From the beta7 release:

'Remember that you'll need to reimport your .fc files, as previous .anki files
won't work.'

'Please don't use this beta for serious work. You may lose data, and your deck
will not be importable in the next beta.'

Perhaps I should put it in caps :-) If you have important work you want recovered, email me your deck and I'll do it manually.

(Note that the deck format is pretty close to being frozen - once it is, binary compatibility between future versions will be ensured)

Thanks for the report, though :-)

Beta 8

The backend has stabilized, and unless some unforeseen bug reveals a flaw in
the design, the deck format will not change again.

I've done more work on importing, making it better at both guessing if you
have one-sided cards, and also more correctly differentiating between furigana
and comments. Even if you only have one sided cards, it should import
automatically this time without the need to go through the importing dialog.
I'd appreciate it if as many people as possible could try starting beta8 with
their old Anki deck and confirm that their old data has been imported
correctly. I'm pretty sure it'll work for JimmySeal's decks now - yorkii, how about you?

Also new is auto-generation of readings for Mandarin and Chinese (per-kanji -
not very intelligent at the moment) and proper cross-platform font support (at
least for 明朝).

And of course, a number of bugs have been fixed.

**NOTE**

I'm almost ready to declare the beta stable enough to use, but I'd like
another day or so of testing. Please wait just a little longer.

Also note that the upgrade process needs to know your previously selected
'target language', in order to decide whether your deck should be 'Japanese'
or 'English'. If you want the upgrade to do the right thing, remove the old
configuration file (Mac/Linux: rm -rf ~/.ankidev/config.db, Windows, find it in
Documents and Settings) and then copy the config.db file from ~/.anki to
~/.ankidev

The test server is at the same location as last time. Any data left there will
be lost when I release 0.3.0. Please don't rely on it.

Let me know how you go!

Study: http://repose.ath.cx:5001/study
Win: http://repose.cx/dump/anki-0.3.0-beta8.zip
Mac: http://repose.cx/dump/Anki.dmg
Linux: http://repose.cx/dump/beta8.tgz

Hi resolve,

I upgraded several of my 2.8 decks to the beta8 format (under Linux), and it mostly went as I expected. Two things however might interest you.

1) It is possible to lose cards in the process. I can send you a deck which does that, if need be, but it's simple enough to reproduce with the following three cards (using either "open" or "import"):
violette -> violet
violet -> purple
purple -> violet

This example can seem a bit contrived, but I got the same situation in a real-life deck ("capitals of the world") which I had cleaned after downloading raw data from Wikipedia (so at the end it was neither completely one-sided nor two-sided). I would rather have the upgrade process stop (even if it requires my cleaning the old deck in some way) than have it silently drop cards which it cannot fit in the new format.

2) it looks like the "next due" dates are not always correctly assigned. In particular, on that same deck (I haven't checked on my Japanese decks 'cause I don't want to mess with those after all the hard input work...), many cards are found to be new, others have different due dates, and I can see no logic in the way the due dates are modified. But it might, of course.

Finally, I find Anki beta8 has many improvements over an already very good Anki 2.8!!! Thanks a lot for this excellent piece of software!

Hi Laurent,

Thanks for the feedback.

1) is difficult to fix without inconveniencing people. If you deck has many of these 'triangle' cards, Anki will choose a different importing algorithm, but if there are only a few, it decides that correcting spelling mistakes should be a higher priority. For example, if you have 2 cards:

２年間 : 2 years
2 years : 2年間

This will fail without fuzzy matching as the number is half width in one of the cards and full width in the other. People who've been using Anki for quite a while invariably have quite a few of these cards.

Note that you can still import those cards properly if you do a manual import via the 'import' dialog.

2) That is more serious. Could you send me the decks in question, in .fc format? Are you sure it's not related to number 1)? Also, upgrading doesn't touch the old file (a new file is made), so it should be safe to try it on your Japanese decks too.

Just wondering, resolve -- why is that when I add new cards to my deck, it adds them as a 'fact' and not as a card? I am unable to find recently added 'facts' in my 'Edit Deck' window too.

Edit: Okay, I've re-enabled the Recognition feature, and new cards that I add appear normally now. However, how may I retrieve the 'hidden' facts card which I've already added sometime back?

Cards show a given fact in a particular way. When you click on a card in the 'edit deck' window, the bottom half of the screen, which allows you to edit fields, is the 'fact editor'. I don't understand what you mean about being unable to find recently added ones, though. If you sort by creation date, you can't see them?

I've tried sorting them by creation date, but the cards which I've input during while the Recognition feature was disabled still do not appear inside the 'Edit Decks' window.

Ah. You're not supposed to be able to disable all the cards - the user interface won't let you. Unfortunately I forgot about that when I manually upgraded your deck. If you send it to me again I'll attach cards to the new information you've added. :-)

Thank you! File sent.

resolve Wrote:1) is difficult to fix without inconveniencing people.

[...]

Note that you can still import those cards properly if you do a manual import via the 'import' dialog.

As a matter of fact, I did try to import the deck, and that failed. Something to do with it not being UTF-8 (which is strange, as I switched my whole system to UTF-8 some time ago. Besides, the characters involved are in the ASCII range. Oh well.).

Actually the import worked (when I made the right choices) but some cards (the same ones) are still missing.

Quote:2) That is more serious. Could you send me the decks in question, in .fc format? Are you sure it's not related to number 1)?

Yes, I'm going to send you one deck where the discrepancy appears. And no, I'm not sure this behaviour is not related to the previous one, but I noted that the cards involved are not the same.

Quote:Also, upgrading doesn't touch the old file (a new file is made), so it should be safe to try it on your Japanese decks too.

Yeah, well I was probably being a bit paranoid... thanks for the reassurance :/

If you're getting errors about UTF8, you're trying to import an Anki deck as a CSV file. You need to choose the format from the top drop-down list (select "Old V2 deck")

yes, silly me, I realized that (about UTF-8) a few minutes after posting my message (I corrected it immediately, but too late for you...).

I sent two decks to the address indicated on your website.

Laurent: I've addressed both these points in Beta9 - I'll send you an email about it.

Beta 9

I have frozen the deck structure now, so decks created with beta8+ will
continue to function on future versions. You can now start using Anki to study
seriously. Please save regularly (don't rely on the 'auto save after every __
questions', as it doesn't make a backup).

Also, please make sure that your deck has been upgraded properly before you
start to use the beta seriously.

New in Beta9 is a different sorting system for the cardlist, some more tweaks
to the importing code, and some small bugfixes.

**NOTE**

Beta9 still syncs to a testing server. Anything stored there will be lost when
0.3.0 is released, so please keep your local deck safe.

Also note that the upgrade process needs to know your previously selected
'target language', in order to decide whether your deck should be 'Japanese'
or 'English'. If you want the upgrade to do the right thing, remove the old
configuration file (Mac/Linux: rm -rf ~/.ankidev/config.db, Windows, find it in
Documents and Settings) and then copy the config.db file from ~/.anki to
~/.ankidev

Let me know how you go!

Study: http://repose.ath.cx:5001/study
Win: http://repose.cx/dump/anki-0.3.0-beta9.zip
Mac: http://repose.cx/dump/Anki.dmg
Linux: http://repose.cx/dump/beta9.tgz

(edit: silly typo)

Just downloaded Anki after I read about it on this forum. It looks like what I've been looking for. Adding Hiragana for my sentences had always been unnecessary work imho, but something you just had to do. Looking forward to using Anki.

One thing that bothers me though is this: "By default, a reverse entry will be added as well, so you study question to answer, and answer to question."

I'm using the AJATT method, so I don't need any reverse entries. Is it possible to switch this off?

watashimo: I'm assuming you're talking about 0.2.8. It certainly is possible to turn it off - just uncheck the 'add reverse entry' box at the bottom of the 'add cards' dialog.

Ah, thanks, didn't see that box. Gotta play around with it now. I'm currently reviewing Heisig kanjis, so it will probably take another two weeks, before I can really test Anki.

Thanks for the reply.

I just started using Anki 0.2.8 a week or so ago. Loving it so far. I know priority is on 0.3 now but I thought it might not hurt to ask anyway. I have one card that just suddenly jumped to a review gap of 120 days. Any idea how that happened and how I can bring back to a more reasonable gap since I just started reviewing it. All the other cards are fine.

kurojohn: sorry, could you confirm where that happened? Did it jump to that number on an 0.3.0 beta - if so, which? Or did it happen on 0.2.8? There was a bug in the early betas where new cards were scheduled based on the time they were added instead of the time since the first review - but that should be fixed in the later betas.

Sorry for not making it clear. It happened in 0.2.8. I haven't tried out 0.3 yet. Decided to wait for the final release.

I've never seen such behaviour on 0.2.8 - that's very strange indeed. Anyway, if you send me your deck I'll fix it for you.

JimmySeal: The reason why your deck didn't upgrade flawlessly on the server is because your 'target language' was set to English. The web interface used to ask you what language you want to study when you signed up, though that may have been disabled when you signed up. Consequently it's trying to put English text in the first column. If you upgrade the deck locally, it should work automatically.

Thank you for your vigilance but even with a local upgrade, there are big problems with one deck:

-I import with "Mandarin" as the target language (field 0 -> expression, field 1 -> meaning):
The upgrade box reports 516 duplicates. No new cards are created, but about 90% of the cards are mutated into reverse entries. That is, their scheduling and stats remain the same, but answer and question are swapped. Anki is reporting duplicates because what used to be cards with unique questions:
金 - jin1
巾 - jin1

are now cards with identical questions:
jin1 - 金
jin1 - 巾
this is the same result as the automatic upgrade that you performed on the server.

-I import with "Mandarin" as the target language (field 0 -> expression, field 1 -> reading):
About the same result as the previous one except 513 duplicates are found and on a handful of cards (those with English in the question), the answer and question are merged into the answer field, and the question field becomes BLANK! So,

Old:
stammer - 吃
New:
[blank] - stammer 吃


-I import with "Other" as the target language:
The upgrade box reports 516 duplicates, and it creates reverse entries for every card in the deck (total number of cards doubles; 1768 facts, 3536 cards).

If you just use the automatic upgrade locally, it will get it almost right: provided the target language was set to "Japanese" in Anki 0.2.8, and you copy the config.db file to ~/.ankidev as described in the beta instructions, it will automatically detect the deck as one sided. Your novels deck will be upgraded perfectly, and only the English meanings in your Chinese deck will be reversed. All you have to do is load your old .fc file via Deck -> Open.

Anki 0.2.8 only had the concept of "Japanese" and "not Japanese", the latter being primarily used by my English students. The standard Anki format involves two cards:

eat : 食べる
食べる : eat

Anki has to know which field is the 'expression' and which field is the 'meaning'. The reason all of your cards are being switched around is because Anki thinks that your deck is English (because it doesn't have a 'Japanese' tag in it).

You can accomplish the same thing as an automatic import by:

1. creating a new deck with 'Japanese'
2. going into the 'model properties' and turning off 'unique' for the 'meaning' field (since you have the same reading for multiple kanji)
3. on the 'cards' tab, deleting the 'production' card
4. importing your old deck

You can then return to the model properties, rename the model from 'Japanese' to 'Mandarin', change the 'features' and 'tags', and you'll be set. But making sure your target language is 'Japanese' and using the automatic upgrade will be faster :-)

Thanks for all the help. Your instructions almost worked except that all multiline cards (about 150 of them) were reversed as though they were recognized as English or something. When I referred to 90% being swapped in a prior post, this is the 10% that was previously being left alone.

Wouldn't it make sense to have an import template that just leaves things as-is, without any fancy pattern matching, grouping, or auto-corrections? I would think that the Other template would behave that way, but it still seems to be trying to categorize things.