apollo-cache-updater

Status

Under heavy development

Why?

I wanted an updater that steals the magic of refetch queries while keeping the power of apollo local cache, but stripped of the boilerplate usually needed for each mutation update.

Updating the local cache becomes exponentially complicated when it needs to:

• include multiple variables
• include multiple queries
• know which of our target queries has been already fired before our speficific mutation happend
• cover scenarios** where apollo's in-place update may not be sufficient

** Add/remove to list, move from one list to another, update filtered list, etc.

This solution tries to decouple the view from the caching layer by configuring the mutation's result caching behavior through the Apollo's update variable.

Demo

Install

$ npm install --save apollo-cache-updater

OR 

$ yarn add apollo-cache-updater

Usage

Example: Add an Article

The following block of code:

• adds a new article to getArticles queries that contain the published: true variable
• adds 1 to the articleCounts queries that contain the published: true variable

import ApolloCacheUpdater from "apollo-cache-updater";
import { getArticles, articlesCount } form '../../apollo/api'; // your apollo queries

createArticleMutation({ // your mutation
    variables: {
        ...articleVariables // your mutation vars
    },
    update: (proxy, { data: { createArticle = {} } }) => { // your mutation response        
        const mutationResult = createArticle; // mutation result to pass into the updater
        const updates = ApolloCacheUpdater({
            proxy, // apollo proxy
            queriesToUpdate: [getArticles, articlesCount], // queries you want to automatically update
            searchVariables: {
                published: true, // update queries in the cache that have these vars
            },
            mutationResult,
        })
        if (updates) console.log(`Article added`) // if no errors
    },
})

Example: Remove an Article

The following block of code:

• removes an article with a specific id from the getArticles queries that contain the published: true variable
• subtract 1 from the articleCounts queries that contain the published: true variable

removeArticleMutation({ // your mutation
    variables: {
        id: article.id // your mutation vars
    },
    update: (proxy }) => {
        const updates = ApolloCacheUpdater({
            proxy, // mandatory
            queriesToUpdate: [getArticles, articlesCount], // queries you want to automatically update
            searchVariables: {
                published: true, // update queries in the cache that have these vars
            },
            operation: 'REMOVE',
            mutationResult: { id: article.id },
        })
        if (updates) console.log(`Article removed`) // if no errors
    },
})

Query params

This package assumes an exact correspondence between operation's params and query's params. So the updates will not work if you're using queries like this:

query getItems(
  $limit: Int
  $offset: Int
  $sort: String
) {
  getItems(
    options: {
      limit: $limit
      offset: $offset
      sort: $sort
    }
  ) {
...
}

It should be:

query getItems(
  $limit: Int
  $offset: Int
  $sort: String
) {
  getItems(
    limit: $limit
    offset: $offset
    sort: $sort
  ) {
...
}

Advanced Usage

@client directive: remote queries with mixed local state.

In case you have queries like this one:

const GET_TODO = gql`
  query todos {
    todos {
      id
      type
      local @client
    }
  }
`;

It's your duty to extend the mutation results with the local state fields that are going to be missing from the response:

update: (proxy, { data: { addTodo = {} } }) => {
              // your mutation response
              const mutationResult = addTodo; // mutation result to pass into the updater
              const updates = ApolloCacheUpdater({
                proxy, // apollo proxy
                queriesToUpdate: [GET_TODO], // queries you want to automatically update
                searchVariables: {},
                mutationResult: {
                  ...mutationResult,
                  local: client.localState.resolvers[
                    mutationResult.__typename
                  ].local(mutationResult)
                }
              });
              if (updates) console.log(`Todo added`); // if no errors
            }

Check this fully working example here

Example: Match any query

If you need to go through all matching query names ignoring any variables you should use the ANY operator. This does not work with the MOVE operation.

The following code removes the article with the matching id from any getArticles cached items, despite all possibe variables combinations stored in the cache.

removeArticleMutation({ // your mutation
    variables: {
        id: article.id // your mutation vars
    },
    update: (proxy }) => {
        const updates = ApolloCacheUpdater({
            proxy, // mandatory
            operator: 'ANY',
            queriesToUpdate: [getArticles], // queries you want to automatically update
            searchVariables: {},
            operation: 'REMOVE',
            mutationResult: { id: article.id },
        })
        if (updates) console.log(`Article removed`) // if no errors
    },
})

Example: Move an Article

The following block of code:

• removes an article from getArticles queries that contain the published: true variable and adds it to getArticles queries that contain the published: false variables
• adds 1 to the articleCounts that contain the published: true variable and adds it to articleCounts queries that contain the published: false variables

setArticleStatus({
    variables: {
        _id: id,
        published: false, // set the published status to false
    },
    update: (proxy, { data: { setArticleStatus = {} } }) => {    
        const mutationResult = setArticleStatus;
        const updates = ApolloCacheUpdater({
            proxy, // mandatory
            operation: 'MOVE',
            queriesToUpdate: [getArticles, articlesCount],
            searchVariables: {
                published: true, // find the mutation result article that in the cache is still part of the queries with published = true and remove it
            },
            switchVars: {
                published: false, // add the mutation result article to the queries that in the cache were invoked with published = false, if any
            },
            mutationResult,
        })
        if (updates) console.log(`Article moved`)
    },
})

Complete configuration object

{
    proxy, // mandatory
    searchOperator: 'AND', // [AND, AND_EDGE, OR, OR_EDGE, ANY], default AND. If you need to match all searchVariable, just one at least or none (with the latter will ignore any variables)
    searchVariables: {
        ...vars // searchVariables cannot be nested objects
    },
    queriesToUpdate: [...queries],
    operation: { // String || Object, default String ('ADD', 'REMOVE', 'MOVE', default: 'ADD')
        type: 'MOVE', // 'ADD', 'REMOVE', 'MOVE', default: ADD
        row: { // String || Object, default String ('TOP', 'BOTTOM', 'SORT', default: TOP)
            type: 'SORT', // 'TOP', 'BOTTOM', 'SORT', [SORT is effective only for ADD and MOVE], default: TOP
            field: 'createdAt', // if SORT, this indicates the field to be sorted
        },
    },
    switchVars: {
        ...otherVars, // switchVars cannot be nested objects
    },
    mutationResult, // mandatory
    ID: '_id', // Set the id field returned by your queries, default: id
}

Override default actions

For maximum flexibility you can also override the default actions of ADD and REMOVE operations.

Add 1 to all queries which data type is a number:

    operation: {
        type: 'ADD',
        add: ({ query, type, data, variables }) => {
            if (type === 'number') {
                return data + 1;
            }
        }
    }

Pass a custom action for the query articles

    operation: {
        type: 'ADD',
        add: ({ query, type, data, variables }) => {
            if (query === 'articles') {
                return [mutationResult, ...data]; // if you have mixed queries you need to extend the mutationResults with the missing local fields here too
            }
        }
    }

When it's an array of objects mutationResult must contain the right __typename field too.

Use the custom add/remove if you want to:

• override the default behavior for arrays
• override the default behavior for numbers
• add a custom function to handle strings (not handled by default)
• affect other variables depending on the query data
• you have specific needs that default actions do not satisfy

Note:

• when using custom add and/or remove sorting is disabled and will be ignored even if you set it. It's up to you to do the sorting in the custom function
• if you do not return the mutated data (or it is undefined) the custom add/remove function's result will be skipped and default actions'result will be used instead. However the logic inside the custom function will always be executed.
• if the operation type is MOVE you need to pass both custom add and remove. Passing just one of them will not work.

EDGE cases

There are edge situations where the cache includes queries like:

• articles({})
• articles({"sort":null,"limit":null,"start":null,"where":null})

This typically happen when a query with params like

query articles($sort: String, $limit: Int, $start: Int, $where: JSON) {
    articles(sort: $sort, limit: $limit, start: $start, where: $where) {
      _id
      title
      published
      flagged
    }
  }

gets called with either no variables object at all (variables object is not present) or a variables empty object has been passed, such as variables: {}. This may happen when variables are built programmatically.

articles({}) and articles({"sort":null,"limit":null,"start":null,"where":null}) are not handled by default and will be skipped, that is they will not be affected by the update.

However EDGE cases can be handled passing one of the EDGE searchOperator(s) such as AND_EDGE and OR_EDGE.

As an example using searchOperator: 'AND_EDGE' the end result would be:

On the other hand queries with no variables included like:

query articles {
    articles {
      _id
      title
      published
      flagged
    }
  }

are not considered EDGE cases. If included in the queriesToUpdate array they will be always updated like the following despite searchOperator that is used:

In the unlikely case that queriesToUpdate contains exclusively queries with no paramaters, the searchVariables should be an emptyObject:

update: (proxy, { data: { createArticle = {} } }) => { // your mutation response        
        const mutationResult = createArticle;
        const updates = ApolloCacheUpdater({
            proxy,
            queriesToUpdate: [getArticlesNoParams, articlesCountNoParams],
            searchVariables: {},
            mutationResult,
        })
        if (updates) console.log(`Article added`)
    },

License

MIT © ric0