Say I have the below 2 tables:
text
id
text_lang
id
media_id
lang
text
(these are not the actual table structures - but I just have 2 tables with one containing all the localized strings that "belong" to the first table)
I want to join the tables, with a "preferred" language. So basically:
SELECT text.id, text_lang.text
FROM text
LEFT JOIN text_lang ON text.id = text_lang.text_id
AND (only a single row, and if there's a row with lang="en", pick that,
otherwise just pick any language)
The part I'm having trouble with is the "only a single row, and if there's a row with lang="en" pick that, otherwise just pick any language".
I think this does what I'm looking for:
SELECT text.id, text_lang.text
FROM text
JOIN text_lang ON text_lang.text_id = text.id AND
text_lang.lang = (select lang
FROM text_lang sub_text_lang
WHERE sub_text_lang.text_id = text.id
ORDER BY text_lang.lang = "en" DESC
limit 1)
... but I'd like to avoid a sub-query if possible, as I assume that would be quite a hit to the query performance.
CodePudding user response:
I can think of a few ways but since you're trying to avoid subquery, perhaps you can do something like this:
SELECT t.id,
SUBSTRING_INDEX(
GROUP_CONCAT(text ORDER BY CASE WHEN lang='en' THEN 0 ELSE RAND() END),',',1)
AS extracted_text
FROM
text t
LEFT JOIN text_lang tl ON tl.text_id=t.id
GROUP BY t.id;
Use GROUP_CONCAT()
with a customized ORDER BY
then use SUBTRING_INDEX()
to get the first value using comma (the default GROUP_CONCAT()
separator) as the delimiter.
I've made a fiddle with sample data here: https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=0d0501f36af1ba4e181856f761dbd5f7 . You can run the fiddle a few times to see that the text value other than 'en' change.