siddharth's space

1. Install JDK. The latest version is 1.7 and it gets installed at C:Program FilesJavajdk1.7.0_02 on my machine
2. Install Android SDK. It gets installed at C:Program Files (x86)Androidandroid-sdk
3. Install Apache Ant. It gets installed at C:Program Files (x86)apache-ant-1.8.2
4. add following folders to PATH environment variable:
C:Program Files (x86)Androidandroid-sdktools;
C:Program Files (x86)Androidandroid-sdkplatform-tools;
C:Program Files (x86)apache-ant-1.8.2bin;
C:Program FilesJavajdk1.7.0_02bin;
5. create following environment variable:
variable name: JAVA_HOME
variable value: C:Program FilesJavajdk1.7.0_02
6. edit C:Program Files (x86)Androidandroid-sdktoolsantbuild.xml as mentioned in http://stackoverflow.com/a/8225017/147530
7. Follow steps at aplusbi.com to create a virtual device, create your android project from cmd line
8. start emulator (C:\Program Files (x86)\Android\android-studio\sdk\tools\emulator.exe)
9. ant debug install will compile debug binary and install onto device
10. to build release version, create your key, then edit ant.properties file and use ant release install

Signing the app: the google play store will require you to sign the app before you can publish it to the store. Follow steps at http://developer.android.com/tools/publishing/app-signing.html for signing.
keytool.exe can be found in C:\Program Files\Java\jdk1.8.0_25\bin
zipalign.exe is located at C:\Program Files (x86)\Android\android-sdk\build-tools\21.1.1
My ant.properties looks like this:
key.store=c:/users/siddjain/jainaartis/siddjain-release-key.keystore
key.alias=helios

When you run ant release, you will be prompted for two passwords:
[input] Please enter keystore password (store:c:/users/siddjain/jainaartis/siddjain-release-key.keystore):
[input] Please enter password for alias ‘helios’:

These two passwords can be changed on later if you desire, by following two commands:
To change the first password use keytool -storepasswd
To change the second password use keytool -keypasswd

Both these passwords are stored in encrypted form in the .keystore file

Troubleshooting:
– emulator takes a long time to start http://stackoverflow.com/questions/20051073/emulator-for-android-4-4-kitkat-is-not-starting

Developing app in html/js and embedding in webview. follow tutorial on google website. note below:

protected void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        if(Build.VERSION.SDK_INT >= Build.VERSION_CODES.KITKAT) {
            WebView.setWebContentsDebuggingEnabled(true);
        }
        setContentView(R.layout.activity_main);

        mWebView = (WebView) findViewById(R.id.activity_main_webview);
        // Enable Javascript
        WebSettings webSettings = mWebView.getSettings();
        webSettings.setJavaScriptEnabled(true);
        webSettings.setAllowFileAccess(true);
        webSettings.setAllowUniversalAccessFromFileURLs(true);
        mWebView.loadUrl("file:///android_asset/www/index.html");
    } 

You can also directly set the view without being dependent on a layout xml file like so:

WebView webview = new WebView(this);
 setContentView(webview);

Troubleshooting
Problem: No resource identifier found for attribute adsize in package com.google…
Solution: http://stackoverflow.com/questions/5819369/error-no-resource-identifier-found-for-attribute-adsize-in-package-com-googl
Problem: Build system is not able to pick up your android sdk directory. This happened to me after I uninstalled android studio and installed just the sdk without the IDE.
Solution: Define ANDROID_HOME environment variable and set it to android SDK directory. Next time you create a project the local.properties file will automatically have a line like so sdk.dir=C:\\Program Files (x86)\\Android\\android-sdk
Problem: How to add reference to google play services library?
Solution: add android.library.reference.1=../../../../Program Files (x86)/Android/android-sdk/extras/google/google_play_services/libproject/google-play-services_lib to project.properties. Note adding an absolute path instead of a relative path gave me an error.
Problem: Where can I define the API Level against which I want my project to be built?
Solution: edit project.properties and define the target like so target=android-21
Problem: What is minSdkVersion and targetSdkVersion in AndroidManifest.xml?
Solution: minSdkVersion specifies the minimum SDK version your app supports. Android will not install your app on devices running a lower SDK version. Google Play Store will list your app as incompatible on devices running a lower SDK. targetSdkVersion specifies the API Level against which you have tested your app. This should logically be the target against which you build your app and which is defined in project.properties
Problem: How to see list of targets available on my machine?
Solution: android list targets
Problem: How to create android project from command line?
Solution: first create empty directory for the project and then run e.g., android create project –target 2 –name adMobExample –path . –activity Main –package com.siddjain.admobexample
Problem: What should I use for the .gitignore file?
Solution: My .gitignore file contains:
bin\
gen\

### References ###

[How to reference a library project in android](http://developer.android.com/tools/projects/projects-eclipse.html#ReferencingLibraryProject)
[Creating and Managing Android Projects from the cmd line](http://developer.android.com/tools/projects/projects-cmdline.html)
[Debugging without eclipse or android studio](http://realmike.org/blog/2010/10/31/android-development-without-eclipse/)
[Another link to debug from cmd line](http://codeseekah.com/2012/02/16/command-line-android-development-debugging/)
[Android Page on Debugging](http://developer.android.com/tools/debugging/index.html)

Recently I had to read a hindi text file in java. I learnt a couple of lessons in the process:
1. First, eclipse kept raising a FileNotFoundException even though the file existed. Turns out if you put the ctor of FileStream in a try-catch the problem goes away.
2. The second lesson was learning how to run java programs w/o eclipse. For this do:
javac Foo.java
java -cp . Foo
3. The third lesson was how to display hindi characters in eclipse. For this see http://paranoid-engineering.blogspot.com/2008/05/getting-unicode-output-in-eclipse.html

Problem: Devise strategy for a computer to play hangman.

Details: The computer needs to guess a word using as few guesses as possible, and make no more than maxWrongGuesses incorrect guesses.
At end of the game, calculate the score for a word to be:

# letter guesses + # number of incorrect word guesses if the computer guessed the word right before making maxWrongGuesses incorrect guesses

or

25 if the computer lost the game before guessing the word correctly.

Solution: The computer starts with a set P of candidate words that initially contains all words from dictionary. Then at any point of time, the computer will guess letter x that will result in discarding maximum # of words from P. The computer finds the answer when only one word is left in P.
Let n = # of words in P
Let G = the computer’s guess so far. To take an example, lets say the word to be guessed is DREAM. When game starts G = —– and P will contain all five letter words in the English dictionary
To calculate the expected cardinality of P, conditional on guessing x, and assuming each word in the dictionary is equally likely to be the secret word that the computer has been asked to guess: 

foreach (var x in C)
{
   int sum = 0;
   foreach (var word in P)
   {
      // calculate cardinality of P that will result if
      // secret word to be guessed is word, and x is guessed
      // by the computer, and G is the guess so far.
      sum += Cardinality(x, word, G);
   }
   e[x] = sum;
}

The computer guesses x that minimizes e over the range of possible characters C to guess from.
When computer has guessed x in a turn, it will never consider guessing x again in subsequent turns (whether it was wrong or right)
i.e., after guessing x, remove x from C, and also revise P appropriately.
If x is a wrong guess, discard all words from P that have x in them;
else if x is a correct guess, update G and filter out words in P that don’t match G.

Implementation: (full code is not shown below)

using System.Collections.Generic;
using System.Diagnostics;
using System.Linq;

namespace Hangman
{
    public class SidsStrategy : GuessingStrategy
    {
        private HashSet<char> letters = new HashSet<char>(new char[] { 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z' });
        private readonly HashSet<string> words = new HashSet<string>();
        private char lastGuess;
        private string lastSofar;

        public SidsStrategy(IEnumerable<string> dictionary)
        {
            foreach (var word in dictionary)
            {
                words.Add(word.ToUpper());
            }
        }

        public Guess nextGuess(HangmanGame game)
        {
            var sofar = game.getGuessedSoFar();
            Filter(game);
            int N = words.Count;
            if (N == 1)
            {
                return new GuessWord(words.First());
            }
            else
            {
                Dictionary<char, double> dict = new Dictionary<char, double>();
                foreach (var ch in letters)
                {
                    int sum = 0;
                    foreach (var word in words)
                    {
                        // calculate cardinality of P assuming secret word to be guessed is word, and we guess ch
                        sum += Cardinality(ch, word, sofar);
                    }
                    dict[ch] = sum;
                }
                lastGuess = ArgMin(dict);
                letters.Remove(lastGuess);
                lastSofar = sofar;
                return new GuessLetter(lastGuess);
            }
        }

        private int Cardinality(char ch, string secretWord, string sofar)
        {
            Debug.Assert(secretWord.Length == sofar.Length);
            char[] array = new char[sofar.Length];
            bool flag = false;
            for (int i = 0; i < sofar.Length; i++)
            {
                if (sofar[i] == HangmanGame.MYSTERY_LETTER &&
                    secretWord[i] == ch)
                {
                    array[i] = ch;
                    flag = true;    // ch will be a correct guess
                }
                else
                {
                    array[i] = sofar[i];
                }
            }
            string postGuess = new string(array);   // if we guess ch, postGuess is what we will end up with
            if (flag)
            {
                int answer = 0;
                foreach(var word in words)
                {
                    if (isEligible(word, postGuess))
                    {
                        answer++;
                    }
                }
                return answer;
            }
            else
            {
                // ch is not present in the secret word
                int answer = 0;
                foreach (var word in words)
                {
                    if (!word.Contains(ch))
                    {
                        answer++;
                    }
                }
                return answer;
            }
        }

        private char ArgMin(Dictionary<char, double> dict)
        {
            double min = double.MaxValue;
            char answer = '0';
            foreach (var kvp in dict)
            {
                if (kvp.Value < min)
                {
                    min = kvp.Value;
                    answer = kvp.Key;
                }
            }
            return answer;
        }

        private void Filter(HangmanGame game)
        {
            var sofar = game.getGuessedSoFar();
            HashSet<string> discard = new HashSet<string>();
            if (sofar == lastSofar)
            {
                // we guessed incorrectly, discard all words that have lastGuess in them
                foreach (var word in words)
                {
                    if (word.Contains(lastGuess))
                    {
                        discard.Add(word);
                    }
                }
            }
            else
            {
                foreach (var word in words)
                {
                    if (!isEligible(word, sofar))
                    {
                        discard.Add(word);
                    }
                }
            }
            foreach (var word in discard)
            {
                words.Remove(word);
            }
        }

        private bool isEligible(string candidate, string sofar)
        {
            if (candidate.Length != sofar.Length)
            {
                return false;
            }
            for (int i = 0; i < candidate.Length; i++)
            {
                var ch1 = candidate[i];
                var ch2 = sofar[i];
                if (ch1 != ch2 && ch2 != HangmanGame.MYSTERY_LETTER)
                {
                    return false;
                }
            }
            return true;
        }
    }
}

Performance: This algorithm turns out to be woefully slow, and cannot be used in any practical scenario.

Revision: Instead of computing expected cardinality conditional on guessing x, we just consider two cases:

Case1: x is a right guess

Case2: x is a wrong guess

f(x) = p₁*n₁ + p₂*n₂

p₁ = probability of case 1 = n₁ / (n₁ + n₂)

p₂ = probability of case 2 = n₂ / (n₁ + n₂)

n₁ = # of words in P that contain x

n₂ = # of words in P that do not contain x

Strategy: Choose x that minimizes f. This algorithm is practical and scores for sample test words are shown below (maxWrongGuesses is set to 5):

COMAKER = 9

CUMULATE = 9

ERUPTIVE = 25

FACTUAL = 7

MONADISM = 6

MUS = 25

NAGGING = 25

OSES = 4

REMEMBERED = 25

SPODUMENES = 7

STEREOISOMERS = 25

TOXICS = 25

TRICHROMATS = 5

TRIOSE = 6

UNIFORMED = 25

MODEL = 8

CAR = 7

RAILWAY = 5

ROAD = 5

TELEPHONE = 4

EXCHANGE = 6

AUTHORITY = 5

COMPLAINTS = 7

FURNITURE = 5

HOUSE = 7

NOTEBOOK = 5

LONG = 6

SHORT = 8

TALL = 25

THIN = 25

FAT = 6

PLAZA = 6

RELOCATIONS = 6

ACCOUNTS = 7

ADDRESS = 7

WEBSITE = 8

AVERAGE SCORE = 11

The problem with this algorithm is that there is no penalty for making a wrong guess. Consider guesses x and y.
x is correct guess, and y is wrong guess.
If both reduce the search space P by equal amounts, both are weighed equally. We would like to choose x over
y in such a case. To factor this in, we make following empirical modification:

f(x) = p₁*n₁ + (1 + α/maxWrongGuesses)*p₂*n₂

The new scores are shown below (α set to 5):

COMAKER = 25

CUMULATE = 10

ERUPTIVE = 6

FACTUAL = 9

MONADISM = 6

MUS = 25

NAGGING = 5

OSES = 4

REMEMBERED = 6

SPODUMENES = 6

STEREOISOMERS = 3

TOXICS = 9

TRICHROMATS = 5

TRIOSE = 10

UNIFORMED = 6

MODEL = 8

CAR = 7

RAILWAY = 7

ROAD = 5

TELEPHONE = 6

EXCHANGE = 6

AUTHORITY = 7

COMPLAINTS = 6

FURNITURE = 6

HOUSE = 7

NOTEBOOK = 6

LONG = 6

SHORT = 8

TALL = 25

THIN = 25

FAT = 6

PLAZA = 6

RELOCATIONS = 9

ACCOUNTS = 6

ADDRESS = 5

WEBSITE = 7

AVERAGE SCORE = 8.58333333333333

An alternative to try if you are one of those who don’t like arbitrary constants like α in the algorithm above, is to set α=0 and if the computer guessed wrong, do not update P. This was actually a bug in my program. I was not updating P in case of a wrong guess, and was using f(x) = p₁*n₁ + p₂*n₂. When I realized the bug, and updated P in case of wrong guess, I found much to my surprise that my scores actually increased by quite a bit instead of decreasing! After much debugging, I realized the need of an α. Anyway the results with α=0, and not updating P in case of a wrong guess are as follows:

COMAKER = 8
CUMULATE = 10
ERUPTIVE = 10
FACTUAL = 8
MONADISM = 8
MUS = 25
NAGGING = 5
OSES = 4
REMEMBERED = 25
SPODUMENES = 9
STEREOISOMERS = 5
TOXICS = 11
TRICHROMATS = 9
TRIOSE = 7
UNIFORMED = 8
MODEL = 10
CAR = 6
RAILWAY = 6
ROAD = 5
TELEPHONE = 6
EXCHANGE = 8
AUTHORITY = 6
COMPLAINTS = 7
FURNITURE = 8
HOUSE = 7
NOTEBOOK = 6
LONG = 8
SHORT = 8
TALL = 6
THIN = 25
FAT = 6
PLAZA = 9
RELOCATIONS = 6
ACCOUNTS = 10
ADDRESS = 8
WEBSITE = 7
AVERAGE SCORE = 8.88888888888889

Download code from https://skydrive.live.com/self.aspx/.Public/Hangman.zip?cid=c1da670a9c473d39&sc=documents

import sys
sys.path.append(‘c:/program files (x86)/google/google_appengine’)
sys.path.append(‘c:/Program Files (x86)/Google/google_appengine/lib/yaml/lib’)
from google.appengine.ext import db
import datetime

class A(db.Model):
date = db.DateProperty()

class B():
date = datetime.date.today()

a = A(date=datetime.date.today())
print a.date.month
b = B()
print b.date.month

prints:
>>10
>>10

the db.Model class does some magic to ensure that the date argument passed to ctor of A is of type datetime.date

Note that a.date is of type datetime.date. It is NOT a DateProperty.

>>> a.date
datetime.date(2011, 10, 2)
>>> b.date
datetime.date(2011, 10, 2)

No one hates CSS more than I do. Some essential tips:

• if you are floating an element, make sure to set its width
• if you set the width of a div in html, it has no effect. the width must be set in css
• say you have a div and set its width to 100%. it means the div will take up 100% of the width of its parent container
• if you set a div’s width, and then apply padding and/or margins to the div – its width will change. to get around this, use the technique of the inner div as explained in stylin’ with css by charles smith

Understanding Z Index
Example 1:

<div id="div1">
    <div id="div2"></div>
</div>

div
{
    position:absolute;
    width:100px;
    height:100px;
}

#div1
{
    z-index:2;
    background:rgba(255,0,0,0.8);
    left:0px;
    top:0px;
}

#div2
{
    z-index:1;
    background:rgba(0,255,0,0.8);
    left:20px;
    top:20px;
}

Result: div2 appears on top of div1 even though its z-index is less than div1
Lesson: Descendants will always appear on top of parent even if their z-index is less than that of parent.

Look at Example 2 vs Example 3 below to understand the difference between z-index:auto and z-index:inherit
Example 2

<div id="div1">
    <div id="div2"></div>
    <div id="div3"></div>
</div>

div
{
    width:100px;
    height:100px;
    position:absolute;
}

#div1
{
    top:0;
    left:0;
    z-index:10;
    background:rgba(255,0,0,0.8);
}

#div2
{
    top:10px;
    left:10px;
    z-index:1;
    background:rgba(0,255,0,0.8);
}

#div3
{
/* z-index is left unspecified */
    top:20px;
    left:20px;
    background:rgba(0,0,255,0.8);
}

Result: div2 appears at the top followed by div3, then div1.
Lesson: if z-index is left unspecified, element shares stacking context of its parent and as seen in example 1 will appear on top of parent.
Contrast above with
Example 3:

#div3
{
    z-index:inherit;
    top:20px;
    left:20px;
    background:rgba(0,0,255,0.8);
}

Lesson: Now div3 will inherit z-index of the parent and in addition establish a new stacking context which will compete and win against stacking context of div2

Example 4:

<div id="div1">
    <div id="div2"></div>
</div>
<div id="div3"></div>

div
{
    width:100px;
    height:100px;
    position:absolute;
}

#div1
{
    top:0;
    left:0;
    z-index:0;
    background:rgba(0,0,255,0.8);
}

#div2
{
    top:10px;
    left:10px;
    z-index:10;
    background:rgba(0,255,0,0.8);
}

#div3
{
    top:20px;
    left:20px;
    z-index:5;
    background:rgba(255,0,0,0.8);
}

div1 and div3 are at same level in the DOM hierarchy. div3 has higher z-index. so it will appear at the top. div1 forms container for div2. by lesson #1, div2 will render on top of div1.

Example 5:
This one is unexpected:

<div id="div0">
<div id="div1">
    <div id="div2"></div>
</div>
<div id="div3"></div>
</div>    

div
{
    width:100px;
    height:100px;
    position:absolute;
}

#div0
{
z-index:69;    
}

#div1
{
    top:0;
    left:0;
    background:rgba(0,0,255,0.8);
}

#div2
{
    top:20px;
    left:20px;
    z-index:10;
    background:rgba(0,255,0,0.8);
}

#div3
{
    top:10px;
    left:10px;
    z-index:5;
    background:rgba(255,0,0,0.8);
}

Result: div2 is at top, followed by div3, then div1.
Expected: div3 to be at the top since it will create a new stacking context whereas div2 will share stacking context of the parent. This example seems to violate what we learnt from Example 2.

Some References:

Z-Index And The CSS Stack: Which Element Displays First?

http://webdesign.tutsplus.com/articles/what-you-may-not-know-about-the-z-index-property–webdesign-16892

1. install python 2.5
2. install GAE SDK
3. download gdata, atom, webapp2 and copy to your application directory (no need to install them, since the production machines won’t have them installed and they will need to pick them up from application dir. However, do install them so that you can write python scripts from IDLE and have them accessible)
4. install git for windows
5. unittest2 (to install: python.exe setup.py install)
6. PIL (if you don’t install PIL, importing testbed will give error. for 64 bit windows you will have to install from uci.edu)
7. install pycharm
i know i am talking to myself in this post