This set of functions aims to implement many (but not all) of the features of the package Zetteldeft, while circumventing and eliminating any dependency on Deft, or any other external packages for that matter. It does not use any backend cache or database, preferring instead to query a directory of plaintext notes directly, treating and utilizing that directory as a sufficient database unto itself.
To that end, these functions rely, at the lowest level, on simple calls to grep, which returns lists of files, links, and tags to the Emacs completion function completing-read, from which files can be opened and links and tags can be inserted into an open buffer.
Out of the box, links are clickable buttons made with the built-in button.el. This means that links will work the same way in (almost) any major mode: fundamental-mode, text-mode, outline-mode, markdown-mode, etc. The key exception is Org-Mode, where a minor change is necessary to enable clickable zk-links. (See below for details.)
The structural simplicity of this set of functions is—one hopes, at least—in line with the structural simplicity of the so-called “Zettelkasten method,” of which much can be read in many places, including at https://www.zettelkasten.de. Ultimately, this package aims to be a lean, understandable, and eminently forkable Zettelkasten implementation for Emacs. Fork away, and make it your own.
Video demonstrations:
Notes are all kept in a single directory, set to the variable zk-directory, with no subdirectories.
Each note is a separate file, named as follows: a unique ID number followed by the title of the note followed by the file extension (set in the variable zk-file-extension), e.g. “202012091130 On the origin of species.txt”.
The primary connector between notes is the simple link, which takes the form of an ID number enclosed in double-brackets, eg, [[202012091130]]. A note’s ID number, by default, is a twelve-digit string corresponding to the date and time the note was originally created. For example, a note created on December 9th, 2020 at 11:30 will have the ID “202012091130”. Linking to such a note involves nothing more than placing the string [[202012091130]] into another note in the directory.
A key consequence of this ID and file-naming scheme is that a note’s title can change without any existing links to the note being broken, wherever they might be in the directory.
For the best experience completing filenames and links, it is highly recommended to use the orderless completion style, from the package of the same name. Another high recommendation is Vertico, for completion in the minibuffer.
The easiest way to install is from MELPA.
Or, manually add zk.el to your loadpath and include (require 'zk) in your init.el file.
At a minimum, you must set the variables zk-directory and zk-file-extension:
(setq zk-directory "~/path/to/your/zk-directory") (setq zk-file-extension "md") ;; any plaintext file extension, eg, "org" or "txt"
Once zk is loaded, call M-x zk-new-note to create a note or M-x zk-find-file to open an existing note.
(zk-setup-auto-link-buttons) in your init config. This ensures that zk-enable-link-buttons is set to t and adds zk-make-link-buttons to Emacs’s find-file-hook.(zk-setup-embark) in your init config.use-package
(use-package zk :custom (zk-directory "~/path/to/zk-directory") (zk-file-extension "md") :config (zk-setup-auto-link-buttons) (zk-setup-embark))Sample setup with
straight.el to include optional zk-consult.el functions
See Alternative Search Functions, using Consult-Grep
(use-package zk
:straight (zk :files (:defaults "zk-consult.el"))
:custom
(zk-directory "~/path/to/zk-directory")
(zk-file-extension "md")
:config
(require 'zk-consult)
(zk-setup-auto-link-buttons)
(zk-setup-embark)
(setq zk-tag-grep-function #'zk-consult-grep-tag-search
zk-grep-function #'zk-consult-grep))
Links are buttons made with the built-in package button.el: they are clickable text that work the same way in any major mode. Whether in fundamental-mode, text-mode, outline-mode, or markdown-mode, etc., clicking or pressing RET on a zk-link opens the corresponding note. The only exception is Org-Mode. (See below.) configuring clickable links in Org-Mode, see below.)
It is also possible to call the command zk-follow-link-at-point when a link is at point, or call the command zk-links-in-note to be presented with a completing-read list of all links in the current note.
In Org-Mode, links in the default format zk-link-format (an ID in double-brackets) will be treated as internal links. This means that when they are clicked, Org will, by default, look for an in-buffer heading or target that is named, or contains, the given ID. To make Org treat zk-links as zk-links and open the corresponding note, it is only necessary to advise the function org-open-at-point as follows:
(defun zk-org-try-to-follow-link (fn &optional arg)
"When 'org-open-at-point' FN fails, try 'zk-follow-link-at-point'.
Optional ARG."
(let ((org-link-search-must-match-exact-headline t))
(condition-case nil
(apply fn arg)
(error (zk-follow-link-at-point)))))
(advice-add 'org-open-at-point :around #'zk-org-try-to-follow-link)
Briefly, this function instructs org-open-at-point to try calling zk-follow-link-at-point when a link is not an internal link.
An alternative solution for using Org-Mode would be to change zk-link-format to use, for example, single brackets instead of double brackets. With this change, the default link buttons will work as expected.
Note that using Org links makes the creation of link buttons, via zk-make-link-buttons, redundant. This link button aspects of the package can be disabled by setting zk-enable-link-buttons to nil.
The companion package zk-org-link.el provides a custom Org-link type called zk, such that links will be styled [[zk:201812101245]] instead of [[201812101245]]. Using Org-links allows notes to be followed as expected, as well as exported to various formats via org-export, stored via org-store-link, and completed via org-insert-link.
The link styles cannot be combined — they are not mutually compatible. Use one style or the other. That is, either use zk-org-link.el or don’t. (I do not, but here it is anyway.)
To use org-links, include the following in your init.el:
(with-eval-after-load 'org
(with-eval-after-load 'zk
(require 'zk-org-link)))
This will set create the zk Org-link type and set necessary values for several variables. Be sure to load zk-org-link.el after zk, as the above code snippet does.
NOTE: zk-completion-at-point functionality is not available when using zk-org-link.el.
To allow link-hint.el to find zk-links, it is necessary to add a new link type, as follows:
(defun zk-link-hint--zk-link-at-point-p ()
"Return the id of the zk-link at point or nil."
(thing-at-point-looking-at (zk-link-regexp)))
(defun zk-link-hint--next-zk-link (&optional bound)
"Find the next zk-link.
Only search the range between just after the point and BOUND."
(link-hint--next-regexp zk-id-regexp bound))
(eval-when-compile
(link-hint-define-type 'zk-link
:next #'zk-link-hint--next-zk-link
:at-point-p #'zk-link-hint--zk-link-at-point-p
:open #'zk-follow-link-at-point
:copy #'kill-new))
(push 'link-hint-zk-link link-hint-types)
Calling zk-backlinks in any note presents a list, with completion, of all notes that contain at least one link to the current note.
The function zk-new-note prompts for a title and generates a unique ID number for the new note based on the current date and time. A new file with that ID and title will be created in the zk-directory.
The header of the new note is inserted by means of a function, the name of which must be set to the variable zk-new-note-header-function.
The default header function, zk-new-note-header, behaves differently depending on the context in which zk-new-note is initiated. If zk-new-note is called within an existing note, from within the zk-directory, the new note’s header will contain a backlink to that note. If zk-new-note is called from outside of the zk-directory, there are two possible behaviors, depending on the setting of the variable zk-default-backlink. If this variable is set to nil, the header of the new note will contain no backlink. If this variable is set to an ID (as a string), the header will contain a link and title corresponding with that ID. This can be useful if the directory contains a something like a “home” note or an “inbox” note.
By default, a link to the new note, along with the new note’s title, will be placed at point wherever zk-new-note was called. This behavior can be configured with the variable zk-new-note-link-insert: when set to t, a link is always inserted; when set to zk, a link is inserted only when zk-new-note is initiated inside an existing note in zk-directory; when set to ask, the user is asked whether or not a link should be inserted; when set to nil, a link is not inserted. Calling zk-new-note with a prefix-argument will insert a link regardless of setting of zk-new-note-link-insert.
By default, the date/time of a generated ID only goes to the minute, though this can be configured with the variable zk-id-time-string-format. In the default case, however, if more than one note is created in the same minute, the ID will be incremented by 1 until it is unique, allowing for rapid note creation.
Finally, a new note can be created from a selected region of text. The convention for this feature is that the first line of the region will be used as the new note’s title, while the subsequent lines will be used as the body, with the exception of a single separator line between title and body. To clarify, consider the following as the region selected swhen zk-new-note is called:
On the origin of species It is not knowledge we lack. What is missing is the courage to understand what we know and to draw conclusions.
The title of the new note in this case will be “On the origin of species.” The body will be the two sentences that follow it. The empty line separating title from body is necessary and should not be excluded.
Note: This behavior is derived from the behavior of an earlier, long-used Zettelkasten implementation and it persists here by custom only. It would be trivial to alter this function to behave perhaps more sensibly, for example by using the selected region in its entirety as the body and prompting for a title. For now, though, custom prevails.
Insert Links via FunctionCalling zk-insert-link presents a list, with completion, of all notes in the zk-directory. By default this function inserts only the link itself, like so: [[202012091130]].
To insert both a link and title, either use a prefix-argument before calling zk-insert-link or set the variable zk-link-insert-title to t, to always insert link and title together. Note that when zk-link-insert-title is set to t, calling zk-insert-link with a prefix-argument temporarily restores the default behavior and inserts the link without a title.
To be prompted with a yes-or-no query, asking whether to insert a title with the link or insert only a link by itself, set zk-link-insert-title to ask. With this setting, a prefix-argument also restores the default behavior of inserting only a link.
The format in which link and title are inserted can be configured with the variable zk-link-and-title-format.
This package includes a completion-at-point-function, zk-completion-at-point, for inserting links. Completion candidates are formatted as links followed by a title, i.e., [[202012091130]] On the origin of species, such that typing [[ will initiate completion. To enable this functionality, add zk-completion-at-point function to completion-at-point-functions, by evaluating the following:
(add-hook 'completion-at-point-functions #'zk-completion-at-point 'append)
Consider using Corfu or Company as a convenient interface for such completions.
The default search behavior of zk-search calls the built-in function lgrep to search for a regexp in all files in zk-directory. Results are presented in a grep buffer.
The function zk-find-file-by-full-text-search presents, via completing-read, a list of all files containing at least a single instance of a give search string somewhere in the body of the note. Compare this to zk-file-file which returns matches only from the filename.
There are two functions that query all notes in the zk-directory for tags in following form: #tag. One of the functions, zk-tag-search, opens a grep buffer listing all notes that contain the selected tag. The other function, zk-tag-insert, inserts the selected tag into the current buffer.
The file zk-consult.el includes two alternative functions, for use with the Consult package, that display the results using completing-read.
To use, make sure Consult is loaded, then load zk-consult.el, and set the following variables accordingly:
(setq zk-grep-function 'zk-consult-grep) (setq zk-tag-grep-function 'zk-consult-grep-tag-search)
This package includes support for Embark, both on links-at-point and in the minibuffer.
To enable Embark integration, evaluate the function zk-setup-embark. Include this function in your config file to setup Embark integration on startup.
When Embark is loaded, calling embark-act on a zk-id at point makes available the functions in the keymap zk-id-map. This is a convenient way to follow links or to search for instances of the ID in all notes using zk-search.
Calling embark-act in the minibuffer makes available the functions in zk-file-map. This is a convenient way to open notes or insert links.
Additionally, note that because the function zk-current-notes uses read-buffer by default, all Embark buffer actions are automatically available through embark-act. This makes killing open notes a snap!
Last note: adding zk-search to other Embark keymaps is a convenient way to search all notes for a given Embark target. Consider adding it to the embark-region-map, for example, with a memorable keybinding — like “z”!
The function zk-current-notes presents a list of all currently open notes. Selecting a note opens it in the current frame.
The command can be set to use custom function, however, by setting the variable zk-current-note-function to the name of a function.
One such function is available in zk-consult.el: zk-consult-current-notes presents the list of current notes as a narrowed consult-buffer-source. Note that this source can also be included in the primary consult-buffer interface by adding zk-consult-source to list consult-buffer-sources. (This is not done by default.)
The package zk-index.el is a companion to zk that offers two buffer-based interfaces for working with notes in your zk-directory.
For a video demonstration, see: https://youtu.be/7qNT87dphiA
This package is available on MELPA.
Sample setup with use-package:
(use-package zk-index :after zk :config (zk-index-setup-embark))
The function zk-index pops up a buffer listing of all note titles, each of which is a clickable button. Clicking a title will pop the note into the above window.
The ZK-Index buffer is in a major mode with a dedicated keymap:
(defvar zk-index-mode-map
(let ((map (make-sparse-keymap)))
(define-key map (kbd "n") #'zk-index-next-line)
(define-key map (kbd "p") #'zk-index-previous-line)
(define-key map (kbd "v") #'zk-index-view-note)
(define-key map (kbd "o") #'other-window)
(define-key map (kbd "f") #'zk-index-focus)
(define-key map (kbd "s") #'zk-index-search)
(define-key map (kbd "g") #'zk-index-query-refresh)
(define-key map (kbd "c") #'zk-index-current-notes)
(define-key map (kbd "i") #'zk-index-refresh)
(define-key map (kbd "S") #'zk-index-sort-size)
(define-key map (kbd "M") #'zk-index-sort-modified)
(define-key map (kbd "C") #'zk-index-sort-created)
(define-key map (kbd "RET") #'zk-index-open-note)
(define-key map (kbd "q") #'delete-window)
(make-composed-keymap map tabulated-list-mode-map))
"Keymap for ZK-Index buffer.")
The keys n and p move the point to the next/previous index item, previewing the note at point in the above window. (This previewing behavior can be disabled by setting zk-index-auto-scroll to nil.) In contrast, using C-n and C-p will move the point up and down the list without previewing notes.
Pressing v (short for for ‘view’) on an index item will open the corresponding note in read-only-mode, such that pressing q will quit the buffer and return the point to the index. Pressing RET on an index item will open the corresponding note the expected major mode.
The key f (for ‘focus’) filters notes by matching a string in the note’s TITLE. For example, pressing f and entering the string “nature” will produce an index of all notes with the word “nature” in their titles.
The focus feature is cumulative, so pressing f again and entering another string, say, “climate,” will narrow down the index down further, to notes with the words “nature” and “climate” in the title.
The key s (for ‘search’) for filters notes by matching a string in their full text. So, pressing s and entering the string “nature” will produce an index of all notes that contain the word “nature” anywhere in the note itself.
The search feature is also cumulative.
Moreover, focus and search can be combined: you can focus by title and then search by content, or the other way around.
The key i refreshes the index, canceling any filtering/narrowing, returning all notes to the list.
By default the index is sorted by time of last modification, with most recently modified notes being sorted to the top of the index. The key M (for ‘modified’) enacts this sorting method.
The key C (for ‘created’) sorts the index by time of creation, with the most recently created notes sorted to the top.
The key S (for ‘size’) sorts the index by size of note, with largest notes sorted to the top.
The feature zk-desktop allows users to select and organize groups of notes relevant to specific projects. The only necessary setup is setting a directory for saved desktops. A convenient and unobtrusive option is to simply use the zk-directory itself:
(use-package zk-desktop :after zk-index :config (zk-desktop-setup-embark) :custom (zk-desktop-directory "path/to/zk-directory"))
Think of zk-desktop as allowing you to achieve something like pulling project-specific note cards from a physical file cabinet and laying them out on a desktop in front of you, to be grouped and rearranged any way you like. In this case, however, the “desktop” is a simple plaintext file saved in the zk-directory and the “note cards” are just note titles, each a clickable button, just like in zk-index.
In contrast to zk-index, all notes on a given desktop are selected and placed there individually by the user, note-by-note, rather than en masse and programmatically. Additionally, the notes placed on the desktop can be rearranged, grouped, and commented on in-line.
It is possible to have several desktops at once, each an individual file, and each corresponding to a different project. Use the function zk-desktop-select to switch from working with one desktop to working with another.
The notes listed on in the zk-desktop buffer can be rearranged, a single note can appear more than once, and the user can type on the desktop just like in a normal buffer — for example, to create headings or simply to type notes.
A zk-desktop buffers open in fundamental-mode by default, but this can be changed by setting the variable zk-desktop-major-mode to the symbol for a major mode. Consider setting this to text-mode, outline-mode, or org-mode.
(setq zk-desktop-major-mode 'outline-mode)Adding notes to a desktop
Each method of adding notes to the currently active desktop is accomplished via the same function: zk-desktop-send-to-desktop.
When this function is called in the zk-index buffer itself, the note at point is sent to the desktop. If several notes are selected in the index, all notes in the active region are sent to the current deskop. This selection feature is usefully combined with the focus/search feature of zk-index, to allow for sending a lot of relevant notes to a desktop at once.
To enable integration with Embark, include (zk-index-setup-embark) and (zk-desktop-setup-embark) in your init config.
This setup allows all index and desktop items to be recognized as zk-id Embark targets, making available all Embark actions in the zk-id-map.
The latter adds zk-desktop-send-to-deskop to zk-id-map and zk-file-map, to facilitate sending files to desktop from the minibuffer or via embark-act in the zk-index buffer.
Use embark-select to mark candidates, including zk-links, items in zk-index, and zk-files in the minibuffer. These selected items can then be acted on via embark-act-all. For example, zk-copy-link-and-title will add to the kill-ring a nicely formatted list of links to the selected notes. Similarly, zk-insert-link will insert a nicely formatted list of links into the appropriate buffer.
When zk-index is loaded, calling embark-export on selected zk-files in the minibuffer or items in a zk-index export those items to a new ZK-Index buffer. Calling zk-index-narrow on selected items will narrow the primary ZK-Index buffer to those files.
You should! They are great. I used each one of them for a least some time, some for longer than others. At a certain point with each, however, I found that I couldn’t make them do exactly what I wanted. My sense, eventually, was that the best implementation of a Zettelkasten is the one in which a user has as much control as possible over its structure, over its behavior, and, frankly, over its future viability. At first, this primarily meant using only plaintext files — no proprietary formats, no opaque databases. Eventually, however, it also meant seeking out malleability and extensibility in the means of dealing with those plaintext files, ie, in the software.
My best experiences in this regard were with “The Archive” and, after I discovered Emacs, with “Zetteldeft.” The former is highly extensible, largely by virtue (at least at this point) of the macro editor “KeyboardMaestro,” through which one can do nearly anything with a directory of text files, in terms of editing, querying, inserting tags and links, etc. If I hadn’t fallen into Emacs, I would definitely still be using “The Archive” in combination with “KeyboardMaestro.” Little about my note-taking practices and preferences has changed since I used “The Archive.” As for “Zetteldeft,” the notable differences between it and the present package are only to be found under-the-hood, so to speak. The only reason I’m not still using it is that, over time, it became this.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4