Duplicate reviews in the reviews log/history before a certain date

I’ve already figured out a solution, so I am writing this mainly for checking with you whether my solution is good and for documentation purposes (as it is quite a strange bug).

Problem description

So, my review log from before 17 September 2019 was really messed up. It seemed to contain duplicates (multiplicates, actually) of all the reviews that I did. Eyeballing it, I saw a 20-fold excess.

Starting from 17 September 2019, the review numbers were on a realistic level.

Fortunately, I didn’t detect any multiple logs in the review history of any individual card (Ctrl+Shift+I in the browser).

Evidence

From the new stats screen:

imagen

imagen

From the classic stats screen:

Total: 3886418 reviews (allegedly!)

From Throughput Historical

Consequences

One obvious problem created by this is that I couldn’t sync to Ankiweb. The collection was “too large to send to AnkiWeb” (287MB > 250MB). I already posted about this in September 2019, but I didn’t know the underlying cause.

I don’t have a clue which other problems were caused.

Possible cause

The multiplicates were presumably caused by some exporting and importing of decks which I did at that time (iirc).

Solution(s?)

Solution that works

I have exported the collection (Ctrl+E) in the .apkg format. Then I created a new profile and imported the .apkg. My assumption was that this way, Anki will get the review history directly from the cards. Thanks to @abdo for the inspiration.

I’ve also moved certain contents from \Anki2\oldprofile\ to \Anki2\newprofile\ (backups, collection.media, media.trash, deleted.txt, anki_killstreaks.db)

Evidence

Looks good:

imagen

Interestingly (or shall I say, suspiciously?), 17 September 2019 now has 322 reviews instead of the 323 from the above screenshot…:

imagen

Total: 268856 reviews (down from 3886418)

imagen

The collection.anki2 file now has only 69.2MB.

And for the first time in 3 years I can sync with Ankiweb again, hurray!

Alternatives?

I haven’t looked at the review log in a readable form [edit: This is actually possible once you’ve extracted the revlog.csv, apparently]. Perhaps I could detect and delete duplicate entries with some database/table/text editing software.

Question

Could my working solution cause any problems that I have not foreseen? Any valuable information that is stripped from the collection when I export as .apkg?

marking topic as solved

You’ll lose collection-global config, which is usually not a big deal. You’ve effectively removed any review log entries associated with cards that have subsequently been deleted.

1 Like