in the upcoming daily unburying, this could lead to a state
where the remote end unburies just at the start of sync
and clobbers more recent changes made on the local end
First, burying changes:
- unburying now happens on day rollover, or when manually unburying from
overview screen
- burying is not performed when returning to deck list, or when closing
collection, so burying now must mark cards as modified to ensure sync
consistent
- because they're no longer temporary to a session, make sure we exclude them
in filtered decks in -is:suspended
Sibling spacing changes:
- core behaviour now based on automatically burying related cards when we
answer a card
- applies to reviews, optionally to new cards, and never to cards in the
learning queue (partly because we can't suspend/bury cards in that queue at
the moment)
- this means spacing works consistently in filtered decks now, works on
reviews even when user is late to review, and provides better separation of
new cards
- if burying new cards disabled, we just discard them from the current queue.
an option to set due=ord*space+due would be nicer, but would require
changing a lot of code and is more appropriate for a future major version
change. discarding from queue suffers from the same issue as the new card
cycling in that queue rebuilds may cause cards to be shown close together,
so the default burying behaviour is preferable
- refer to them as 'related cards' rather than 'siblings'
These changes don't require any changes to the database format, so they
should hopefully coexist with older clients without issue.
Python used to use C-style division, where division of two ints was
truncated, and division involving a float resulted in a float.
This is confusing, because you often can't tell from looking at a
line of code in isolation what sort of division it's supposed to do.
With 'from __future__ import division' Python ensures that division is
always explicit.
// means (floored) integer division
/ means float division
regardless of argument types.
This should make the source a bit clearer now, as well as removing one
obstacle if Anki is ever ported to Python 3.
