Question

Let us suppose I have an app which has articles and authors. It persists its data in a Redis base. The articles are stored as hashes, as the one below represented in JSON syntax:

{ "title" : "My title", "content" : "This is a content", "authorId" : 1}

Also, my authors are represented by hashes:

{ "name": "John Smith", "username" : "jsmith", "password" : "secret", "id" : 1}

The question is, how would you retrieve all articles given an author id? Is there some command in Redis to do it? Would you retrieve and filter them manually? Or is my way of representing this relationship flawed? What would you suggest?

Answer 1

In Redis there’s no direct way to retrieve hashes using a field value as a selection criteria. With your current data model only way to it is to retrieve all article hashes and then discarding those that don’t match, but this method is far from efficient.

The fact is, you’ll have to maintain an index by yourself. As you can’t have sets inside hashes, I’d do this by having separate sets containing IDs of articles of an author. That way, you’ll have hashes of authors:

author_1: { "name": "John Smith", ... "id": 1 }
author_2: { "name": "Jane Doe", ... "id": 2 }

And sets for this mapping:

author_1_articles: [1, 3]
author_2_articles: [2]

And it maps to these articles:

article_1: { "title": "My title", ... "authorId": 1 }
article_2: { "title": "Another one", ... "authorId": 2 }
article_3: { "title": "What a story", ... "authorId": 3 }

Now, when you want to get articles for an author X, you’ll just get SMEMBERS author_X_articles and you’ll get the list of article IDs.

I’d use sets, because Redis will just ignore it if you end up adding same article for an author multiple times. So the list of articles you’ll get will never have duplicates, and you can use SADD without first checking if the article is already in the set.

This might sound like a lot of work, and it will be if your data model is complex and you have a lot of different access patterns you need to make fast. What you’ll end up doing is maintaining your own indexes, and for those you must duplicate data and keep it up to date. That’s the NoSQL tradeoff…