Pandas DataFrame is a very useful data type in Python to format and order data in different ways. One of its use is to write Dataframe to a file and read back to a DataFrame when it is needed.

1. DataFrame to CSV

In this example, we will write DataFrame into a CSV file

import pandas as pd

d = {'Name': ['John Doe', 'Foo', 'Tony Stark'], 'Age': [10, 12, 48]}
df = pd.DataFrame(data=d)

df.to_csv('test.csv', index=False)

This will create a file called test.csv in the directory, you have run the python code.

Note that here we have provided index=False so it will not provide the index as a column. If you want the index you can just run to_csv function without params

df.to_csv('test.csv')

You can read this CSV by using the read_csv function.

import pandas as pd

dfr = pd.read_csv('test.csv')
print(dfr)

This will print the data in the console as a formated table

When you have data with commas, you can change the seperator to a different one using sep= parameter. Here I have used | pipe symbol and following is the result

import pandas as pd

d = {'Name': ['John Doe', 'Foo', 'Tony Stark'], 'Age': [10, 12, 48]}
df = pd.DataFrame(data=d)

df.to_csv('test.csv', index=False, sep="|")

dfr = pd.read_csv('test.csv')
print(dfr)

---

2. Dataframe to JSON

When saving JSON, we have to provide the rows, columns, and indexes as well if you need to guarantee order of data.

For that, we can provide orient parameter when converting to JSON.

import pandas as pd

d = {'Name': ['John Doe', 'Foo', 'Tony Stark'], 'Age': [10, 12, 48]}
df = pd.DataFrame(data=d)

df.to_json('test.json', orient='split')

dfr = pd.read_json('test.json', orient="split")
print(dfr)

This saves the test.json file in the following format

You can see the three main attributes columns, index, data defines how the data should be in the DataFrame.

There are other orients that you can save the DataFrame to JSON files

orient=records

Array of records as each row. No loss as the column title is available. Takes space as column title is always duplicated

orient=columns

Each column data is represented by the index

orient=values

Array of each rows of data as arrays. Similar to records but data row represented as an array. Saves space but column header is missing.

orient=table

Much more details as a table structure. Data type and pandas version all reserved

Following are some useful References to write Dataframe to a file and read back

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.to_json.html

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.to_csv.html