Add ekg-edit-hook

[qingshuizheng]

Senario: auto save without having to finalize

I don't finalize the ekg capture/edit buffer too often, just to keep them open for more info to come in. Very often I mistakenly kill an unsaved note buffer, and those are the moments I'm dying inside. :-(

So I have a workflow for auto-saving current ekg-capture/edit instance without having to finalize using ekg-capture/edit-finalize. The note auto saves every 1 sec using the goodies of the package super-save but with an advised super-save-command

   (define-advice super-save-command (:override (&rest _) with-ekg-support)
     "Save the current buffer if needed."
     (if (and (or (bound-and-true-p ekg-edit-mode)
                  (bound-and-true-p ekg-capture-mode))
              (not (ekg--in-metadata-p)) ; don't auto save while in overlay, to avoid auto-saving useless tags
              (buffer-modified-p (current-buffer)))
         ;; ekg buffer
         (progn
           (ekg--update-from-metadata)
           (ekg--save-note-in-buffer)
           (message "Saved."))
       ;; other buffer
       (when (super-save-p) (save-buffer))))

If the note is saved via the above auto-save mechanism, the note will be tagged as "raw", so I have a clue to find them later on; If the note is finalized normally, the "raw" tag will be converted to "draft".

The workflow relies on the state of buffer-modified-p in capture/edit buffer. At the moment the capture/edit buffer's initial modified state is always t. It should be set default to nil, otherwise it will trigger super-save to save when a note is captured or opened.

I have tried it in my personal fork commit (bec068f) for one month or so, which works great. But I cannot think of a good reason to ask for adding a (set-buffer-modified-p nil) to both ekg-capture and ekg-edit. Or maybe It's okay to add it?

A more generalized solution: add a new HOOK ekg-edit-hook

So what I have in my mind is to add a HOOK ekg-edit-hook within ekg-edit, thus user could use it to add customizations like other -hook does.

For my use case:

(defun z/ekg-set-buffer-not-modified (&optional _note)
  "Set buffer state as not modified."
  (set-buffer-modified-p nil))

(defvar ekg-edit-hook nil  "Hook run after opening a note.")
(setq ekg-edit-hook '(z/ekg-set-buffer-not-modified))
(setq ekg-pre-save-hook '(z/ekg-set-buffer-not-modified))

Like to hear you thoughts.

[ahyatt]

I think setting the modified state to nil seems pretty reasonable. After all, when you first open / capture, it really isn't modified even though we add metadata. Perhaps it could be better, though, maybe the best thing to do is to note before changing the metadata section if the buffer is unmodified, and if it is, reset it to unmodified after changing the metadata.

Also, your use-case seems a lot like #19, and I have the same objection, which I don't think I've really thought of a great solution for yet. If you have any ideas, let me know!

[ahyatt]

Hm, interesting. The first output shouldn't have shown up as two identical titles. I see only one actual title, and no dups.

But the lack of indexes means that dups could occur. After you updated (I assume this was via ekg-force-update), it is fixed, which is great. But it's disturbing you were missing those indexes, did you force an update before, or was this your first time?

[qingshuizheng]

But the lack of indexes means that dups could occur.

Good to know that!

did you force an update before, or was this your first time?

I did that several times, I don't remember when the first time was. More ekg-force-upgrade were done in the past 2-3 weeks.

Anyway, the db works again, and I'm gonna live with that duplicate titles.

[qingshuizheng]

Another UPDATE:

I inspected several of my notes that have dup titles via DB Browser for SQLite.app, and found these (:t t) in properties, are they valid values?

Note 1:

Note 2:

Thanks!

[ahyatt]

The (:t t) is valid, and how emacsql denotes empty property values. I can't remember exactly the reason, but I think for some reason emacsql couldn't handle the simpler "()" value there. I'm now realizing I should be upgrading these when you switch to sqlite, although that means if you switch back there might be dups again.

[qingshuizheng]

The (:t t) is valid, and how emacsql denotes empty property values.

TIL.

that means if you switch back there might be dups again.

I won't try emacsql again, the performance is terrible. It takes a thousand year to return queries. 😆

[jayrajput]

This is a good change. I recently discovered that s-k in Emacs on Macos kills the buffer. I am new to using Emacs on Macos. But I did lost my drafts. Thanks for making this change. I have also remapped s-k to bury-buffer rather than kill-buffer.

So the draft tag be automatically applied on a new capture or edit? And only on C-c C-c gets removed?

[qingshuizheng]

Indeed. I almost smashed my keyboard when I hit s-k without save.

According to the code implemented, the "draft" tag will only be pushed to the note tags if triggered by ekg-save-draft, which is binded to the save-buffer kbd. The default is C-x C-s. I haven't figured out how Andrew remove the "draft" tag yet (I haven't go deep into the commit code). But the tag will eventually be removed after C-c C-c.

[ahyatt]

Thanks!

I only removed the draft tag when setting up the edit buffer. So it just won't appear in the list of tags in the edit buffer, and therefore, next time you save it, it won't be saved.

[qingshuizheng]

Andrew has pushed some great changes in both ekg and triples recently.

ekg now has better saving mechanism, and the dup records and dup titles issues are well taken care of. The latter see these two commits in triples repo: ahyatt/triples@01f8c7c , ahyatt/triples@ddeb338 , respectively.

Thanks for addressing them. @ahyatt