rix0rrr
commented
1 month ago @Annelein this should have thrown an error locally before deploying and running against an actual DynamoDB, so my apologies that it didn't. But also, update() is supposed to only get the fields you want to change, while put() should be used if you really want to overwrite all fields of an entire record (and delete all non-included fields).
# Insert or replace an entire record (all fields not part of this write are gone)
table.put(entire_record)
# Update a single field (all other fields remain unchanged)
table.update(record['id'], { 'field': 'new_value', 'another_field': 'another_value' })update() is safe against concurrent updates(), while put() will just overwrite all fields (even if another request changed any of the fields in the mean time).
mergify[bot]
Annelein
When logged in, we automatically save a program as the user runs it.
In a recent change, we update the program counter only when a program is modified from the example code. But we wrote the entire 'program' record back to the database in an
update()call, which includes theidfield; Dynamo does not allow including theidfield in anupdate()call.This is also not necessary: a DynamoDB
update()call only needs to include the fields we actually want to change, and not any of the fields that are unchanged (this may even lead to a race condition if any of those fields were changed in between the read and update calls, we would undo the intermediate change again).Instead, only update the
is_modifiedfield, and update the Dynamo abstraction layer to check for this error condition.How to test
Log in, and run a program.