Automatic chunking on assets.delete fails

[Deleted user]

Assume you have a list (assetIds) of asset ids you want to delete. The list of asset ids represents an asset hierarchy with 1 root and 1000 direct children. The list looks like this: [root, child1, child2, ..., child999, child1000].

If you call client.assets.delete(assetIds.map(assetId => ({ id: assetId }))) #176 will automatically chunk your delete-request into two API-calls:

1. Delete [root, child1, child2, ..., child999]
2. Delete [child1000]

If the API-calls is executed in that order then the 2nd request will fail with 400 (didn't find the asset). The reason for this is that the root-asset was deleted in the 1st API-call and therefore deleted the entire subtree.

Since the execution order matters (API-call 2 and then API-call 1 will not fail) and since the SDK cannot control it if we want to parallelise then it doesn't make sense to return info about failed requests to the end-user.

We can solve this by parsing the 400-response and look for objects in the missing array (array of asset ids that don't exist) and remove them from our requests before we retry the request.

[Deleted user]

Cognite is currently discussing internally if we want to introduce a new parameter to the delete-asset endpoint for silently ignoring unknown IDs. Putting SDK fix on hold

[eirikdahlen]

Moved to JIRA