Pachmarhi visit

Some random place

Rajat prapat

Towards chauragarh

An exercise in simplification

I recently needed to solve a problem:

Problem
There is a file which contains pairs of user ids and timestamps. Each line contains one pair. user ids are sorted in ascending order, and for each user id, timestamps are sorted in ascending order. Here is an example file:

user1    2012-09-18-14:30:29
user1    2012-09-18-14:30:55
user1    2012-09-18-14:31:12
user1    2012-09-18-15:31:12
user1    2012-09-18-15:31:12
user1    2012-09-18-16:31:12
user2    2012-09-19-14:30:29
user2    2012-09-19-14:30:55
user2    2012-09-19-14:31:12
user2    2012-09-19-15:31:12
user2    2012-09-19-15:31:13
user2    2012-09-19-15:31:16

For each user, we need to output various session lengths. A session is defined as a sequence of consecutive visits so that no two visits are more than 30 minutes apart. We need not output the user ids.

For instance, in above example, user1 has three sessions: consisting of first three visits, next two visits and the next visit.

Output for above file would be:

3, 2, 1
3, 3

First cut
Initial thought is to

• Scan through the file checking when the user id changes
• For a given user id, keep checking if next timestamp is more than 30 minutes after the previous one.

Here is the python code embodying this idea:

from datetime import datetime, timedelta

NEW = 'new'
def are_close(t1, t2):
    fmt = '%Y-%m-%d-%H:%M:%S'
    diff = datetime.strptime(t2, fmt) - datetime.strptime(t1, fmt)                              
    return diff < timedelta(minutes = 30)                                                       

def write_details(sessions):                                                                    
    print ', '.join([str(len(session)) for session in sessions])                                
        
def write_one_id_details(viz_id, page_views):
    if viz_id == NEW:                                                                           
        return
    if len(page_views) == 0:                                                                    
        return
    sessions = []
    session = [page_views[0]]
    if len(page_views) == 1:
        sessions.append(session)                                                                
        write_details(session)                                                                  
        return
    for i in range(1, len(page_views)):                                                         
        if are_close(page_views[i - 1], page_views[i]):
            session.append(page_views[i])
        else:
            sessions.append(session)
            session = [page_views[i]]                                                           
    sessions.append(session)                                                                    
    write_details(sessions)
                                                                                                
def process():
    f = open('visits.txt')
    curr_viz_id = NEW
    page_views = []
    for line in f:
        viz_id, time = line.strip().split('\t')                                                 
        if viz_id != curr_viz_id:                                                               
            write_one_id_details(curr_viz_id, page_views)                                       
            curr_viz_id = viz_id                                                                
            page_views = [] 
        page_views.append(time)
    write_one_id_details(curr_viz_id, page_views)
    f.close()

if __name__ == '__main__':
    process()

While this works, there are several problems with this code:

• We have a variable called NEW to boostrap the process
• In write_one_id_details(), we need to initialize sessions to [page_views[0]]. This is hindrance in achieving harmony.
• When a loop ranges from 1 to n - 1, I do not like it. Loops should range from 0< to n - 1.
• sessions = [] or page_views = [] are indications of procedural programming. I would rather want easy to understand declarative style

 

Just Chill Chill

I just heard the song Just Chill Chill Just Chill.
I think I am hearing it for the first time in India.
I heard it last in Cbhatt's Dabba!

When once I got lost on 237 (when I was a new driver),
this was the song playing as I was getting frustrated
over losing my way and having low gas in Cbhatt's
dabba! I tried initially to get motivated by the song,
but soon gave up and shut up the song.

http://www.youtube.com/watch?v=1g_6TmOyXAM
Just Chill Chill,
--Pandey

Impossible triangles

In Macroeconomics, there is an impossible triangle: You cannot simultaneously control

1. Interest rate
2. Inflation rate
3. Exchange rate

If you try to tackle two of them, the third one is automatically determined. Let us see how:

Suppose RBI wants a certain interest rate and a certain rate of inflation. To have a fixed interest rate, it will need to buy/sell bonds. That will increase/decrease money supply in the economy and lead to some inflation. To control inflation, RBI needs to make sure money supply in the system is same as previously. So, it would buy/sell forex. That would lead to an exchange rate over which RBI does not have control.

Similar reasoning can be applied for other two cases.

Even at VMware, there was an impossible triangle. There was tradeoff between

1. Number of hidden page faults
2. Context switch cost
3. (I forgot what it was.)

Of course, there are two way "On the one hand x, on the other hand y" types of things, but triangles are more interesting!

Cypress Talk

Today, there was pre placement talk by Cypress Semiconductors. The person who presented (his name was Pashupathy, I think) was a VP at Cypress. He had worked around 14 years at Cypress in San Jose and then returned to India to start the Chennai office.
He seemed very excited about his job. He told us that Chennai office was not just a BPO/KPO type of thing. It handled things that were important for Cypress. He mostly seemed to handle investments, M&A etc. for Cypress.
This looks like a nice company to work for. However, given my background, if I decide to go to a Tech company, I would rather go for yahoo/google/microsoft etc. which are software companies.
It is a complicated tradeoff to make between earning a lot of money and having a relatively cool life. (Work is interesting in any case, I believe)

अजीब दुनिया

ये दुनिया अजीब है:

1. अगर आप बैंक से पैसे लेने जाएं, तो अगर आप की credit history अच्छी है तो आपको कम ब्याज पर पैसे मिल जायेंगे, और अगर ख़राब है तो आपको ज्यादा ब्याज देना पड़ेगा। मतलब ये कि गरीब आदमी ज्यादा ब्याज देगा और अमीर को सस्ते में उधार मिलेगा। वाह!

2. चलो आपने credit history देखकर ब्याज की दर निर्धारित कर ली। ब्याज की दर निर्धारित करते समय आपने ये ध्यान रखा था कि कुछ लोग default कर जायेंगे। मतलब ये कि जो लोग default करेंगे उनको जो पैसे देने चाहिए थे, वो आपने उन लोगों से ले लिए जिन्होंने ने ईमानदारी से पैसे दिए। मतलब ईमानदार लोग बेईमान (या बदकिस्मत) लोगों की भरपाई करें। अगर आप पैसे लेकर भाग गए टू भाग गए, मगर अगर आपने वापस किए तो जो भागे हैं, उनकी भी भरपाई करें. वाह!

कुल मिला कर शायद system का optimal configuration यही है, मगर थोड़ा अजीब लगता है.

Quotes

1. Play every game for all you're worth, but when the game is over, it is over.
i.e. don't dwell on past gains and loses.