Rakesh Rajan's Blog

Change at lifeblob

2010-11-30T17:20:00.002+05:30

I still remember the day when I put in my papers at Trilogy over 3 years back to start lifeblob. Since then, the journey that Pranav, I and the rest of the team took, was nothing short of a roller-coaster ride! We had a fun filled and exciting period working on something that we believed so passionately about. I hope to share some of the insights and my learnings in the future.

Last week lifeblob was acquired by Printo, and the details can be found here. I am quite excited about this step that brings together the largest retail printing player in India and an elegant personalized photo merchandising platform, which results in a great and consistent customer experience when they visit Printo both online and offline.

Post this change, the team has already started work to integrate and quickly launch the platform for both online ( printo.in ) and offline (over 15 stores and increasing) channels.

If you need to reach out to me, you can email me at rakeshxp at gmail dot com or catch me on @xp.

Today, I became an uncle! I have a nephew :)

2010-10-13T00:08:00.001+05:30

Baby boy born today at Cradle. Both sister and nephew are doing great!

Auto Complete with Redis & Java

2010-09-14T23:06:00.002+05:30

Salvatore ( @antirez Redis creator ) recently wrote a nice article on implementing auto completion using Redis Set. Since we use redis at lifeblob and we have been thinking of enabling auto complete for some search options, I came up with a java equivalent. You can grab it from github.

Sample usage:

Note that I am using Jedis java library since that looks like the only java based redis library that supports ZRank.

Reverse Geocoding Using GeoName Data

2008-01-14T02:03:00.002+05:30

I have be using GeoNames data for implementing reverse geocoding.

It took me a while to get it working and thought of sharing the steps that I followed ( Postgres + Postgis ) to get the reverse geocoding work ( and it is fast! )
1) Created and loaded the table by following link or link.

2) Created a geometry column
SELECT AddGeometryColumn( 'public', 'geoname', 'latlon_point', 2163, 'POINT', 2 );
Note that I am using 2163 as SRID ( the unit is meter )

3) Populated the column
update geoname set latlon_point =
transform(GeomFromText('POINT(' || longitude || ' ' || latitude || ')',4326),2163)

4) Created a clustered gist index
CREATE INDEX geoname _latlon_place_index
ON geoname
USING gist
(latlon_point);
ALTER TABLE geoname CLUSTER ON geoname _latlon_place_index;

5) To find nearest 5 records (within 5 kms) for the given lat/long ( 12.97199/77.60483)

SELECT * FROM geoname
WHERE feature_class = 'P'
and st_dwithin(latlon_point, transform(GeomFromText('POINT(77.60483 12.97199)',4326),2163),5000)
order by ST_Distance(latlon_point, transform(GeomFromText('POINT(77.60483 12.97199)',4326),2163))
limit 5

The above query should be extremely fast if the gist index is created properly. One way to confirm whether is it using the index or not, is by looking at the query planner output :

"Limit (cost=9.82..9.82 rows=1 width=175)"
" -> Sort (cost=9.82..9.82 rows=1 width=175)"
" Sort Key: st_distance(latlon_point, '...'::geometry)"
" -> Index Scan using geoname1_latlon_place_index on geoname1 (cost=0.00..9.81 rows=1 width=175)"
" Index Cond: (latlon_point && '..'::geometry)"
" Filter: ((feature_class = 'P'::bpchar) AND (latlon_point && '..'::geometry) AND ('..'::geometry && st_expand(latlon_point, 5000::double precision))

As with any user generated data, there are dirty data present in this dump like duplicate place record, empty place name etc. Also since I am interested only in cities, I added a filter on feature_code = 'PPL' and feature_class = 'P'.

This is just the first pass, and I still need to do couple more processing for cases where the place name contains punctuation, duplicate places across state/country etc. I'm calling it a night! The alarm clock is going to go off in just six hours.

-XP

I'll be there @ BCB5

2007-11-14T03:04:00.000+05:30

LifeBlob is Live !!!

2007-10-19T22:10:00.003+05:30

We are feature complete and passed out invites to our first set of users!

My Timeline
-XP

Running Memcached on Windows

2007-09-18T12:03:00.000+05:30

I was trying to run memcached server on windows as that is my current development OS. Unfortunately the memcached's windows port has given only source code for libevent ( a dependency) and it needs VC to compile :P. With obvious lack ( read: interest ) in VC tools, I set on a big google hunt for finding the windows build of libevent.

I was finally able to get some luck from this site. Thanks to them, I was able to get memcached running successfully.

I have attached the memcached ( 1.2.1 ) setup here

Sets to run the server
1) Unzip the folder to any directory
2) Within the folder, run memcached.exe -d install ( One time )
3) For
starting the server: memcached.exe -d start
stopping the server: memcached.exe -d stop
4) To uninstall the service, run memcached.exe -d uninstall

You can also run memcached.exe -h to find all the properties that can be configured.

-XP

Comic

2007-08-31T13:57:00.000+05:30

Comic

2007-08-23T00:31:00.000+05:30

Excellent Y! Support

2007-08-22T22:25:00.000+05:30

The closure Y! photo's was indeed a bad news. I was totally in love with their interface and obviously unlimited storage:P The day the news came in, I decided to move my photos ( > 2 GB ) quickly ( read: bad judgment ) to flickr. The decision was pretty much based on Flickr's brand name over other options. After the image transfer, I realized that I would have been better off using either Shutterfly/SnapFish service which offers unlimited storage.

But as expected, after doing the photo transfer to Flickr, my Y! photo account got locked. ( It was mentioned in their FAQ that the transfer is a one time only operation). I shot a mail to Y! support ( with the least expectation ) on whether it is possible for them to reactivate my Y! account so that I could migrate to another photo service.

I received a mail from their support within 2 days and was a positive reply :)

Hello,

Thank you for writing to Yahoo! Photos.

Thank you for contacting us regarding your attempted move of your Yahoo! Photos to another service. We are very sorry to hear that you are having problems moving to the affiliate of your choice.

We have released your Yahoo! Photos account so that you can attempt the move again. If you are having continuing problems moving to the same affiliate, you may want to try moving to a different affiliate or simply downloading your images to your computer yourself and then manually uploading them to the affiliate of your choice.

If you have any additional questions or concerns please let us know as soon as possible as we'd be more than happy to help!

- http://help.yahoo.com/l/us/yahoo/photos/

We appreciate your time in writing to us -- your input helps us to identify ways to help make this the easiest and most hassle-free way to transition all of your favorite photos to one of these great services above!

Thank you again for contacting Yahoo! Photos.

Regards,

Jamie Lynn

Yahoo! Photos Customer Care

This kind of support is truly amazing and very much appreciated!

PS: I moved my photos to ShutterFly. My new preference is Google Photos but sadly they don't provide "free" unlimited storage :(

Struts2 + Custom URLs

2007-08-17T03:45:00.000+05:30

For the web application that I was building using Struts, I needed custom ( read: cool ) URLs.

Traditionally, struts URL would be of the format
http://www.myapp.com/login.do
where do is the action extension.

But rather, I wanted a clean URL like http://www.myapp.com/login

To achieve this, I had to do the following

-> In my custom struts.xml, I added a line which overrides the default ActionMapper class

-> Write the custom class which should extend ActionMapper

With a custom Mapper Class, the scope of formats of URL is limited only by imagination :P
To take up an example, lets say we want to have a "clean" search URL
http://www.myapp.com/search/abc+xyz
To achieve the above, we need the following

Action class called Search and corresponding function called getResults(). This class also needs to implement ServletRequestAware to get hold of the ServletRequest ( which would contain the search parameters )

In the custom Action Mapper Class, we would need to use regex ( split on "/" ) to understand the URL and if we find the first token is "search", we could set

actionMapping.setNamespace(namespace);
actionMapping.setName("search");
actionMapping.setMethod("getResults");
request.setAttribute(SEARCH_KEY, searchParameters);

-XP

MS Live Search WTF

2007-04-03T15:51:00.000+05:30

Check out the results for "how long does it take to get a patent" on MS Live Search ( roftl ).

Looks like the live team is doing a great job in catching up google :P

Funny Traffic Signal

2007-02-21T13:10:00.000+05:30

Encountered a weired traffic signal near RajajiNagar Entrance. Was wondering what to do next: "Jump" the signal or face the wrath of the vehicles behind me :P

PS: I had taken the pic using my phone ( Nokia 6270 ) without flash and while moving and hence the bad quality ;)

-XP

Family Tree

2007-02-21T01:24:00.000+05:30

I am very bad at remembering detailed family relations and for that matter even names :P When ever I visit my native place (which is very rare) with my parents, my parents/I make sure that I get a "crash" course on all the relations again ;). So being a tech guy, I wanted a tech solution and hence entered the domain of family tree. For long I had be wanting to find a site that would really simplify creation of the family tree. I had tried various sites like Ancestry, TribalPages etc, but none of them allowed me an "easy" or "flexible" way to create a family tree :( Couple of hrs back, Pavan ( Friend and Colleague ), pinged me with the URL to this brand new "web 2.0 family tree"!

It is called Geni( That is one short name!). The first thing that strike me was with the ease it allowed me to create an account ( No email validation to start off! But then we need to validate later). The UI is flash based and mind blowing. The whole experience of creating family tree is taken to the next level. It was so much of fun that within 30 mins or so, I had added over 20+ members to my tree. The product is still in "beta" and has couple of privacy issues to be answered ( Like mother's maiden name being shown, access control etc). But in all, I really liked the whole user experience. I recommend that you take a shot at this.

A sample tree

I also manage to find couple of other new Family Tree service

But none the above has an UI that is as intuitive as the Geni's one

Update

Check out the way they incite users to complete the profile.

-XP

Eclipse Callisto

2007-02-20T16:52:00.000+05:30

I had been using Eclipse 3.1 for a long time. So thought I should give 3.2 a shot at. Eclipse site mentioned about "Callisto".

In their words

Callisto is about improving the productivity of the developers working on top of Eclipse frameworks by providing a more transparent and predictable development cycle. By releasing 10 projects at the same time, the goal is to eliminate uncertainty about version compatibility and make it easier to incorporate multiple projects into your environment.

The best part I like was the customization that it allows to choose features from various projects. Further googling landed me on Yoxos. They provide an excellent UI ( Rich AJAX Platform )create a custom Eclipse installation.

Allows to create/share custom scenarios ( Eclipse installation)
Allows to pick standard features and 3rd party plugins.
Provides Yoxos Install Manager (YIM) which is an eclipse based update mechanism.
Provides 3 months free subscription. After that YIM will not work.

I created a custom ( warning: 500 MB+ ) Eclipse Installation ( I added every plugin that I could think of:P ). It is accessible out here.

Screenshots of features for my custom Eclipse

-XP

Resolutions

2007-02-18T22:15:00.000+05:30

Have been thinking a lot to make life more exciting. Decided to "try" to do 2 things from now on

1) Wake up early and define a morning ritual ( Thanks to this nice read by Steve Pavlina).
I have been very erratic w.r.t to sleeping. These days I sleep at 2-3 and wake up at 11. I end up reaching office only around 2-3 :( Also alarm never helps :P. Need to bring some order back into my life. So decide to start waking up around 6. Yoga sounds a good idea.

2) Start blogging ;)
Set up my blogger account's template, widgets etc, yest night. Still not very happy about the template. Will change once I get hold of a nice template ( I really liked this page look )

-XP

Simility Launched !!!

2005-10-15T22:37:00.000+05:30

Simility Beta has finally been launched!

-> simility is the best way to find similar content on the web.
-> simility toolbar recommends web pages that are similar to the page you are currntly viewing.

simility is easy

The simility toobar is the fast, easy way to find great new content on the web. With just the click of the "recommend" button, we suggest new pages similar to the page you are currently viewing.

Better than search

simility is better than search engines because the recommended pages are generated from other users. We only show you recommendations of sites that other users find interesting and helpful, so you don't need to spend time paging through hundreds of search results.

WikiPedia Article

BikeWala Launched

2005-10-01T01:35:00.000+05:30

One of our TU project : BikeWala has been launched.
It is one of the best places to buy motorbikes in India. It was launched 2 days back. So far it has got over 1000 hits!!! Have a look at the site

Google Map

2005-04-07T03:25:00.000+05:30

I was amazed at the way Google Maps works! It is way faster than Yahoo or MSN or MapQuest ( I think www.map.search.ch is equally fast, i like the zooming effect )

Google Maps uses two built-in browser components:
-> XMLHttpRequest ( very famous, used in Google Suggest) and
-> XSLTProcessor

Google Map communicates with the server to get the map tiles and search results
It basically gets 3 types of images from the server ( based on the selection)

1) JPEG Image in case of satellite image
It sends a GET request to Keyhole Server 2.4 at kh.google.com
Sample Get request
http://kh.google.com/kh?v=1&t=tqtrtsttrqqtrtq

Value of
v = No idea, seems to work with different values, maybe for some internal usage
t = direction to be taken to get the desired JPEG image tile

Q	R
T	S

-> t can have maximum 15 characters ( i.e 15 zoom levels)
-> suppose intially t = tqt ( so it will fetch 4 tiles, for top-left, top-right, bottom-left, bottom-right ) next time if t = tqtt , it means fetch the bottom-left tile and zoom it ( ie it will again fetch 4 more tiles )

2) GIF Image in case of map image
It sends a GET request to Keyhole Server 2.4 at mt.google.com
Sample Get request
http://mt.google.com/mt?v=.3&x=-2360&y=-1033&zoom=3
where,
v = No Idea
x = X coordinate of the tile to be fetched
y = Y coordinate of the tile to be fetched
zoom = zoom level ( 1 – 5 )
So it is possible to fetch the entire image database of Google maps, if we write a script which will fetch all image tiles for a given zoom level but obviously it is copyrighted!

3) PNG Image is returned which is basically the route between 2 locations (It is a white image with the route displayed in blue) and it is superimposed onto the image tiles.

It is again a GET request to Geocode/Map Server

Sample request:
http://www.google.com/maplinedraw?width=144&height=260&path={F{MGBGKZONVPVr@hA@@r@f@v@j@JFJJ?RQN?V?B?L?R@@?L?x@?ZA??JO?@LH??X}oR

where,
width/height = width/height of the resulting PNG file.
Path = yet to figure out ( i think the encoded string should contain the source and destination location ???? )

When we search for any city, say New York, it sends a HTTP GET request to mfe server. e.g.
http://maps.google.com/maps?q=new+york&btnG=Search&sll=33.748889%2C-84.388056&sspn=0.102539%2C0.230971&z=5&t=&f=q&output=js&hl=en

where,
q = name of the location
sll = latitude & longitude
sspn = span size
z = zoom level
the output of such a request is XML

<?xml version="1.0"?>
<page>
<title>new york</title>
<query>new york</query>
<request>
<url>
http://maps.google.com/maps?q=new+york&
btnG=Search&
sll=33.748889%2C-84.388056&
sspn=0.102539%2C0.230971&amp;amp;amp;
z=5&
t=&
f=q&
hl=en&
num=10
</url>
<query>new york</query>
</request>
<center lat="40.714167" lng="-74.006389"/>
<span lat="0.089988" lng="-0.118722"/>
<overlay panelStyle="/maps?file=gp&hl=en">
<location infoStyle="/maps?file=gi&hl=en" id="A">
<point lat="40.714167" lng="-74.006389"/>
<icon class="noicon"/>
<info>
<address>
<line>
New York, NY
</line>
</address>
</info>
</location>
</overlay>
</page>

Location of XSL files
1) panelStyle = http://maps.google.com/maps?file=gp&hl=en
2) infoStyle = http://maps.google.com/maps?file=gi&hl=en

I read somewhere that this kind of apps are called AJAX ( Advanced Javascript & XML )

I personally think that Google might have pre-calculated routes between all locations ( if not all, atleast between important places). So when we try to find route between, say, A to B, Google would have precomputed it and would be in memory. So it can retrieve the path and send it back to the client.

One addition what Google could probably add is Traffic so that route displayed is the route with least traffic ( As I assume they have precomputed routes and put it in memory, assume 5 routes are there between A to B, all it has to do is to add weight ( traffic density) at run-time to all the 5 routes and return the route with the least traffic.

This is all I could manage to figure out in 2 hrs. Will post, once I find out more .