psycopg2.DataError: invalid byte sequence for encoding "UTF8": 0xa0

Question

I have done quite a bit of googling on this error and have boiled it down to the fact that the databases I am working with are in different encodings.

The AIX server I am working with is running

psql 8.2.4

 server_encoding                 | LATIN1          |      | Client Connection Defaults / Locale and Formatting                | Sets the server (database) character set encoding.

The windows 2008 R2 server I am working with is running

psql (9.3.4)

CREATE DATABASE postgres
WITH OWNER = postgres
   ENCODING = 'UTF8'
   TABLESPACE = pg_default
   LC_COLLATE = 'English_Australia.1252'
   LC_CTYPE = 'English_Australia.1252'
   CONNECTION LIMIT = -1;

COMMENT ON DATABASE postgres
IS 'default administrative connection database';

Now when i try execute my below python script I get this error

Traceback (most recent call last):
File "datamain.py", line 39, in <module>
sys.exit(main())
File "datamain.py", line 33, in main
write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
File "datamain.py", line 21, in write_file_to_table
cur.copy_from(f, table, ",")
psycopg2.DataError: invalid byte sequence for encoding "UTF8": 0xa0
CONTEXT:  COPY cms_jobdef, line 15209    

Here is my script

import psycopg2
import StringIO
import sys
import pdb

def connect_db(db, usr, pw, hst, prt):
    conn = psycopg2.connect(database=db, user=usr,
    password=pw, host=hst, port=prt)
    return conn

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    cur = connection.cursor()
    cur.copy_to(f, table, ",")
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor()
    cur.copy_from(f, table, ",")
    f.close()
    cur.close()

def main():
    login = open('login.txt','r')
    con_tctmsv64 = connect_db("x", "y",
    login.readline().strip(),
    "d.domain", "c")
    con_S104838 = connect_db("x", "y", "z", "a", "b")
    try:
        write_table_to_file("cms_jobdef.txt", "cms_jobdef", con_tctmsv64)
        write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
    finally:
        con_tctmsv64.close()
        con_S104838.close()

if __name__ == "__main__":
    sys.exit(main())

have removed some sensitive data.

So I'm not sure how I can proceed. As far as I can tell the copy_expert method might help by exporting as a UTF8 encoding. But because the server I am pulling the data from is running 8.2.4 I Dont think it supports COPY encoding format.

I think my best shot is to try and reinstall the postgre database with an encoding of LATIN1 on the windows server. When I try and do that I get the below error.

So im quite stuck,any help would be greatly appreciated!

Update I installed the postgre db on the windows as LATIN1 encoding by changing the default local to 'C'. This however gave me the below error and doesnt seem like a likely successful/correct approach

I have also tried encoding the files in BINARY using the PSQL COPY function

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    cur = connection.cursor()
    #cur.copy_to(f, table, ",")
    cur.copy_expert("COPY cms_jobdef TO STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor()
    #cur.copy_from(f, table)
    cur.copy_expert("COPY cms_jobdef FROM STDOUT WITH BINARY", f)
    f.close()
    cur.close()

Still no luck I get the same error

DataError: invalid byte sequence for encoding "UTF8": 0xa0
CONTEXT:  COPY cms_jobdef, line 15209, column descript

In relation to Phils answer I have tried this approach with still no success.

import psycopg2
import StringIO
import sys
import pdb
import codecs

def connect_db(db, usr, pw, hst, prt):
    conn = psycopg2.connect(database=db, user=usr,
    password=pw, host=hst, port=prt)
    return conn

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    #fx = codecs.EncodedFile(f,"LATIN1", "UTF8")
    cur = connection.cursor()
    cur.execute("SHOW client_encoding;")
    print cur.fetchone()
    cur.copy_to(f, table)
    #cur.copy_expert("COPY cms_jobdef TO STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor() 
    cur.execute("SET CLIENT_ENCODING TO 'LATIN1';")
    cur.execute("SHOW client_encoding;")
    print cur.fetchone()
    cur.copy_from(f, table)
    #cur.copy_expert("COPY cms_jobdef FROM STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def main(): 
    login = open('login.txt','r')
    con_tctmsv64 = connect_db("x", "y",
    login.readline().strip(),
    "ctmtest1.int.corp.sun", "5436")
    con_S104838 = connect_db("x", "y", "z", "t", "5432")
    try:
        write_table_to_file("cms_jobdef.txt", "cms_jobdef", con_tctmsv64)
        write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
    finally:
        con_tctmsv64.close()
        con_S104838.close()

if __name__ == "__main__":
    sys.exit(main())

output

In [4]: %run datamain.py
('sql_ascii',)
('LATIN1',)

In [5]: 

This completes successfully but when i run a

select * from cms_jobdef;

Nothing is in the new database

I have even tried converting the file format from LATIN1 to UTF8. Still no luck

The weird thing is when I do this process manually by only using the postgre COPY function it works. I have no idea why. Once again any help would be greatly appreciated.

Solution 1

Turns out there are a few options to solve this problem.

The option to change the clients encoding suggested by Phil does work.

cur.execute("SET CLIENT_ENCODING TO 'LATIN1';")

Another option is to convert the data in the fly. I used a python module called codecs to do this.

f = open(file, "w")
fx = codecs.EncodedFile(f,"LATIN1", "UTF8")
cur = connection.cursor()
cur.execute("SHOW client_encoding;")
print cur.fetchone()
cur.copy_to(fx, table)

The key line being

fx = codecs.EncodedFile(f,"LATIN1", "UTF8")

My main problem was that I was not committing my changes to the database! Silly me :)

Solution 2

I'm in the process of migrating from an SQL_ASCII database to a UTF8 database, and ran into the same problem. Based on this answer, I simply added this statement to the start of my import script:

set client_encoding to 'latin1'

and everything appears to have imported correctly.