PO file translations - MultipleObjectsReturned error

[byandrej]

I was trying to translate a page using PO editor. But I got error: get() returned more than one String -- it returned 2!
An error was made by one of query selector in wagtail_localize/models.py:

    for index, entry in enumerate(po):
        try:
            string = String.objects.get( …
                locale_id=self.source.locale_id, data=entry.msgid
            )
           ...

with kwargs {'data': 'Lokacija', 'locale_id': 1},
clone:<QuerySet [<String: String object (662)>, <String: String object (663)>

In wagtail_localize_string there were two strings with the same name but one started with a capital letter and the other with a small letter (One is probably title and the other is slug). In DB I found 10 duplicated strings, all of them with same situation.

Modify	id	data_hash	data	locale_id
uredi	662	331eb28504df5b48966cf4b1ec541ba8	Lokacija	1
uredi	663	8d7c8d74ab4f5957ab6a9670d04e6b7c	lokacija	1

Any idea how to solve this?

Django Version: 4.0.6
Python Version: 3.10.4
Wagtail: 2.16.2
wagtail-localize: 1.2.1

[zerolab]

@byandrej what database (and version) are you using?

I would expect String.objects.get(locale_id=1, data="Lokacija") to return just 662 🤔

[chris48s]

@byandrej what database (and version) are you using?

I'm going to go out on a limb and guess: MySQL with a case-insensitive table collation?

[zerolab]

that was my thinking too...
I guess we could change the code to something like:

string = String.objects.filter(
    # ...
    locale_id=self.source.locale_id, data=entry.msgid
).first()

to cater for setups like this, even though I would very much push for a case sensitive collation which should be straightforward to do (via ALTER DATABASE db_name COLLATE collation_name_cs; https://dev.mysql.com/doc/refman/8.0/en/charset-database.html)

[byandrej]

True, I'm using MySQL 5.7 with utf8 case insensitive collection. Better support for utf8 cs was introduced with v8. In v5.7, only option is to change it to utf8_bin, not ideal, but it works.

In the future I suggest that code should be changed to work with _ci db too. But for now, I'll accept this solution.

[Nigel2392]

I have adressed this in the following issue: #758