I had a question regarding the usage of Multi.run/3. I am using pg as my database. Let’s say I have a Multi with a couple different updates to my db. At the very end, I want to run a custom function that does not update the db (so no Repo.insert etc. It updates an ets table) using Multi.run/3.
My question is, does it make sense to use Multi.run/3 to run functions that do not update the db? I would like to do this because my custom function can return {:error, value} and all the preceding updates can be rolled back.
You use Multi.run for something not DB oriented but where you still want the DB to rollback if it fails, so yes, this is the use-case it is for.
Just remember that an entire Multi pipeline is in a transaction, so the rows it accessed in the database are locked until it’s complete, so make sure it doesn’t take too long to run.
Edit: I was also wondering, if you use a Multi.run to actually update the DB, does the update get rolled back when some other operation down the line fails? If so, how does Ecto figure out which Multi runs update the DB and which ones don’t (In order to rollback any updates on failure)?
@apr Ecto itself doesn’t “undo” the operations, rather it opens a transaction and then all operations within that transaction will be rolled back by the database unless the transaction is committed.
In order for an operation in a Multi.run block to be included in the transaction must the run block use the Repo argument passed from the earlier Multi step or can it call a function that makes arbitrary Repo calls while not using the repo passed in from the Multi? This could be such as calling Repo.one, Repo.insert, or something like that
