Convert object to DateRange

Question

I'm querying an underlying PostgreSQL database using Pandas 0.8. Pandas returns the DataFrame properly but the underlying timestamp column in my database returns as a generic "object" type in Pandas. As I would eventually like to do a seasonal normalization of my data and I am curious how to convert this generic "object" column into something that is suitable for an analysis.

Here is my current code to retrieve the data:

# get timestamp with time zone Pandas example
import pandas.io.sql as psql
import psycopg2

# define query
QRY = """
    select 
        i i, 
        i * random() f,
        case when random() > 0.5 
        then 
            true 
        else 
            false 
        end b, 
        (current_date - (i*random())::int)::timestamp with time zone tsz 
    from 
        generate_series(1,1000) as s(i)
    order by
        4
    ;
"""
CONN_STRING = "host='localhost' port=5432 dbname='postgres' user='postgres'"

# connect to db
conn = psycopg2.connect(CONN_STRING)

# get some data set index on relid column
df = psql.frame_query(QRY, con=conn)

print "Row count retrieved: %i" % (len(df),)

Result in Python:

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1000 entries, 0 to 999
Data columns:
i      1000  non-null values
f      1000  non-null values
b      1000  non-null values
tsz    1000  non-null values
dtypes: bool(1), float64(1), int64(1), object(1)

Interesting to note that the first column, "i", is an Integer col in PG. I'm not sure why Pandas thinks this is a "bool" type column. My real issue though is the "object" column which I think needs to be of some type of timestamp.

Answer 1

The dtypes listed are in alphabetical order. The tsz column probably contains datetime.datetime Python objects (what is usually returned by database drivers for timestamp columns), have you looked?