Possible Data Corruption Bug.

I noticed this evening when testing with the latest extension that when I switch from one document to another the log information retains text from the prior documents. For example, when working with a test account I first opened up a blank doc and typed 'Hippidy skip' I then opened up the shared assignment doc in the google classroom and typed some other random gibberish, and then turned to a third doc with new data. Yet the running log contents appear to show all of the file text being added together in the same record:

2021-08-19T03:30:39.680398: pipeline/handler/incoming_websocket_handler Processing PubSub message google_docs_save from org.mitros.writing-analytics
[.process at 0x7fa45dac9ca0>]
{'text': 'Testing eiyh TeThis is pe\n\n\n\nA new thaing baby big fun baby\n\n\nHippidy \n\nsdfaskd\n\nWe atrdoing \n\nA thing', 'position': 16, 'edit_metadata': {'cursor': [9, 10, 11, 18, 46, 47, 49, 50, 51, 53, 55, 57, 59, 60, 61, 63, 65, 67, 68, 2, 3, 4, 5, 7, 9, 11, 12, 7, 10, 13, 14, 16, 18, 20, 21, 22, 24, 25, 28, 30, 32, 34, 36, 37, 38, 39, 3, 8, 11, 14, 15, 16], 'length': [8, 9, 10, 17, 18, 19, 21, 22, 23, 25, 27, 29, 31, 32, 33, 35, 37, 39, 40, 41, 42, 43, 44, 46, 48, 50, 51, 47, 50, 52, 53, 55, 57, 59, 60, 61, 63, 64, 67, 69, 71, 73, 75, 76, 77, 78, 80, 85, 88, 101, 102, 103]}}

I have yet to investigate whether this affects other log data or the pub-sub but it is showing up in the log.

Start at pipeline/handler/incoming_websocket_handler for the google_docs_save PubSub message and reproduce the documented sequence of switching among three Google Docs. Trace whether text from the prior document remains in the record; done means each log record contains only the current document’s text.

Written by the indexing model from the issue text.