Efficient method to get item by id with Appsync and DynamoDB

Question

I want to create and use slug urls for use with a blog site built on top of Appsync and DynamoDB via AWS Amplify. The slug is a combination of the article title and a cuid; for example /article/an-interesting-post-on-bats-123abc would return the post with title "An Interesting Post on Bats." This would be relatively straight forward but for a few constraints:

DynamoDB's GetItem action only works with the primary key.
The primary key of an item cannot be changed. In order to "change" the primary key, you have to delete and recreate an item.
The Article entity data is malleable over time in that a user can change the Article's title, text, etc. A change to the title may warrant a change to the slug. In fact, a user can save an Article without having entered a title yet, so the slug could not be the basis for the primary key (as it is required to create an item).

It seems pretty clear that the slug itself is a poor choice for a primary key; however, the slug is the only input (as a path param) currently available upon which to create the query. Given these constraints, what is the most efficient way to query an Article if I only have the slug?

At this point, I'm early enough in the design phase that several resources can be modified: (1) the article urls, (2) the DynamoDB design (i.e. what the primary key and sort keys are); and (3) the Appsync schema.

Currently, the Article schema contains:

type Article @model {
  id: ID! // the primary key
  authorId: String
  title: String
  text: String
  slug: String
}

Although it is possible to scan for the slug, this seems like an inefficient way to utilize Dynamo for a regular access pattern. One thought I had was to make authorId the primary key with slug as a sort key or local secondary index. This appears to be similar to how Medium and StackOverflow create post urls.

Neill Neill · Accepted Answer · 2019-04-17T16:58:38

You can create a secondary index on the slug attribute. Then you would have to create another GraphQL query & it's respective resolver using the dynamoDB query operation. Query is far more efficient than Scan.

To do this, you have two options. In the first one, you can follow the design of the query operation and return a list of Articles instead of a single Article. This can be achieved using connectionModels or simply a type that returns a list.

For example

## Schema
type ArticleConnection{
  items: [Article]
}

type Query {
    getArticle(id: ID!): Article
    getArticleBySlug(slug: String): ArticleConnection // Query on secundary index
}


## Resolver
## Request Mapping Template
{
    "version": "2017-02-28",
    "operation": "Query",
    "scanIndexForward": true,
    "index" : "slug",
    "select" : "ALL_ATTRIBUTES",
    "query" : {
        "expression" : "slug = :slug",
        "expressionValues" : {
            ":slug" : $util.dynamodb.toDynamoDBJson("${ctx.args.slug}")
        }
    }
}

## Response Mapping Template
$util.toJson($ctx.result)

Or if you want to keep consistent with the first query, you can do the following

## Schema
type Query {
    getArticle(id: ID!): Article
    getArticleBySlug(slug: String): Article // Query on secundary index
}


## Resolver
## Request Mapping Template
{
    "version": "2017-02-28",
    "operation": "Query",
    "scanIndexForward": true,
    "index" : "slug",
    "select" : "ALL_ATTRIBUTES",
    "query" : {
        "expression" : "slug = :slug",
        "expressionValues" : {
            ":slug" : $util.dynamodb.toDynamoDBJson("${ctx.args.slug}")
        }
    }
}

## Response Mapping Template
#if($ctx.result.items.size() > 0)
   $util.toJson($ctx.result.items[0])
#else
    null
#end

Efficient method to get item by id with Appsync and DynamoDB

1 Answers