来自两个集合的带有updatedocuments的AQL查询是否作为事务运行？

Question

I have a question about transactions in arangodb, if I run below AQL query, will it be executed as one transaction or will it be separated into two transaction? My backend is php:

LET r1 = (FOR u IN Users UPDATE u WITH { status: "inactive" } IN Users)
LET r2 = (FOR b IN Blogs UPDATE b WITH { status: "inactive" } IN Blogs)
RETURN true

For now I am using transaction as arangodb documentation suggests (using javascript code), but if using AQL query is possible, I'd prefer to remove js code from my php code!

If it's possible, do you suggest this solution for commiting transactions or using js way is preferred?

Answer 1

(edited from original)

According to the docs:

Each UPDATE operation is restricted to a single collection, and the collection name must not be dynamic. Only a single UPDATE statement per collection is allowed per AQL query, and it cannot be followed by read or write operations that access the same collection, by traversal operations, or AQL functions that can read documents.

In order to trigger a transaction you need to do so explicitly .

There are no individual BEGIN, COMMIT or ROLLBACK transaction commands in ArangoDB. Instead, a transaction in ArangoDB is started by providing a description of the transaction to the db._executeTransaction JavaScript function:

db._executeTransaction(description);

For example:

db._executeTransaction({
  collections: {
    write: [ "users", "logins" ],
    read: [ "recommendations" ]
  }
});

So, to answer the question, you need to use a client library to trigger a transaction as it will not happen automatically.

That being said, the main benefit of document/graph databases is the horizontal scaling, sharding, and cluster abilities. If you absolutely need to rely on transactions you might want to rethink why you are using a document based database. Eventual Consistency is typically good enough for a lot of use cases but in other cases you absolutely need ACID . A blog probably doesn't need ACID.

Answer 2

AQL in a single server environment is executed in an ACID fashion. This is automatically within a single transaction without any further need of a separate _executeTransaction . Therefore you can use the above AQL statement to update the two collections within a single AQL statement.

There are some caveat to keep in mind:

• after the update statement for a collection you cannot use this particular collection in any further update or read statement
• if you enable intermediate commits then the update will no longer be atomic