There are probably a lot of approaches, but something like this works to get minute data for 5 business days (ie trading days) and will not be affected by any half days.

import pandas as pd  
previous_days = 5

# There are 6.5 hrs in a trading day. Get an extra day to be safe.  
max_minutes = 6.5 * 60 * (previous_days  + 1)  
price_hist = data.history(security, 'price', max_minutes ,'1m')  


# Normalize the times to be all 00 hours  
dates = price_hist.index.normalize()


# Find unique dates. Normalize the datetimes first so all the times begin at 00:00  
unique_dates = pd.unique(dates)

# Get date 'previous_days' ago. Note this is trade days not calendar days.  
date_5_trading_days_ago = unique_dates[-5]


# Then simply use the loc method to select only those dates after date_5_trading_days_ago  
price_hist.loc[date_5_trading_days_ago : ]

That will select all the minutes of each day for the past 5 trading days whether a half day or not.

Now, for the second question, the simplistic method is to assume there are 5 trading days for every 7 calendar days. If one wants 90 calendar days then select 5/7 x 90 ~ 65 trading days.

price_hist = data.history(security, 'price', 65 ,'1d')

If one truly wants 90 calendar days then something like this.

import pandas as pd  
previous_days = 90

# Get a little more data than we need  
price_hist = data.history(security, 'price', previous_days+1 ,'1d')  


# Find unique dates  
unique_dates = pd.unique(price_hist.index)


# Get the last date  
last_date = unique_dates[-1]


# Get date 90 days ago. Note this is now calendar days.  
date_90_trading_days_ago = last_date - pd.Timedelta(days=previous_days)


# Then simply use the loc method to select only those dates after date_90_trading_days_ago  
prices_last_90_calendar_days = price_hist.loc[date_90_trading_days_ago : ]

See attached notebook for ideas.

Hi Dan,

Thanks a ton for your detailed and educating answer. I have one (perhaps noob level) query for the line:

# Find unique dates  
unique_dates = pd.unique(dates)  

What is dates here? I am asking this because it is not a literal (like a string), and not a variable too in your code.

Thanks.
...- anil

@ Anil Bhatnagar

Oops, sorry had the wrong variable name. I had 'price_dates' in the original. I updated the post above. Basically, 'dates' is just the index of all the datetimes in 'price_hist'. It's normalized so all the time data is changed to midnight 00:00. This effectively gives just the dates of each data point.

dates = price_hist.index.normalize()