PostgreSQL + PHP + UTF8 = invalid byte sequence for encoding

Question

I'm migrating a db from mysql to postgresql. The mysql db's default collation is UTF8, postgres is also using UTF8, and I'm encoding the data with pg_escape_string(). For whatever reason however, I'm running into some funky errors about bad encoding:

pg_query() [function.pg-query]: Query failed: ERROR: invalid byte sequence for encoding "UTF8": 0xeb7374 HINT: This error can also happen if the byte sequence does not match the encoding expected by the server, which is controlled by "client"

I've been poking around trying to figure this out, and noticed that php is doing something weird; if a string has only ascii chars in it (eg. "hello"), the encoding is ASCII. If the string contains any non ascii chars, it says the encoding is UTF8 (eg. "Hëllo").

When I use utf8_encode() on strings that are already UTF8, it kills the special chars and makes them all messed up, so.. what can I do to get this to work?

(the exact char hanging it up right now is "�", but instead of just search/replace, i'd like to find a better solution so this kinda problem doesn't happen again)

Answer 1

BTW, an ASCII string is exactly the same in UTF-8 because they share the same first 127 characters; so "Hello" in ASCII is exactly the same as "Hello" in UTF-8, there's no conversion needed.

The collation in the table may be UTF-8 but you may not be fetching information from it in the same encoding. Now if you have trouble with information you give to pg_escape_string it's probably because you're assuming content fetched from MySQL is encoded in UTF-8 while it's not. I suggest you look at this page on MySQL documentation and see the encoding of your connection; you're probably fetching from a table where the collation is UTF-8 but you're connection is something like Latin-1 (where special characters such as çéèêöà etc won't be encoded in UTF-8).

Answer 2

Most likely, the data in your MySQL database isn't UTF8. It's a pretty common scenario. MySQL at least used to not do any proper validation at all on the data, so it accepted anything you threw at it as UTF8 as long as your client claimed it was UTF8. They may have fixed that by now (or not, I don't know if they even consider it a problem), but you may already have incorrectly encoded data in the db. PostgreSQL, of course, performs full validation when you load it, and thus it may fail.

You may want to feed the data through something like iconv that can be set to ignore unknown characters, or transform them to "best guess".