Question

I’m trying to model data similar to amazon.com categories (e.g., books, movies, electronics, etc.) in a Redis database. Order matters to me when they’re rendered on an HTML page so that the user is presented with a consistent user interface. As such, I stored the categories in a Sorted Set:

ZADD categories 0 "Books"
ZADD categories 1 "Movies"
ZADD categories 2 "Electronics"

I then created another Sorted Set for each of the sub-categories.

ZADD categories:books 0 "Fiction"
ZADD categories:books 1 "Non-Fiction"
ZADD categories:movies 0 "Horror"
[...]

From here, I figured I could store products in a Hash.

HMSET product:1000 category 0 subcategory 0 title "Redis Cookbook"
HMSET product:1001 category 1 subcategory 0 title "Nightmare on Elm Street"
[...]

I’m very new to both Redis and Key/Value database stores, so I’m anything but confident in my approach. Is this model going to work for me long-term? Is there a better/alternative approach I should be aware of? One concern I have is keeping the names “synchronized”. For example, if I change the top-level category from “Books” to “Literature” (terrible example, I know) all of the keys for the sub-categories that “reference” books should also be updated.

Answer 1

Lets answer your question backwards.

Storing Products in a Hash

That is a good idea. In general, you want to store entities in a hashmap.

Changing Category Names

You should use an internal identifier for each category. Then store the category display name in a hashmap just like you store products. This indirection is useful – because it lets you store additional information for a category. For example, if you later decide to display “Most Popular Product” for each category, such a schema would be useful.

Storing Products in Category

You can create a key like category:books:products and then store the ids for each product. This can be a list, a set or sortedset – depending on what you want to do with the data.

1. If there is an intrinsic order to the products, such as recently added products, then it makes sense to use a List. In this way, the first element in the list is the most recent product.
2. If you want the products to be sorted on some criteria such as “most downloaded” or “most viewed”, you should use a sorted set.
3. Use a Set if there is no other relationship between products in a category.

Using a Set or SortedSet lets you perform intersections. These come in handy when you want a subset of items, such as ‘Books published by O’Reilly”. To do this, you will have to maintain another set “publisher:preilly” that contains book ids.

Storing a collection of Categories

Now, you suggested a SortedSet, which is one possible solution. But you can also use a list if the order of categories isn’t going to change dynamically. For example, if you always want to display categories in ascending order, a list is good enough.

A SortedSet proves useful if the order will change dynamically. For example, if you want to display categories on the basis of the number of products in each category – then storing them in a SortedSet is a good idea.

Also, remember, you can use several lists or sortedsets – each collection can represent a different use case. Lets assume you don’t want to sell books to overseas customers. In that case, you can have another key categories:overseas and store the categories you want to display.