Textoo Pro ; in the works

May 16th

Posted by Kiran Kumar in Projects where I used Titanium framework

July 17th 2011 : Textoo Pro now in the apple app store. Download it here : http://textoo.in

July 10th 2011 : Textoo v2.0 renamed to Textoo Pro

Release of Textoo Pro has been postponed to July 15th 2011 due to app store codesign issues.

Here’s a sneak peek into the brand new version of Textoo Pro – the iOS app which lets you send free real SMS to any mobile phone in India and other countries.
Based on the numerous feedback emails we have got, we have tried our best to incorporate all the enhancements into this version.

A brand new custom designed UI with retina support
With an intent to keep the UI similar on both iOS and android versions, the entire UI was rewritten by avoiding the use the standard components provided by apple SDK as much as possible. The UI design was inspired by Apple’s iBook application and each page was designed to work well on both portrait and landscape views with support for retina and non retina displays. Animations and transitions were added to boost the look and feel of the app.

Native support for group messages and multi select contacts Continue reading “Textoo Pro ; in the works” »

Incoming search terms:

android, appcelerator titanium UI framework, Auto, automatic message, character limit, contact, contacts groups titanium, Draft, feedback, feedback emails, fix bugs textoo, gatew, gatewa, group messages, iBook, iOS, landscape, landscape views, message thread, message threads, messaging app using titanium, multi select contacts in titanium, new version of textoo, page, peek, phone, pretttyUI, problem in textoo fixed, progressbar, ProRelease, Queuing, retina, retina images @2x titanium, sdk, selector, sneak peek, store, textoo pro 1 1 ipa, textoo release date july 1st 2011, threaded views, ui design, unsent messages, use, version

What’s my twitter reputation score ?

Dec 4th

Posted by Kiran Kumar in Uncategorized

1 comment

Last month, Twitter founder Evan Williams announced during the Web 2.0 summit that Twitter.com internally keeps a record of every users reputation score. This made me wonder how do they calculate each users reputation mathematically. After some thought, I charted out a few parameters based on which they could possibly arrive at a quantitative measure of a users popularity on twitter. Here are some of my conclusions,

Number of followers: This factor contributes heavily to the score. More the followers, more is ones reputation. But its not just the number which decides the reputation. Each follower has his own reputation on twitter which will contribute to the user he follows. If a users reputation was only dependant on his followers, then a recursive algorithm would be required to accurately determine one’s reputation. Again the problem will arise when there are cyclic links, i.e. you are followed by a follower who is being followed by someone whom you are following, which will make it an infinite cycle. So the algorithm should stop the branching process as soon as cyclic links are detected. But most probably twitter.com would decide on the depth of penetration for each calculation beforehand.

Also there are certain users who back follow other users who are following them. Such cases will give rise to high rates of cyclic links and hence should not contribute much to the reputation calculation.

Number of retweets: You are definitely reputed if a significant number of users retweet your tweets regularly. So number of retweets coupled with the individual reputation of the user who is retweeting your tweet will add to your reputation score.

Number of mentions: If someone with a high individual reputation @mentions you, then your reputation score should increase.

mentions, favourites, listings etc….. todo

In the meantime checkout http://www.twitter-reputation.com. It is a small experimental page which I set up for finding out your own twitter reputation score based on the parameters listed above. I used the twitter @anywhere javascript API inspite of its limitations , instead of using the more stable PHP api, just to minimize server side bandwidth being used up for fetching user info.

Incoming search terms:

algorithm, API, Auto, branching process, calculate twitter reputation, calculation, com, Draft, evan williams, follower, founder, how reputed are you on twitter.com, infinite cycle, inspite, meantime, meantime checkout, measure, month, quantitative measure, record, recursive algorithm, reputation, reputation score, reputation score on twitter, rise, score, score number, summit, todo, tweeprank, tweet, tweets, tweputation, twitrep, twitreput, twitter, twitter popularity index, twitter reputation score, web, web2.0 summit

Java resource bundle modification to identify hardcoded strings

Nov 13th

Posted by Kiran Kumar in [...]

1 comment

In Java, Internationalization of applications is supported through the ResourceBundle class. The strings which need to be translated are packaged in individual resource bundles and each language will have its own resource bundle which is loaded by the ResourceBundle Class based on the application logic.

I was involved in the modification of the Java ResourceBundle class to make it optionally decorate all the strings coming from any specific resource bundle with certain special characters like {{ or [[ so that it will be possible to visually identify any hard coded strings in the application. My role in this project was to specifically hack the XliffResourceBundle class which decorates the strings from Xliff resource bundle files. The decoration process had to be turned off or on by the user and had to obey certain rules/filters which allows the user to decorate specific strings (exclude any numbers/dates etc).
As all strings which support internationalization typically use the resource bundle class to fetch the translated strings, modifying the class itself at the root level will help one to indentify the ones which are not using this class (and hence hardcoded).
But one might ask “why not write a tool to scan through hardcoded strings using regex”. Yes , you could do this, but there are cases where this method wont be sufficient. The strings could still be following your regex pattern and might not be using the resourcebundle class to fetch the translated strings. Only modifying resource nundle class can provide a foolproof way to identify them.

Why ?
Identifying hardcoded strings is very much necessary for any software application which has to be internationalized. Especially when the software is used all over the world. Typically Arab speaking and Chinese users contribute to a major percentage of profits and they prefer localized versions of the user interface. Localizing the UI after a software has become successful will take minimal effort and the results of such a move are quite astounding.

But usually during inital phases of development, no developer will give much importance to externalizing string and coding in a way that makes localization of each UI easier. The tendency is to concentrate on the functionality of the software and make things work well. The problem comes when the functionality breaks during the localiztion process. Also from the profits point of view, localizing the software becomes important. But to the developer, who feels that putting additional efforts to just make the software work under different locales and obey the rules of different regions in showing the dates/currency and timezone implications , it is something of least interest. They are the developers who just copy past code samples from the internet and never give importance to how things work. Such negligence gives rise to performance issues too. But internationalizing existing programs also become difficult is this is the case.