I have a dataframe such as:

and I would like to sort the columns in a certain order. I tried using sort_values and passing a list with the order I want them in but it is not recognizing '2010-01-01' or any other dates in the list. I assume it is because they are in date format, so how would I sort these labels?

What I currently am trying:

order = ['more data', 'data', '2011-01-01', '2010-01-01']
df = df.sort_values(order)

CodePudding user response：

I'm not pretty sure how you want to handle non-datetime columns. I assume that you prefer these columns to show at the beginning of the dataframe. With that in mind, I suggest you use sorted function:

import time
import datetime
def sort_date(value):
  try:
    return -1 * time.mktime(datetime.datetime.strptime(value, "%Y-%m-%d").timetuple())
  except:
    return float("-inf")
columns = ['data', '2011-01-01', "value", '2010-01-01']

sorted(columns, key=sort_date)

The abovementioned code has nothing to do with pandas dataframe - rather, gives you a better idea of how to handle such a situation. The sorted function mentioned returns the following output:

['data', 'value', '2011-01-01', '2010-01-01']

Therefore, you can simply use the returned value of the abovementioned code to sort your columns out:

df[sorted(columns, key=sort_date)]

Note that you can access the columns of the dataframe by simply using df.columns.

CodePudding user response：

If you only want to re-sort the date columns and not any of the others you have to write a bit more code. Something like:

# Separate date columns from non-date columns
non_date_cols = ['name']
date_cols = [col for col in df.columns if col not in non_date_cols] # Get a list of the date column names

# Sort the column names (not the actual columns)
date_cols.sort()

# Rebuild your dataframe by taking the non date cols and joining the rest of the columns in the order specified
df_sorted = df[non_date_cols   date_cols]