Building upon @frank-van-puffelen's accepted answer on the old question, this will now delete the message IDs from their sender's user data as part of the same atomic delete operation without firing off a Cloud Function for every message deleted.
Method 1: Restructure for concurrency
Before being able to use this method, you must restructure how you store entries in /User/someUserId/myMessages to follow best practices for concurrent arrays to the following:
{
"/User/someUserId/myMessages": {
"-Lfq460_5tm6x7dchhOn": true,
"-Lfq483gGzmpB_Jt6Wg5": true,
...
}
}
This allows you to modify the previous function to:
// Cut off time. Child nodes older than this will be deleted.
const CUT_OFF_TIME = 24 * 60 * 60 * 1000; // 2 Hours in milliseconds.
exports.deleteOldMessages = functions.database.ref('/Message/{chatRoomId}').onWrite(async (change) => {
const rootRef = admin.database().ref(); // needed top level reference for multi-path update
const now = Date.now();
const cutoff = (now - CUT_OFF_TIME) / 1000; // convert to seconds
const oldItemsQuery = ref.orderByChild('seconds').endAt(cutoff);
const snapshot = await oldItemsQuery.once('value');
// create a map with all children that need to be removed
const updates = {};
snapshot.forEach(messageSnapshot => {
let senderId = messageSnapshot.child('senderId').val();
updates['Message/' + messageSnapshot.key] = null; // to delete message
updates['User/' + senderId + '/myMessages/' + messageSnapshot.key] = null; // to delete entry in user data
});
// execute all updates in one go and return the result to end the function
return rootRef.update(updates);
});
Method 2: Use an array
Warning: This method falls prey to concurrency issues. If a user was to post a new message during the delete operation, it's ID could be removed while evaluating the deletion. Use method 1 where possible to avoid this.
This method assumes your /User/someUserId/myMessages object looks like this (a plain array):
{
"/User/someUserId/myMessages": {
"0": "-Lfq460_5tm6x7dchhOn",
"1": "-Lfq483gGzmpB_Jt6Wg5",
...
}
}
The leanest, most cost-effective, anti-collision function I can come up for this data structure is the following:
// Cut off time. Child nodes older than this will be deleted.
const CUT_OFF_TIME = 24 * 60 * 60 * 1000; // 2 Hours in milliseconds.
exports.deleteOldMessages = functions.database.ref('/Message/{chatRoomId}').onWrite(async (change) => {
const rootRef = admin.database().ref(); // needed top level reference for multi-path update
const now = Date.now();
const cutoff = (now - CUT_OFF_TIME) / 1000; // convert to seconds
const oldItemsQuery = ref.orderByChild('seconds').endAt(cutoff);
const snapshot = await oldItemsQuery.once('value');
// create a map with all children that need to be removed
const updates = {};
const messagesByUser = {};
snapshot.forEach(messageSnapshot => {
updates['Message/' + messageSnapshot.key] = null; // to delete message
// cache message IDs by user for next step
let senderId = messageSnapshot.child('senderId').val();
if (!messagesByUser[senderId]) { messagesByUser[senderId] = []; }
messagesByUser[senderId].push(messageSnapshot.key);
});
// Get each user's list of message IDs and remove those that were deleted.
let pendingOperations = [];
for (let [senderId, messageIdsToRemove] of Object.entries(messagesByUser)) {
pendingOperations.push(admin.database.ref('User/' + senderId + '/myMessages').once('value')
.then((messageArraySnapshot) => {
let messageIds = messageArraySnapshot.val();
messageIds.filter((id) => !messageIdsToRemove.includes(id));
updates['User/' + senderId + '/myMessages'] = messageIds; // to update array with non-deleted values
}));
}
// wait for each user's new /myMessages value to be added to the pending updates
await Promise.all(pendingOperations);
// execute all updates in one go and return the result to end the function
return ref.update(updates);
});
