command promt - it works fine

for example in case with auto-generated build.xml

this one return error unknown help target if you open it in eclipse
after that you are not able to compile you app!!! 
As a not very good solution is moving  build.xml file outside project root dir in this case eclipce will not see it and you can add to get it working from command line

Java final String optimisation

Some times I need a sub string of a string and I noticed that if you add final keyword, Java compiler makes a small optimization, isn't it?

For example :

String name = "this is a simple string";
String simple = name.substring(0,4);

Here Java creates two strings
but if you write this code in different way you get rid of one extra allocation

final String name = "this is a simple string";
final String simple = name.substring(0,4);

Both objects points to the same memory block and String object itself has a constant context
Am I right?

I am not sure about this conclusions, so, I will check it again when I have time.

If you know something how to use final for Strings and what exactly java compiler does, 

I'll be pleased if someone share his knowledges in this area

Bugs aligned to CString

Today I noticed a funny thing. Class CString C++ sometimes works like class String in Java
For example :
class Example
{
  CString mName;

public: 
  void setName(const CString & name)
  {
     mName = name; 
  }
  void modify()
  {
     memset((VPVOID)(LPTSTR) mName,65,mName.GetLength()*sizeof(TCHAR));
  }
}

class Other
{
// we set up this name somewere in code 
public: 
  CString getName() const
  {
     return mName; 
  } 
}


void main()
{
   std::autoptr o1 = std::autoptr(new  Other(_T("simple string")));
   Example ex1;
   ex.setName( o1->getName() );
   ex.modyfy();
   // here you will have empty string for o1 and for ex classes 
   // because it refers to the same string data in memory 
   // so both classes will contain modified data
 
} 


Conclusion : don't mix API because you never know about class implementation.
Fix is simple, just use mText.Empty() or mText[n] to modify string and never (LPTSTR)(mText)[n]

Windows messaging system

How I understand windows messaging system

I have taken a look to MFC, WTL frameworks and desided to explain some moments aligned to window creation and user interaction on Windows API level

Creating  a Window

To create window in Windows OS you call CreateWindow(className,...)
There is a enum of predefined classes
COMBOBOX, BUTTON, STATIC, LISTBOX and other default classes

As far as thay are already pressent in the system they asosiated with WindowProc. Actually every window class is asociated with windowProc inside OS.
What is window proc? it is just a function like below


INT_PTR CALLBACK MyDialog_DialogProc( HWND hwnd,
UINT messageId,
WPARAM wParam,
LPARAM lParam )


WindowProc is responsible for the window message processing. Thats why BUTTON for example looks like button :) , so to create a dialog not from window resources you need. To understand how OS GUI works lets create a dialog manually and not using resource file for that.


Creating Dialog window not from recources


1. register class

WNDCLASS wndClass = {0};
wndClass.lpfnWndProc = (WNDPROC)MyDialog_DialogProc;
wndClass.lpszClassName = L"DIALOG";
ATOM atom = RegisterClass(&wndClass);
HWND dialogHwnd = CreateWindow(L"DIALOG", L"MyDlg", ~WS_DLGFRAME &    ~WS_CLIPSIBLINGS, 10,10, 200,200, NULL, (HMENU)0, instance, 0); 

2. write a WindowProc

INT_PTR CALLBACK MyDialog_DialogProc( HWND hwnd,

UINT messageId,

WPARAM wParam,

LPARAM lParam )

{

if (messageId == WM_CREATE)

{

     // create button

   HINSTANCE hInstance = GetModuleHandle(NULL);

     HWND myButton = CreateWindow(L"BUTTON", L"ClickMe"
           , WS_CHILD, 10,10, 100,20, hwnd
           ,(HMENU)1, hInstance,0);

     if (!myButton)

     {

         // log error

     }

         ShowWindow(myButton, SW_SHOWNORMAL); 

  // create othe controls if there is a need!

  return TRUE;

}

return DefWindowProc( hwnd,  messageId, wParam, lParam); 

}

You can change this default WindowProc and this process is so calling subclassing
There is a lot of windows messages so if you don't need (or you don't want) to process all of them you need to call oldWindowProc to process messages that your new WindowProc doesn't process

I found that class "DIALOG" is not a default class in windows. This class creates when you are going to create a dialog based on recource file. When you create dialog from recource (using CreateDialog for example) this function register dialog class read resource file read styles inside dialog and raise WM_INITDIALOG and othe windows messages (WM_FONT is GWS_FONT defined) and after that call ShowWindow (if GWS_VISIBLE is defined)
all this steps is strictly synchronous because SendMessage is sync function.
How windows looks depends on window class and on windows style

Message handling 

As you may know some messages are received in thread queue others processed directly in WindowProc so you will never get WM_SIZE using GetMessage() or similar function

But messages from mouse and keybord will be sent to system queue
for example for erlier created messages in the same thread we use:


MSG msg;
while (GetMessage(&msg,0,0,0))
{
TCHAR buffer[MAX_PATH] = {0};
TCHAR windowName[MAX_PATH] = {0};
GetWindowText( msg.hwnd, windowName, MAX_PATH);
swprintf(buffer,L"Handle [0x%X] WindowName [%s] Message: [0x%X] \n",msg.hwnd, windowName , msg.message);
OutputDebugString(buffer);
DispatchMessage(&msg);
}

GetMessage Log:

Handle [0x0] WindowName [] Message: [0xC0B1] 
Handle [0x0] WindowName [] Message: [0xC0B1] 
Handle [0x60878] WindowName [MyDlg] Message: [0xF]  // WM_PAINT
Handle [0x60870] WindowName [ClickMe] Message: [0xF]  // WM_PAINT
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B7] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B9] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B9] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B9] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B9] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B9] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6] 
Handle [0x6086E] WindowName [CicMarshalWndAJCB] Message: [0xC0B6]


from the log we can see that only WM_PAINT message we receive from GetMessage. Messages with 0 handle are messages not aligned to windows. There is also a CicMarshalWndAJCB but we didn't create window with such mane. I created only two windows "ClickMe" and "MyDlg" so what is "CicMarshalWndAJCB" for?

of cource when I move my mouse to that window I receive all mouse/keyboard input messages like WM_MOUSEMOVE WM_KEYDOWN and othe hardware messages

In  the Internet I also could not find any information about that window , so if someone reading this could help, please do

To complete  this article I put log in dialog window proc to better understand what messages come  first

WindowProc Log



Handle [0x3B086A] Message: [0x24]  //WM_GETMINMAXINFO
Handle [0x3B086A] Message: [0x81]  //WM_NCCREATE
Handle [0x3B086A] Message: [0x83]  //WM_NCCALCSIZE
Handle [0x3B086A] Message: [0x1]   //WM_CREATE
Handle [0x3B086A] Message: [0xD]   //WM_GETTEXT my call
Handle [0x3B086A] WindowName [MyDlg] Message: [0x1] //WM_CREATE we receive only one wm create but I put log just to be sure that windowname is MyDlg
Handle [0x3B086A] Message: [0x210] // WM_PARENTNOTIFY
Handle [0x3B086A] Message: [0x18]  //WM_SHOWWINDOW
Handle [0x3B086A] Message: [0x46]  // WM_WINDOWPOSCHANGING
Handle [0x3B086A] Message: [0x46]  // WM_WINDOWPOSCHANGING // this call from thread
Handle [0x3B086A] Message: [0x1C]  // WM_ACTIVATEAPP
Handle [0x3B086A] Message: [0x86]  // WM_NCACTIVATE
Handle [0x3B086A] Message: [0xD]   // WM_GETTEXT
Handle [0x3B086A] Message: [0x6]   // WM_ACTIVATE
Handle [0x3B086A] Message: [0x7]   // WM_SETFOCUS
Handle [0x3B086A] Message: [0x85]  // WM_NCPAINT
Handle [0x3B086A] Message: [0xD]   // WM_GETTEXT
Handle [0x3B086A] Message: [0x14]  // WM_ERASEBKGND !! I have handled it
Handle [0x3B086A] Message: [0x47]  // WM_WINDOWPOSCHANGED
Handle [0x3B086A] Message: [0x5]   // WM_SIZE
Handle [0x3B086A] Message: [0x3]   // WM_MOVE
Handle [0x3B086A] Message: [0x7F]  // WM_GETICON
Handle [0x3B086A] Message: [0x7F]  // WM_GETICON
Handle [0x3B086A] Message: [0x7F]  // WM_GETICON
Handle [0x3B086A] Message: [0xF]   // WM_PAINT
Handle [0x3B086A] Message: [0x135] // WM_CTLCOLORBTN ClickMe send this message to parent


So this mechanism works very good
all other stuff and errors and a lot of complains alligned to higher level of API in user mode. Teoretically it is possible to rewrite user mode API to have very simple and ea sy to use operating system. I think this is might be done by 3rd party vendows that works with Windows CE OS

NDK Android

Working in Android in the native level is quite interesting

To start working under Windows platform you need

1.  Cygwin you need version for Windows
2. Go to Android NDK framework and download NDK, SDK
3. Install Eclipce and ADT plugin

Also you need JDK 1.5 or 1.6 installed
To Edit C++ sources you could install  CDT C/C++ plugin Eclipce or use Microsoft Viusal Studio or  Notepad++ with c++ plugin

Also download Android cpp sources. Using sources as manual you will understand more about what android consist of and will be able to all native library or drivers to extend Android functionality

Subclassing child window using WTL

To create custom control  you may use WM_DRAW_ITEM but this message you will receive in DLGPROC
So you have to draw control not in his own procedure but in parent one.

To resolve this complexity you can use subclassing with the child control.
If you are using WTL - all need to do is create class dediven not from standard WTL controls with other template parameter

For example for ATL::CListBox use

class CListBoxEx;

typedef CWindowImpl< CListBoxEx, CWindow, CControlWinTraits> CMyListBox;
template <>
HWND CListBoxT::Create(HWND hWndParent, ATL::_U_RECT rect, LPCTSTR szWindowName,
DWORD dwStyle, DWORD dwExStyle,
ATL::_U_MENUorID MenuOrID, LPVOID lpCreateParam)
{
      return CMyListBox::Create( hWndParent
                               , rect
                               , szWindowName
                               , dwStyle
                               , dwExStyle
                               , MenuOrID
                               , lpCreateParam);
}

template <>
CListBoxT::CListBoxT(HWND hParentHwnd)
:CMyListBox()
{
   if (::IsWindow(hParentHwnd))
   {
      CMyListBox::SetParent(hParentHwnd);
   }
}

class CListBoxEx :
public CListBoxT
{
...
}
 
This code allows you create a control and using Subclass-window replace WNDPROC for that control class
Now we can create or subclass a control


class CList_boxDialog : public CAppStdDialogResizeImpl,
public CUpdateUI,
public CMessageFilter, public CIdleHandler
{
... 
LRESULT OnInitDialog(UINT uMsg, WPARAM /*wParam*/, LPARAM /*lParam*/, BOOL& bHandled) 
{
    ..
    CWindow wnd = GetDlgItem(IDC_LIST_ADAPTERS);  
    mListBox.SubclassWindow(wnd);
    mListBox.SetParent(m_hWnd);

    mListBox.ResetContent();
    mListBox.AddString(_T("First"));
    mListBox.AddString(_T("Second"));
    ..
}

..
CListBoxEx mListBox;
} 

Code above get controls window handle and WTL replace controls window procedure


You should allow all unhandled messages to pass to old WNDPROC to do it just return 'false' from
ProcessWindowMessage()  

in CListBoxEx class:


BEGIN_MSG_MAP(CListBoxEx)
   MESSAGE_HANDLER(OCM_DRAWITEM, OnDrawItem)
   MESSAGE_HANDLER(OCM_MESUREITEM, OnMesureItem)
END_MSG_MAP()


Next step is message reflection

One of the most difficulties is to follow code in templates .. 
In your CDialog class you need to reflect all messages so they will be handled in child control class
to reflect all unhanded messages use

class CMyDialog;
BEGIN_MSG_MAP(CMyDialog)

    REFLECT_NOTIFICATIONS()
    CHAIN_MSG_MAP(CAppStdDialogResizeImpl<CMyDialog>)END_MSG_MAP()

Now you should be able to receive OCM_XXX and handle them in child control

This trick allow you to hold code in one place and increase portability and reduce time to create applications
As usual people don't create GUI in C++  even for Windows CE, Mobile but who knows ;)

Nano-X window system

Recently I have worked in Linux environment with a great NanoX project
It is about 300k size. NanoX provide solution for timer support, windows management, user input.

To use Nano-X you need only gcc c/c++ compiler that it is included in many(most) Linux systems

Architecture:

Features:
- message based system
- small size
- simple API
- add/replace any keyboard/mouse/screen driver
- font support. It is possible to add Unicode TrueType fonts

Lets discuss how to use it:

To create nano-X window you need :


#include  "nano-x.h"
  
int main(int argv, char[][] argv)
{
   GR_WINDOW_ID wnd = 
     GrNewWindow( GR_ROOT_WINDOW_ID
                  , 0,0,100,100
                  , WINDOW_NO_BORDER
                  , FOREGROUND_COLOR
                  , BLACKGROUND_COLOR); // create a window

   gc = GrNewGC(); // create a graphic context


   //select events. you are able to select 
   //only that events that you need. this will boost your application  
   GrSelectEvents( wnd, GR_EVENT_MASK_ALL); 


   GrMapWindow( wnd ); // show window


   // you may create as many windows as you wish
   // there is only one message queue for all Nano-X windows


   GR_EVENT anEvent;
   while (true)
   {
     GrCheckNextEvent(&anEvent);
     if (anEvent.type == GR_EVENT_TYPE_NONE )
     {
       sleep(1); 
       continue;
     }
     if (HandleEvent(anEvent) == NEED_TO_EXIT) break; 
   }
   // close all handles and exit
   GrUnmapWindow(wnd); // hide window
   GrDestroyGC( gc ); //
   GrDestroyWindow( wnd ); 
}

Now you might understand nano-x weak side: 

• Messages might not be synchronized. ( for example when you kill window but some messages that was in queue might come)
• Timer messages might not work proper under some circumstances  (for example when you create a timer event, then change time in your system) 
• Pure focus management (for example you can use GrSetFocus() then destroy window and then GrGetFocus() doesn't return you focus to a parent window like it is in MS Window OS,
• Font support is available but you have to dig into sources to add it. 

iPhone SDK

Yesterday, I start investigating IPhone SDK
there are few links that may help to start developing

MAC OS reference library SDK and framework

Objective C /C/C++  compiler for windows platform

Yii framework

Not so long time ago I have learned php5 Yii framework
It is quite simple and well documented MVC  framework
I like it!

So why I think it will be popular?
because it is easy to use.. and it is free
Here is  few the most important features for me

Localization 
first of all I pay attention to frameworks localization
only one static function  

Yii::t("section", "string")

Security
It is very important to have only one entry point for all user requests
in this case you can wrap this entry with security patterns and framework itself already did it for you

All requests have the same entry for example


http://mysite.com/index.php?r=action&...

of cause you are able to create more attractive requests like


http://mysite.com/home/livingroom
http://mysite.com/2010/09/01/school

there is a trick.. all this urls redirect your request to one php file. remove php file entry name like "index.php" is done using .htaccess or if you are able to configure apache yourself it is could be done in *.conf file

Database
You are able to request data  from database even if you don't know sql and even if you are not sure what database is installed on you web server. I mean code that Yii framework produces is working with any sql database platform (MySQL, PostageSQL, SQLite or other PDO php extention. 
As results of the request it returns associative array and all other low-level stuff framework hides inside.

Widgets and Components
Widgets is a way to reuse you code - very powerful thing

I am not going to describe all Yii features, there is a manual for that

Android Streaming Audio

Android audio framework

I will try to explain as detailed as possible about streaming audio in android. For this purposes in SDK 1.5 has been added AudioRecord and AudioTrack classes. AudioRecord is responsible to get samples from microphone and AudioTrack is responsible for playback samples.

This classes works only with PCM encoded data (it means that they are not encoded at all :)

Examples:
Play one buffer loop

Initialzation

AudioRecord record = null;
AudioTrack track = null;
 
record = new AudioRecord(MediaRecorder.AudioSource.MIC, 8000,
  AudioFormat.CHANNEL_CONFIGURATION_MONO,
  AudioFormat.ENCODING_PCM_16BIT, 
    AudioRecord.getMinBufferSize(8000, AudioFormat.CHANNEL_CONFIGURATION_MONO, AudioFormat.ENCODING_PCM_16BIT) * 4
  );

log.d("AudioTrack min buffer size: "
+ AudioTrack.getMinBufferSize(8000, AudioFormat.CHANNEL_CONFIGURATION_MONO,
AudioFormat.ENCODING_PCM_16BIT));

track = new AudioTrack(AudioManager.STREAM_VOICE_CALL, 8000, AudioFormat.CHANNEL_CONFIGURATION_MONO,
AudioFormat.ENCODING_PCM_16BIT, AUDIO_SYSTEM_BUFFER_SIZE, AudioTrack.MODE_STREAM);
 
Initialization is quite simple. All you have to do is to create AudioRecord and AudioTrack

Start audio player
track.play();


Start audio recorder

record.startRecording(); 

Main Loop 
while (needMoreAudioData)
{
     recorder.read(buffer, bufferSize);
     player.write(buffer, bufferSize);
} 

Stop audio player/recorder
track.stop();
record.stop(); 

Remove player
Because this objects works with system resources developer has to release them when they are not in use. Java garbage collector will not release them automatically because instances hold system handlers.
 
track.release();
record.release();


Hope you have got the idea ...

Now few practical tips

When you call  record.startRecording();  recording is already started! and to reduce latency you have to read that data immediately.

You should realize that android is not a real-time operating system and often call backs come not exactly at certain. You will end-up with increased latency over time unless you apply a sort of throttling algorithm to adjust playback side.