Fix/filter out manual and cram reviews in true retention & use _revlogLimit()

[L-M-Sherlock]

fix:

• Discrepancy between add-on and native True Retention ankitects/anki#3453

[user1823]

I don't know why but this doesn't seem to have any effect (apart from the "Deck" vs "Collection" issue).

[user1823]

Well, that's because the code for true retention already filters the revlogs with ease = 0 and I don't have any revlogs with type = 3 and factor = 0.

[L-M-Sherlock]

Well, that's because the code for true retention already filters the revlogs with ease = 0

The previous version doesn't.

[user1823]

The previous version doesn't.

It does filter them in the following lines:

fsrs4anki-helper/stats.py

Lines 389 to 392 in a459532

	sum(case when lastIvl < %(i)d and ease = 1 and (type = 1 OR lastIvl <= -86400 OR lastIvl >= 1) then 1 else 0 end), /* flunked young */
	sum(case when lastIvl < %(i)d and ease > 1 and (type = 1 OR lastIvl <= -86400 OR lastIvl >= 1) then 1 else 0 end), /* passed young */
	sum(case when lastIvl >= %(i)d and ease = 1 and (type = 1 OR lastIvl <= -86400 OR lastIvl >= 1) then 1 else 0 end), /* flunked mature */
	sum(case when lastIvl >= %(i)d and ease > 1 and (type = 1 OR lastIvl <= -86400 OR lastIvl >= 1) then 1 else 0 end), /* passed mature */

Anyway, the changes LGTM. They don't solve the issue reported by me, though.

[L-M-Sherlock]

It does filter them in the following lines:

Nope. It just treats them as failed (0). That causes the previous discrepancy.

[user1823]

Well, these lines include ease = 1 and ease > 1. So, I don't know why you think that the add-on is counting the revlogs with ease = 0.

I think that the discrepancy is corrected because the latest commit in Anki uses the correct code to remove the crammed revlogs.
(I am now sure about this because the number of revlog entries with type = 3 and factor != 0 is exactly equal to the size of the discrepancy.)

[L-M-Sherlock]

Fine. The bug has been fixed. It's a good news.