~l3kn assigned ~l3kn 3 years ago

~l3kn added Feature 3 years ago

~l3kn 2 years ago*

I've pushed a prototype of a caching indexer, it can enabled / disabled with org-fc-cache-mode.

Assuming only a small subset of files managed by org-fc changes during each review, it's much faster than indexing all files all the time.

The code is untested but the dashboard and reviewing cards work (and are noticably faster).

In case this cache becomes incoherent due to some bug in the code, numbers shown in the dashboard will be inaccurate. The worst thing that can happen when the cache is incoherent is that cards are scheduled for review before they are due. To prevent that, before each review the data of the current card is compared to the cached data and if they differ, an error is raised.

In that case, you can disable the cache with org-fc-cache-mode and continue using org-fc as usual, without any risk of "damaging" the review/spacing data.

~l3kn referenced this from #27 2 years ago

~l3kn 2 years ago

I've been using the cache mode without problems for the last few days. File titles were missing during the review but that has been fixed with the last commit.

~kindablue 2 years ago

I have started using the cache mode and began getting the error org-fc-review-next-card: Symbol’s value as variable is void: cur.

I think it is because the macro org-fc-review-with-current-item expects the current card to be setup/session to be setup in order to read from the org-fc--session variable. However, the function org-fc-cache-coherence-check is added to the org-fc-before-setup-hook upon activation of org-fc-cache-mode.

Would there be any significant affect if that is changed to org-fc-after-setup-hook?

(defun org-fc-cache-coherence-check ()
  "Check if the entry at point is coherent with its cache representation.
This is especially relevant w.r.t a card's due date / suspension state before review."
  (org-fc-review-with-current-item cur
    (if (org-fc-suspended-entry-p)
        (error "Trying to review a suspended card"))
    (let* ((position (plist-get cur :position))
           (review-data (org-fc-get-review-data))
           (row (assoc position review-data #'string=))
           (due (parse-iso8601-time-string (nth 4 row))))
      (unless (time-less-p due (current-time))
        (error "Trying to review a non-due card")))))

~l3kn 2 years ago

The before-setup-hook runs after a new card is set and before it is set up using the setup function of the card type, so I don't think the session not being defined is the core problem here.

However, I had placed the (require 'org-fc-cache) at a point where the review-with-current-item macro is not yet defined, loading the cache code it would be considered as a function with an undefined argument cur which might explain your error.

Another problem I found is that reviewing single buffers outside of the org-fc-directories doesn't work when cache-mode is active. (not fixed yet)

Did you notice any speedup with your setup when using the cache mode?

The coherence check is a bit of a hack to make sure using the cache doesn't do any damage, I think it could just be moved into the with-current-item macro once we know the cache works as expected (and ideally have some tests for it).

~l3kn 2 years ago

Even when requiring the cache code later, I ran into the same error trying to use cache-mode on a fresh emacs instance. I'm not sure why this is happening, adding an autoload seems to have fixed it.