I'm striving to convert badly encoded data from my table. For instance, I have a field with Nadège which should be Nadège.

I tried using Postgres's functions convert, convert_from, convert_to without much success.

db=# SHOW client_encoding;
(1 row)

db=# SHOW server_encoding;
(1 row)

db=# SELECT "firstName", encode("firstName"::bytea, 'hex') FROM contact;       
 firstName |       encode       
 Nadège    | 4e6164c3a86765
 Nadège   | 4e6164c383c2a86765
(2 rows)

db=# SELECT "firstName", convert_from("firstName"::bytea, 'latin1') FROM contact WHERE "lastName" ILIKE 'crochard';
 firstName |  convert_from  
 Nadège    | Nadège
 Nadège   | NadÃ\u0083¨ge
(2 rows)

db=# SELECT "firstName", convert("firstName"::bytea, 'utf8', 'latin1') FROM contact;                                                                                                                                                       
 firstName |     convert      
 Nadège    | \x4e6164e86765
 Nadège   | \x4e6164c3a86765
(2 rows)

Using python I'm able to get the correct encoding with:


Any hint on how to convert these wrongly encoded data in postgres ?

1 Answer 1


As you have correctly identified, Nadège is the UTF-8 representation of Nadège incorrectly decoded as ISO-8859-1 ("latin-1"). Then, in your case, re-encoded to UTF-8 for storage in the DB.

To fix it you need to:

  • Take the current representation and decode the UTF-8 to latin-1 as a byte string
  • re-interpret the byte string, decoding it as utf-8


test=> SELECT convert_from(convert_to('Nadège', 'latin-1'), 'utf-8');
(1 row)

The Python equivalent would be close to what you wrote, but starts with a unicode representation to illustrate that PostgreSQL stores everything in the database encoding. Something like:

>>> print u"Nadège".encode("latin-1").decode("utf-8")

The problem with all your attempted solutions is that the cast from text to bytea uses the database encoding. So you're starting with the bytes for the utf-8 representation of utf-8 mis-decoded as latin-1. With the cast you'd have to write:

test=> SELECT convert_from(convert_to(convert_from((TEXT 'Nadège')::bytea, 'utf-8'), 'latin-1'), 'utf-8');
(1 row)

because you have to explicitly decode the utf-8 representation produced by the cast before re-interpreting as latin-1 and decoding again.

You just needed to use convert_to(mycol, 'latin-1') instead of mycol::bytea

  • We have solved our encoding issue with perl yesterday, so I cannot try out in a real environment. But your explanation about bytea casting and why we failed to convert seems to be the exact point. I accept your answer, thank you :)
    – Cyrbil
    Dec 17, 2015 at 10:15
  • 1
    The second solution with the ::bytea cast, has one unexpected caveat: Backslashes in the text string need to be escaped (i.e. doubled), since bytea's input function interprets them as special!
    – cstork
    Nov 28, 2020 at 17:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.