Python psycopg2 insert NULL in some rows in postgr

问题:

I have a Python dataframe with NULL value in some rows, while inserting to postgresql, some null in datetype column turns into 'NaT' string or 'NaN', I like it to be a real NULL , which is nothing in that cell.

sample dataframe before insert

import psycopg2
import pandas as pd
import numpy as np

conn=psycopg2.connect(dbname= 'myDB', host='amazonaws.com', 
port= '2222', user= 'mysuser', password= 'mypass')
cur = conn.cursor()

df= pd.DataFrame({ 'zipcode':[1,np.nan,22,88],'city':['A','h','B',np.nan]})


subset = df[['zipcode', 'city']]
data = [tuple(x) for x in subset.values]
records_list_template = ','.join(['%s'] * len(data)) 
insert_query = 'insert into public.MyTable (zipcode, city) values {}'.format(records_list_template)
cur.execute(insert_query, data)
conn.commit()

result in postgresql table

expected result below

回答1:

You can convert NaN to None in this way:

df= pd.DataFrame({
    'zipcode':[1,np.nan,22,88],
    'city':['A','h','B',np.nan],
    'date':['2019-01-01','2019-01-02',pd.NaT,pd.NaT]})

df['date'] = [d.strftime('%Y-%m-%d') if not pd.isnull(d) else None for d in df['date']]

subset = df.where((pd.notnull(df)), None)

See DataFrame.where

回答2:

Convert all instances of NaN in the dataframe by replacing with None, like this:

df = df.replace({pd.np.nan: None})