Question

Please excuse the poor title – not sure how to phrase what I’m trying to do in a short, simple way.

I’m working on a new UI for an app with an existing DB that is localized. There can be any number of languages, and they can be created, editted or removed at any time by admins. The DB is being used by other applications, and I’m not able to modify the structure.

The DB manages a collection of artwork. There are several tables of “assets”, such as media and pieces (and others, but those 2 should suffice as examples).

Each asset table has a field called localization. This is basically a unique identifier, and is a FK to a table of the same name (localizations). for simplicity, assume the media table looks like this

id localization
1  1
2  2
3  5
4  8
5  11
...

and the pieces table uses the same localization increment, so might look like this

id localization
1  3
2  4
3  6
4  9
5  20
...

the localizations table is just a series of auto-incremented PKs.

localization
1
2
3
4
5
...

there is a localization_entries table that ties key:value pairs to a localization:

localization  language  entry_key  entry_value
1             en        title      photo of dog
1             en        detail     this is long text describing the photo of the dog.
1             es        title      foto de perro
2             en        title      photo of cat
...

Any particular localization might have an entry key and value for none, any or all of available languages.

We need to be able to search and sort using localized values. Initially, I thought the only way was to load up the dataset in it’s entirety (multiple queries), populate a PHP array of objects, then perform search and sort with PHP, but I wonder if it’s possible using straight SQL.

Ideally, I’d like to get back rows that have all localized entry values. For example, the row for the first localization would have all the fields for the media item itself, in addition to all localized values for that localization.

Something like this “works”

SELECT media.id, media.name, en.entry_value as en_title, sp.entry_value as sp_title
FROM media
JOIN localization_entries en
ON media.localization=en.localization 
JOIN localization_entries sp
ON media.localization=sp.localization
AND en.language='en' AND en.entry_key = 'title'
AND sp.language='es' AND sp.entry_key = 'title'

But obviously anything that doesn’t have an an entry_key of the appropriate key or language would not be returned (e.g., anything without an ‘es’ entry_key of ‘title’). Further, adding any kind of WHERE or ORDER BY clause fails as well…

SELECT media.id, media.name, en.entry_value AS en_title, sp.entry_value AS sp_title
FROM media
JOIN localization_entries en
ON media.localization=en.localization 
JOIN localization_entries sp
ON media.localization=sp.localization
AND en.language='en' AND en.entry_key = 'title'
AND sp.language='es' AND sp.entry_key = 'title'
WHERE en.entry_value like '%whatever%' -- would not work
ORDER BY en.entry_value

Is there a practical way to search / sort / filter using MySQL for a set of relationships like this?

TYIA

Answer 1

Use an outer join:

SELECT media.id, media.name, en.entry_value as en_title, sp.entry_value as sp_title
FROM media
LEFT OUTER JOIN localization_entries en
ON media.localization=en.localization AND en.language='en' AND en.entry_key = 'title'
LEFT OUTER JOIN localization_entries sp
ON media.localization=sp.localization AND sp.language='es' AND sp.entry_key = 'title'

You’ll get NULL for en_title or sp_title if there’s no matching row in localization_entries.