Postgresql ON CONFLICT in sqlalchemy

Using Postgres 9.6.1, sqlachemy 1.1.4, and psycopg2 2.6.2:

1. Convert your data structure to a dictionary. From Pandas it is

   import pandas
   from sqlalchemy import MetaData
   from sqlalchemy.dialects.postgresql import insert
   import psycopg2
   
   # The dictionary should include all the values including index values
   insrt_vals = df.to_dict(orient='records')
   

2. Connect to database through sqlalchemy . Instead try psycog2 driver underneath and the native COPY function, which bypasses all the postgres indexing.

   csv_data = os.path.realpath('test.csv')
   con = psycopg2.connect(database = 'db01', user = 'postgres')
   cur = con.cursor()
   cur.execute("\copy stamm_data from '%s' DELIMITER ';' csv header" % csv_data)
   con.commit()
   

3. Execute

   results = engine.execute(do_nothing_stmt)
   # Get number of rows inserted
   rowcount = results.rowcount
   

Warning:

This method does not work with NaTs out of the box.

Everything together

tst_df = pd.DataFrame({'colA':['a','b','c','a','z', 'q'],
              'colB': pd.date_range(end=datetime.datetime.now() , periods=6),
              'colC' : ['a1','b2','c3','a4','z5', 'q6']})


insrt_vals = tst_df.to_dict(orient='records')
engine =      sqlalchemy.create_engine("postgresql://user:password@localhost/postgres")
connect = engine.connect()
meta = MetaData(bind=engine)
meta.reflect(bind=engine)
table = meta.tables['tstbl']
insrt_stmnt = insert(table).values(insrt_vals)

do_nothing_stmt  = insrt_stmnt.on_conflict_do_nothing(index_elements=['colA','colB'])
results = engine.execute(do_nothing_stmt)

Instead of step 2 and 3 , using psycog2 driver with the copy command in postgres is faster for larger files (approaching a gig) because it sets all the table indexing off.

csv_data = os.path.realpath('test.csv')

This works with Postgresql 9.5:

from sqlalchemy.ext.compiler import compiles
from sqlalchemy.sql.expression import Insert

@compiles(Insert)
def prefix_inserts(insert, compiler, **kw):
    return compiler.visit_insert(insert, **kw) + " ON CONFLICT DO NOTHING"

I use it for bulk_insert_mappings. It does however not make ON CONFLICT DO NOTHING optional