This manual, written by Protesilaos Stavrou, describes the customization
options for the Emacs package called denote
(or denote.el
), and
provides every other piece of information pertinent to it.
The documentation furnished herein corresponds to stable version {{{stable-version}}}, released on {{{release-date}}}. Any reference to a newer feature which does not yet form part of the latest tagged commit, is explicitly marked as such.
Current development target is {{{development-version}}}.
- Package name (GNU ELPA):
denote
- Official manual: https://protesilaos.com/emacs/denote
- Change log: https://protesilaos.com/emacs/denote-changelog
- Git repo on SourceHut: https://git.sr.ht/~protesilaos/denote
- Mirrors:
- Mailing list: https://lists.sr.ht/~protesilaos/denote
If you are viewing the README.org version of this file, please note that the GNU ELPA machinery automatically generates an Info manual out of it.
Copyright (C) 2022 Free Software Foundation, Inc.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover Texts being “A GNU Manual,” and with the Back-Cover Texts as in (a) below. A copy of the license is included in the section entitled “GNU Free Documentation License.”
(a) The FSF’s Back-Cover Text is: “You have the freedom to copy and modify this GNU manual.”
Denote aims to be a simple-to-use, focused-in-scope, and effective note-taking tool for Emacs. It is based on the following core design principles:
- Predictability
- File names must follow a consistent and descriptive naming convention (The file-naming scheme). The file name alone should offer a clear indication of what the contents are, without reference to any other metadatum. This convention is not specific to note-taking, as it is pertinent to any form of file that is part of the user’s long-term storage (Renaming files).
- Composability
- Be a good Emacs citizen, by integrating with other
packages or built-in functionality instead of re-inventing functions
such as for filtering or greping. The author of Denote (Protesilaos,
aka “Prot”) writes ordinary notes in plain text (
.txt
), switching on demand to an Org file only when its expanded set of functionality is required for the task at hand (Points of entry). - Portability
- Notes are plain text and should remain portable. The way Denote writes file names, the front matter it includes in the note’s header, and the links it establishes must all be adequately usable with standard Unix tools. No need for a database or some specialised software. As Denote develops and this manual is fully fleshed out, there will be concrete examples on how to do the Denote-equivalent on the command-line.
- Flexibility
- Do not assume the user’s preference for a note-taking methodology. Denote is conceptually similar to the Zettelkasten Method, which you can learn more about in this detailed introduction: https://zettelkasten.de/introduction/. Notes are atomic (one file per note) and have a unique identifier. However, Denote does not enforce a particular methodology for knowledge management, such as a restricted vocabulary or mutually exclusive sets of keywords. Denote also does not check if the user writes thematically atomic notes. It is up to the user to apply the requisite rigor and/or creativity in pursuit of their preferred workflow (Writing metanotes).
- Hackability
- Denote’s code base consists of small and reusable functions. They all have documentation strings. The idea is to make it easier for users of varying levels of expertise to understand what is going on and make surgical interventions where necessary (e.g. to tweak some formatting). In this manual, we provide concrete examples on such user-level configurations (Keep a journal or diary).
Now the important part… “Denote” is the familiar word, though it also is a play on the “note” concept. Plus, we can come up with acronyms, recursive or otherwise, of increasingly dubious utility like:
- Don’t Ever Note Only The Epiphenomenal
- Denote Everything Neatly; Omit The Excesses
But we’ll let you get back to work. Don’t Eschew or Neglect your Obligations, Tasks, and Engagements.
There are five ways to write a note with Denote: invoke the denote
,
denote-type
, denote-date
, denote-subdirectory
commands, or
leverage the org-capture-templates
by setting up a template which
calls the function denote-org-capture
. We explain all of those in the
subsequent sections.
The denote
command will prompt for a title. Once that is supplied, it
will ask for keywords. The resulting note will have a file name as
already explained: The file naming scheme
The file type of the new note is determined by the user option
denote-file-type
(Front matter).
The keywords’ prompt supports minibuffer completion. Available
candidates are those defined in the user option denote-known-keywords
.
More candidates can be inferred from the names of existing notes, by
setting denote-infer-keywords
to non-nil (which is the case by
default).
Multiple keywords can be inserted by separating them with a comma (or
whatever the value of the crm-indicator
is—which should be a comma).
When the user option denote-sort-keywords
is non-nil (the default),
keywords are sorted alphabetically (technically, the sorting is done
with string-lessp
).
The interactive behaviour of the denote
command is influenced by the
user option denote-prompts
(The denote-prompts option).
The denote
command can also be called from Lisp. Read its doc string
for the technicalities.
In the interest of discoverability, denote
is also available under the
alias denote-create-note
.
The user option denote-prompts
determines how the denote
command
will behave interactively (Standard note creation).
The value is a list of symbols, which includes any of the following:
title
: Prompt for the title of the new note.keywords
: Prompts with completion for the keywords of the new note. Available candidates are those specified in the user optiondenote-known-keywords
. If the user optiondenote-infer-keywords
is non-nil, keywords in existing note file names are included in the list of candidates. Thekeywords
prompt usescompleting-read-multiple
, meaning that it can accept multiple keywords separated by a comma (or whatever the value ofcrm-sepator
is).file-type
: Prompts with completion for the file type of the new note. Available candidates are those specified in the user optiondenote-file-type
. Without this prompt,denote
uses the value ofdenote-file-type
.subdirectory
: Prompts with completion for a subdirectory in which to create the note. Available candidates are the value of the user optiondenote-directory
and all of its subdirectories. Any subdirectory must already exist: Denote will not create it.date
: Prompts for the date of the new note. It will expect an input like 2022-06-16 or a date plus time: 2022-06-16 14:30. Without thedate
prompt, thedenote
command uses thecurrent-time
.
The prompts occur in the given order.
If the value of this user option is nil, no prompts are used. The
resulting file name will consist of an identifier (i.e. the date and
time) and a supported file type extension (per denote-file-type
).
Recall that Denote’s standard file-naming scheme is defined as follows (The file-naming scheme):
DATE--TITLE__KEYWORDS.EXT
If either or both of the title
and keywords
prompts are not
included in the value of this variable, file names will be any of
those permutations:
DATE.EXT DATE--TITLE.EXT DATE__KEYWORDS.EXT
When in doubt, always include the title
and keywords
prompts.
Finally, this user option only affects the interactive use of the
denote
command (advanced users can call it from Lisp). For ad-hoc
interactive actions that do not change the default behaviour of the
denote
command, users can invoke these convenience commands:
denote-type
, denote-subdirectory
, denote-date
. They are described
in the subsequent section (Convenience commands for note creation).
Sometimes the user needs to create a note that has different
requirements from those of denote
(Standard note creation). While
this can be achieved globally by changing the denote-prompts
user
option, there are cases where an ad-hoc method is the appropriate one
(The denote-prompts option).
To this end, Denote provides the following convenience interactive commands for note creation:
- Create note by specifying file type
- The
denote-type
command creates a note while prompting for a file type.This is the equivalent to calling
denote
whendenote-prompts
is set to'(file-type title keywords)
. In practical terms, this lets you produce, say, a note in Markdown even though you normally write in Org (Standard note creation).The
denote-create-note-using-type
is an alias ofdenote-type
. - Create note using a date
- Normally, Denote reads the current date
and time to construct the unique identifier of a newly created note
(Standard note creation). Sometimes, however, the user needs to set
an explicit date+time value.
This is where the
denote-date
command comes in. It creates a note while prompting for a date. The date can be in YEAR-MONTH-DAY notation like2022-06-30
or that plus the time:2022-06-16 14:30
.This is the equivalent to calling
denote
whendenote-prompts
is set to'(date title keywords)
.The
denote-create-note-using-date
is an alias ofdenote-date
. - Create note in a specific directory
- The
denote-subdirectory
command creates a note while prompting for a subdirectory. Available candidates include the value of the variabledenote-directory
and any subdirectory thereof (Denote does not create subdirectories).This is equivalent to calling
denote
whendenote-prompts
is set to'(subdirectory title keywords)
.The
denote-create-note-in-subdirectory
is a more descriptive alias ofdenote-subdirectory
.
For integration with org-capture
, the user must first add the relevant
template. Such as:
(with-eval-after-load 'org-capture
(require 'denote-org-capture)
(add-to-list 'org-capture-templates
'("n" "New note (with Denote)" plain
(file denote-last-path)
#'denote-org-capture
:no-save t
:immediate-finish nil
:kill-buffer t
:jump-to-captured t)))
[ In the future, we might develop Denote in ways which do not require such manual intervention. More user feedback is required to identify the relevant workflows. ]
Once the template is added, it is accessed from the specified key. If,
for instance, org-capture
is bound to C-c c
, then the note creation
is initiated with C-c c n
, per the above snippet. After that, the
process is the same as with invoking denote
directly, namely: a prompt
for a title followed by a prompt for keywords (Standard note creation).
Users may prefer to leverage org-capture
in order to extend file
creation with the specifiers described in the org-capture-templates
documentation (such as to capture the active region and/or create a
hyperlink pointing to the given context). Due to the particular
file-naming scheme of Denote, which is derived dynamically, such
specifiers cannot be written directly in the template. Instead, they
have to be assigned to the user option denote-org-capture-specifiers
,
which is interpreted by the function denote-org-capture
. Example with
our default value:
(setq denote-org-capture-specifiers "%l\n%i\n%?")
Note that denote-org-capture
ignores the denote-file-type
: it always
sets the Org file extension for the created note to ensure that the
capture process works as intended, especially for the desired output of
the denote-org-capture-specifiers
.
The user option denote-directory
accepts a value that represents the
path to a directory, such as ~/Documents/notes
. Normally, the user
will have one place where they store all their notes, in which case this
arrangement shall suffice.
There is, however, the possibility to maintain separate directories of notes. By “separate”, we mean that they do not communicate with each other: no linking between them, no common keywords, nothing. Think of the scenario where one set of notes is for private use and another is for an employer. We call these separate directories “silos”.
To create silos, the user must specify a local variable at the root of
the desired directory. This is done by creating a .dir-locals.el
file, with the following contents:
;;; Directory Local Variables. For more information evaluate:
;;;
;;; (info "(emacs) Directory Variables")
((nil . ((denote-directory . default-directory))))
When inside the directory that contains this .dir-locals.el
file, all
Denote commands/functions for note creation, linking, the inference of
available keywords, et cetera will use the silo as their point of
reference. They will not read the global value of denote-directory
.
The global value of denote-directory
is read everywhere else except
the silos.
In concrete terms, this is a representation of the directory structures
(notice the .dir-locals.el
file is needed only for the silos):
;; This is the global value of 'denote-directory' (no need for a .dir-locals.el) ~/Documents/notes |-- 20210303T120534--this-is-a-test__journal_philosophy.txt |-- 20220303T120534--another-sample__journal_testing.md `-- 20220620T181255--the-third-test__keyword.org ;; A silo with notes for the employer ~/different/path/to/notes-for-employer |-- .dir-locals.el |-- 20210303T120534--this-is-a-test__conference.txt |-- 20220303T120534--another-sample__meeting.md `-- 20220620T181255--the-third-test__keyword.org ;; Another silo with notes for my volunteering ~/different/path/to/notes-for-volunteering |-- .dir-locals.el |-- 20210303T120534--this-is-a-test__activism.txt |-- 20220303T120534--another-sample__teambuilding.md `-- 20220620T181255--the-third-test__keyword.org
Denote’s file-naming scheme is not specific to notes or text files: it is useful for all sorts of files, such as multimedia and PDFs that form part of the user’s longer-term storage (The file-naming scheme). While Denote does not manage such files, it already has all the mechanisms to facilitate the task of renaming them.
To this end, we provide the denote-dired-rename-file
command. It has
a two-fold purpose: (i) to change the name of an existing file while
retaining its identifier and (ii) to write a Denote-compliant file name
for an item that was not created by denote
or related commands (such
as an image or PDF).
The denote-dired-rename-file
command will target the file at point if
it finds one in the current Dired buffer. Otherwise it prompts with
minibuffer completion for a file name. It then uses the familiar
prompts for a TITLE
and KEYWORDS
the same way the denote
command
does (Points of entry). As a final step, it asks for confirmation
before renaming the file at point, showing a message like:
Rename sample.pdf to 20220612T052900--my-sample-title__testing.pdf? (y or n)
However, if the user option denote-dired-rename-expert
is non-nil,
conduct the renaming operation outright—no questions asked.
When operating on a file that has no identifier, such as sample.pdf
,
Denote reads the file properties to retrieve its last modification time.
If the file was from a past date like 2000-11-31 it will get an
identifier starting with 20001131
followed by the time component (per
our file-naming scheme).
The file type extension (e.g. .pdf
) is read from the underlying file
and is preserved through the renaming process. Files that have no
extension are simply left without one.
Renaming only occurs relative to the current directory. Files are not moved between directories.
The final step of the denote-dired-rename-file
command is to call the
special hook denote-dired-post-rename-functions
. Functions added to
that hook must accept three arguments, as explained in its doc string.
For the time being, the only function we define is the one which updates
the underlying note’s front matter to match the new file name:
denote-dired-rewrite-front-matter
. The function takes care to only
operate on an actual note, instead of arbitrary files.
DEVELOPMENT NOTE: the denote-dired-rewrite-front-matter
needs to be
tested thoroughly. It rewrites file contents so we have to be sure it
does the right thing. To avoid any trouble, it always asks for
confirmation before performing the replacement. This confirmation
ignores denote-dired-rename-expert
for the time being, though we might
want to lift that restriction once everything works as intended.
[ denote-dired-convert-file-to-denote
is part of {{{development-version}}} ]
The command denote-dired-convert-file-to-denote
renames a file to the
Denote format while also inserting the relevant front matter to the top
of the file. The front matter is prepended unconditionally, regardless
of the file’s contents. To rename and existing note, use the command
denote-dired-rename-file
instead (Renaming files).
The only condition for converting a file to a Denote note is that it must be a regular file (i.e. not a directory, named pipe, et cetera).
When in a Dired buffer, denote-dired-convert-file-to-denote
operates
on the file at point. Otherwise it uses the minibuffer to prompt for a
file.
Once the target file is found, denote-dired-convert-file-to-denote
prompts for a title and keywords (Standard note creation). The default
title is retrieved from a line starting with a title field, depending on
the given file type (Front matter). Else, the file name is used.
Notes are stored the denote-directory
. The default path is
~/Documents/notes
. The denote-directory
can be a flat listing,
meaning that it has no subdirectories, or it can be a directory tree.
Either way, Denote takes care to only consider “notes” as valid
candidates in the relevant operations and will omit other files or
directories.
Every note produced by Denote follows this pattern (Points of entry):
DATE--TITLE__KEYWORDS.EXTENSION
The DATE
field represents the date in year-month-day format followed
by the capital letter T
(for “time”) and the current time in
hour-minute-second notation. The presentation is compact:
20220531T091625
. The DATE
serves as the unique identifier of each
note.
The TITLE
field is the title of the note, as provided by the user. It
automatically gets downcased and hyphenated. An entry about “Economics
in the Euro Area” produces an economics-in-the-euro-area
string for
the TITLE
of the file name.
The KEYWORDS
field consists of one or more entries demarcated by an
underscore (the separator is inserted automatically). Each keyword is a
string provided by the user at the relevant prompt which broadly
describes the contents of the entry. Keywords that need to be more than
one-word-long must be written with hyphens: any other character, such as
spaces or the plus sign is automatically converted into a hyphen. So
when emacs_library
appears in a file name, it is interpreted as two
distinct keywords, whereas emacs-library
is one keyword. This is
reflected in how the keywords are recorded in the note (Front matter).
While Denote supports multi-word keywords by default, the user option
denote-allow-multi-word-keywords
can be set to nil to forcibly join
all words into one, meaning that an input of word1 word2
will be
written as word1word2
.
The EXTENSION
is the file type. By default, it is .org
(org-mode
)
though the user option denote-file-type
provides support for Markdown
with YAML or TOML variants (.md
which runs markdown-mode
) and plain
text (.txt
via text-mode
). Consult its doc string for the minutia.
While files end in the .org
extension by default, the Denote code base
does not actually depend on org.el and/or its accoutrements.
Examples:
20220610T043241--initial-thoughts-on-the-zettelkasten-method__notetaking.org 20220610T062201--define-custom-org-hyperlink-type__denote_emacs_package.md 20220610T162327--on-hierarchy-and-taxis__notetaking_philosophy.txt
The different field separators, namely --
and __
introduce an
efficient way to anchor searches (such as with Emacs commands like
isearch
or from the command-line with find
and related). A query
for _word
always matches a keyword, while a regexp in the form of,
say, "\\([0-9T]+?\\)--\\(.*?\\)_"
captures the date in group \1
and
the title in \2
(test any regular expression in the current buffer by
invoking M-x re-builder
).
Features of the file-naming scheme for searching or filtering.
While Denote is an Emacs package, notes should work long-term and not depend on the functionality of a specific program. The file-naming scheme we apply guarantees that a listing is readable in a variety of contexts.
Denote has to be highly opinionated about which characters can be used
in file names and the file’s front matter in order to enforce its
file-naming scheme. The private variable denote--punctuation-regexp
holds the relevant value. In simple terms:
- What we count as “illegal characters” are converted into hyphens.
- Input for a file title is hyphenated and downcased. The original value is preserved in the note’s contents (Front matter).
- Keywords should not have spaces or other delimiters. If they do, they are converted into hyphens. Keywords are always downcased.
File names have three fields and two sets of field delimiters between them:
DATE--TITLE__KEYWORDS.EXTENSION
The first field delimiter is the double hyphen, while the second is the double underscore. These practically serve as anchors for easier searching. Consider this example:
20220621T062327--introduction-to-denote__denote_emacs.txt
You will notice that there are two matches for the word denote
: one in
the title field and another in the keywords’ field. Because of the
distinct field delimiters, if we search for -denote
we only match the
first instance while _denote
targets the second one. When sorting
through your notes, this kind of specificity is invaluable—and you get
it for free from the file names alone!
Users can get a lot of value out of this simple arrangement, even if
they have no knowledge of regular expressions. One thing to consider,
for maximum effect, is to avoid using multi-word keywords as those get
hyphenated like the title and will thus interfere with the above: either
set the user option denote-allow-multi-word-keywords
to nil or simply
insert single words at the relevant prompts.
Notes have their own “front matter”. This is a block of data at the top of the file, with no empty lines between the entries, which is automatically generated at the creation of a new note. The front matter includes the title and keywords (aka “tags” or “filetags”, depending on the file type) which the user specified at the relevant prompt, as well as the date and unique identifier, which are derived automatically.
This is how it looks for Org mode (when denote-file-type
is nil):
#+title: This is a sample note #+date: [2022-06-30 Thu 16:09] #+filetags: denote testing #+identifier: 20220630T160934
For Markdown with YAML (denote-file-type
has the markdown-yaml
value), the front matter looks like this:
--- title: "This is a sample note" date: 2022-06-30T16:09:58+03:00 tags: denote testing identifier: "20220630T160958" ---
For Markdown with TOML (denote-file-type
has the markdown-toml
value), it is:
+++ title = "This is a sample note" date = 2022-06-30T16:10:13+03:00 tags = ["denote", "testing"] identifier = "20220630T161013" +++
And for plain text (denote-file-type
has the text
value), we have
the following:
title: This is a sample note date: 2022-06-30 tags: denote testing identifier: 20220630T161028 ---------------------------
The format of the date in the front matter is controlled by the user
option denote-date-format
. When nil, Denote uses a file-type-specific
format:
- For Org, an inactive timestamp is used, such as
[2022-06-30 Wed 15:31]
. - For Markdown, the RFC3339 standard is applied:
2022-06-30T15:48:00+03:00
. - For plain text, the format is that of ISO 8601:
2022-06-30
.
If the value is a string, ignore the above and use it instead. The
string must include format specifiers for the date. These are described
in the doc string of format-time-string
..
The denote-link
command inserts a link at point to an entry specified
at the minibuffer prompt. Links are formatted depending on the file
type of current note. In Org and plain text buffers, links are
formatted thus: [[denote:IDENTIFIER][TITLE]]
. While in Markdown they
are expressed as [TITLE](denote:IDENTIFIER)
.
When denote-link
is called with a prefix argument (C-u
by default),
it formats links like [[denote:IDENTIFIER]]
. The user might prefer
its simplicity.
Inserted links are automatically buttonized and remain active for as
long as the buffer is available. In Org this is handled by the major
mode: the denote:
hyperlink type works exactly like the standard
file:
. In Markdown and plain text, Denote performs the buttonization
of those links. To buttonize links in existing files while visiting
them, the user must add this snippet to their setup (it already excludes
Org):
(add-hook 'find-file-hook #'denote-link-buttonize-buffer)
Denote has a major-mode-agnostic mechanism to collect all linked file
references in the current buffer and return them as an appropriately
formatted list. This list can then be used in interactive commands.
The denote-link-find-file
is such a command. It uses minibuffer
completion to visit a file that is linked to from the current note. The
candidates have the correct metadata, which is ideal for integration
with other standards-compliant tools (Extending Denote). For instance,
a package such as marginalia
will display accurate annotations, while
the embark
package will be able to work its magic such as in exporting
the list into a filtered Dired buffer (i.e. a familiar Dired listing
with only the files of the current minibuffer session).
The command denote-link-backlinks
produces a bespoke buffer which
displays the file name of all notes linking to the current one. Each
file name appears on its own line and is buttonized so that it performs
the action of visiting the referenced file. The backlinks’ buffer looks
like this:
Backlinks to "On being honest" (20220614T130812) ------------------------------------------------ 20220614T145606--let-this-glance-become-a-stare__journal.txt 20220616T182958--not-feeling-butterflies-in-your-stomach__journal.txt
The backlinks’ buffer is fontified by default, though the user has
access to the denote-link-fontify-backlinks
option to disable this
effect by setting its value to nil.
The placement of the backlinks’ buffer is subject to the user option
denote-link-backlinks-display-buffer-action
. Due to the nature of the
underlying display-buffer
mechanism, this inevitably is a relatively
advanced feature. By default, the backlinks’ buffer is displayed below
the current window. The doc string of our user option includes a sample
configuration that places the buffer in a left side window instead.
Reproducing it here for the sake of convenience:
(setq denote-link-backlinks-display-buffer-action
'((display-buffer-reuse-window
display-buffer-in-side-window)
(side . left)
(slot . 99)
(window-width . 0.3)))
Note that the backlinking facility uses Emacs’ built-in Xref infrastructure. On some operating systems, the user may need to add certain executables to the relevant environment variable.
Why do I get “Search failed with status 1” when I search for backlinks?
The command denote-link-add-links
adds links at point matching a
regular expression or plain string. The links are inserted as a
typographic list, such as:
- link1 - link2 - link3
Each link is formatted according to the file type of the current note,
as explained further above about the denote-link
command. The current
note is excluded from the matching entries (adding a link to itself is
pointless).
When called with a prefix argument (C-u
) denote-link-add-links
will
format all links as [[denote:IDENTIFIER]]
, hence a typographic list:
- [[denote:IDENTIFIER-1]] - [[denote:IDENTIFIER-2]] - [[denote:IDENTIFIER-3]]
Same examples of a regular expression that can be used with this command:
journal
match all files which includejournal
anywhere in their name._journal
match all files which includejournal
as a keyword.^2022.*_journal
match all file names starting with2022
and including the keywordjournal
.\.txt
match all files including.txt
. In practical terms, this only applies to the file extension, as Denote automatically removes dots (and other characters) from the base file name.
If files are created with denote-sort-keywords
as non-nil (the
default), then it is easy to write a regexp that includes multiple
keywords in alphabetic order:
_denote.*_package
match all files that include both thedenote
andpackage
keywords, in this order.\(.*denote.*package.*\)\|\(.*package.*denote.*\)
is the same as above, but out-of-order.
Remember that regexp constructs only need to be escaped once (like \|
)
when done interactively but twice when called from Lisp. What we show
above is for interactive usage.
For convenience, the denote-link
command has an alias called
denote-link-insert-link
. The denote-link-backlinks
can also be used
as denote-link-show-backlinks-buffer
. While denote-link-add-links
is aliased denote-link-insert-links-matching-regexp
. The purpose of
these aliases is to offer alternative, more descriptive names of select
commands.
A “metanote” is an entry that describes other entries who have something in common. Writing metanotes can be part of a workflow where the user periodically reviews their work in search of patterns and deeper insights. For example, you might want to read your journal entries from the past year to reflect on your experiences, evolution as a person, and the like.
The command denote-link-add-links
, which we covered extensively in the
previous section, is suited for this task (Linking notes). You will
create your metanote the way you use Denote ordinarily (metanotes may
have the metanote
keyword), write an introduction or however you want
to go about it, invoke denote-link-add-links
to cite the notes that
match the given regexp, and continue writing.
Metanotes can serve as entry points to groupings of individual notes. They are not the same as a filtered list of files, i.e. what you would do in Dired or the minibuffer where you narrow the list of notes to a given query. Metanotes contain the filtered list plus your thoughts about it. The act of purposefully grouping notes together and contemplating on their shared patterns is what adds value.
Your future self will appreciate metanotes for the function they serve in encapsulating knowledge, while current you will be equipped with the knowledge derived from the deliberate self-reflection.
One of the upsides of Denote’s file-naming scheme is the predictable
pattern it establishes, which appears as a near-tabular presentation in
a listing of notes (i.e. in Dired). The denote-dired-mode
can help
enhance this impression, by fontifying the components of the file name
to make the date (identifier) and keywords stand out.
There are two ways to set the mode. Either use it for all directories, which probably is not needed:
(require 'denote-dired)
(add-hook 'dired-mode-hook #'denote-dired-mode)
Or configure the user option denote-dired-directories
and then set up
the function denote-dired-mode-in-directories
:
(require 'denote-dired)
;; We use different ways to specify a path for demo purposes.
(setq denote-dired-directories
(list denote-directory
(thread-last denote-directory (expand-file-name "attachments"))
(expand-file-name "~/Documents/vlog")))
(add-hook 'dired-mode-hook #'denote-dired-mode-in-directories)
The faces we define are:
denote-faces-date
denote-faces-delimiter
denote-faces-extension
denote-faces-keywords
denote-faces-subdirectory
denote-faces-time
denote-faces-title
For the time being, the diredfl
package is not compatible with this
facility.
The denote-dired-mode
does not only fontify note files that were
created by Denote: it covers every file name that follows our naming
conventions (The file-naming scheme). This is particularly useful for
scenaria where, say, one wants to organise their collection of PDFs and
multimedia in a systematic way (and, perhaps, use them as attachments
for the notes Denote produces).
Denote has a dedicated minibuffer history for each one of its prompts.
This practically means that using M-p
(previous-history-element
) and
M-n
(next-history-element
) will only cycle through the relevant
record of inputs, such as your latest titles in the TITLE
prompt, and
keywords in the KEYWORDS
prompt.
The built-in savehist
library saves minibuffer histories. Sample
configuration:
(require 'savehist)
(setq savehist-file (locate-user-emacs-file "savehist"))
(setq history-length 10000)
(setq history-delete-duplicates t)
(setq savehist-save-minibuffer-history t)
(add-hook 'after-init-hook #'savehist-mode)
Denote is a tool with a narrow scope: create notes and link between them, based on the aforementioned file-naming scheme. For other common operations the user is advised to rely on standard Emacs facilities or specialised third-party packages. This section covers the details.
While there are subtle technical differences between a journal and a diary, we will consider those equivalent in the interest of brevity: they both describe a personal space that holds a record of your thoughts about your experiences and/or view of events in the world.
Suppose you are committed to writing an entry every day. Unlike what we
demonstrated before, your writing will follow a regular naming pattern.
You know that the title of the new note must always look like Tuesday
14 June 2022
and the keyword has to be journal
or diary
. As such,
you want to automate the task instead of being prompted each time, as is
the norm with denote
and the relevant commands (Points of entry).
This is easy to accomplish because denote
can be called from Lisp and
given the required arguments of TITLE
and KEYWORDS
directly. All
you need is a simple wrapper function:
(defun my-denote-journal ()
"Create an entry tagged 'journal' with the date as its title."
(interactive)
(denote
(format-time-string "%A %e %B %Y") ; format like Tuesday 14 June 2022
'("journal"))) ; multiple keywords are a list of strings: '("one" "two")
By invoking my-denote-journal
you will go straight into the newly
created note and commit to your writing outright.
Of course, you can always set up the function so that it asks for a
TITLE
but still automatically applies the journal
tag:
(defun denote-journal-with-title ()
"Create an entry tagged 'journal', while prompting for a title."
(interactive)
(denote
(denote--title-prompt) ; ask for title, instead of using human-readable date
'("journal")))
Sometimes journaling is done with the intent to hone one’s writing skills. Perhaps you are learning a new language or wish to communicate your ideas with greater clarity and precision. As with everything that requires a degree of sophistication, you have to work for it—write, write, write!
One way to test your progress is to set a timer. It helps you gauge
your output and its quality. To use a timer with Emacs, consider the
tmr
package:
(defun my-denote-journal-with-tmr ()
"Like `my-denote-journal', but also set a 10-minute timer.
The `tmr' command is part of the `tmr' package."
(interactive)
(denote
(format-time-string "%A %e %B %Y")
'("journal"))
(tmr "10" "Practice writing in my journal")) ; set 10 minute timer with a description
Once the timer elapses, stop writing and review your performance. Practice makes perfect!
[ As Denote matures, we may add hooks to control what happens before or after the creation of a new note. We shall also document more examples of tasks that can be accomplished with this package. ]
Sources for tmr
:
- Package name (GNU ELPA):
tmr
- Official manual: https://protesilaos.com/emacs/tmr
- Change log: https://protesilaos.com/emacs/denote-changelog
- Git repo on SourceHut: https://git.sr.ht/~protesilaos/tmr
- Mirrors:
- GitHub: https://github.com/protesilaos/tmr
- GitLab: https://gitlab.com/protesilaos/tmr
- Mirrors:
- Mailing list: https://lists.sr.ht/~protesilaos/tmr
Emacs’ standard file manager (or directory editor) can read a regular
expression to mark the matching files. This is the command
dired-mark-files-regexp
, which is bound to % m
by default. For
example, % m _denote
will match all files that have the denote
keyword (Features of the file-naming scheme for searching or filtering).
Once the files are matched, the user has to options: (i) narrow the list to the matching items or (ii) exclude the matching items from the list.
For the former, we want to toggle the marks by typing t
(calls the
command dired-toggle-marks
by default) and then hit the letter k
(for dired-do-kill-lines
). The remaining files are those that match
the regexp that was provided earlier.
For the latter approach of filtering out the matching items, simply
involves the use of the k
command (dired-do-kill-lines
) to omit the
marked files from the list.
These sequences can be combined to incrementally narrow the list. Note
that dired-do-kill-lines
does not delete files: it simply hides them
from the current view.
Revert to the original listing with g
(revert-buffer
).
For a convenient wrapper, consider this example:
(defvar prot-dired--limit-hist '()
"Minibuffer history for `prot-dired-limit-regexp'.")
;;;###autoload
(defun prot-dired-limit-regexp (regexp omit)
"Limit Dired to keep files matching REGEXP.
With optional OMIT argument as a prefix (\\[universal-argument]),
exclude files matching REGEXP.
Restore the buffer with \\<dired-mode-map>`\\[revert-buffer]'."
(interactive
(list
(read-regexp
(concat "Files "
(when current-prefix-arg
(propertize "NOT " 'face 'warning))
"matching PATTERN: ")
nil 'prot-dired--limit-hist)
current-prefix-arg))
(dired-mark-files-regexp regexp)
(unless omit (dired-toggle-marks))
(dired-do-kill-lines))
embark
is a remarkable package that lets you perform relevant,
context-dependent actions using a prefix key (simplifying in the
interest of brevity).
For our purposes, Embark can be used to produce a Dired listing directly
from the minibuffer. Suppose the current note has links to three other
notes. You might use the denote-link-find-file
command to pick one
via the minibuffer. But why not turn those three links into their own
Dired listing? While in the minibuffer, invoke embark-act
which you
may have already bound to C-.
and then follow it up with E
(for the
embark-export
command).
This pattern can be repeated with any list of candidates, meaning that you can narrow the list by providing some input before eventually exporting the results with Embark.
Overall, this is very powerful and you might prefer it over doing the same thing directly in Dired, since you also benefit from all the power of the minibuffer (Narrow the list of files in Dired).
Emacs provides built-in commands which are wrappers of standard Unix
tools: M-x grep
lets the user input the flags of a grep
call and
pass a regular expression to the -e
flag.
The author of Denote uses this thin wrapper instead:
(defvar prot-search--grep-hist '()
"Input history of grep searches.")
;;;###autoload
(defun prot-search-grep (regexp &optional recursive)
"Run grep for REGEXP.
Search in the current directory using `lgrep'. With optional
prefix argument (\\[universal-argument]) for RECURSIVE, run a
search starting from the current directory with `rgrep'."
(interactive
(list
(read-from-minibuffer (concat (if current-prefix-arg
(propertize "Recursive" 'face 'warning)
"Local")
" grep for PATTERN: ")
nil nil nil 'prot-search--grep-hist)
current-prefix-arg))
(unless grep-command
(grep-compute-defaults))
(if recursive
(rgrep regexp "*" default-directory)
(lgrep regexp "*" default-directory)))
Rather than maintain custom code, consider using the excellent consult
package: it provides commands such as consult-grep
and consult-find
which provide live results and are generally easier to use than the
built-in commands.
Part of the reason Denote does not reinvent existing functionality is to encourage you to learn more about Emacs. You do not need a bespoke “jump to my notes” directory because such commands do not scale well. Will you have a “jump to my downloads” then another for multimedia and so on? No.
Emacs has a built-in framework for recording persistent markers to
locations. Visit the denote-directory
(or any dir/file for that
matter) and invoke the bookmark-set
command (bound to C-x r m
by
default). It lets you create a bookmark.
The list of bookmarks can be reviewed with the bookmark-bmenu-list
command (bound to C-x r l
by default). A minibuffer interface is
available with bookmark-jump
(C-x r b
).
If you use the consult
package, its default consult-buffer
command
has the means to group together buffers, recent files, and bookmarks.
Each of those types can be narrowed to with a prefix key. The package
consult-dir
is an extension to consult
which provides useful extras
for working with directories, including bookmarks.
If you are already using consult
(which is a brilliant package), you
will probably like its consult-notes
extension. It uses the familiar
mechanisms of Consult to filter searches via a prefix key. For example:
(setq consult-notes-data-dirs
`(("Notes" ?n ,denote-directory)
("Books" ?b "~/Documents/books")))
With the above, M-x consult-notes
will list the files in those two
directories. If you type n
and space, it narrows the list to just the
notes, while b
does the same for books.
Note that consult-notes
is in its early stages of development. Expect
improvements in the near future (written on 2022-06-22 16:48 +0300).
Emacs a built-in library for treating a directory tree as a “project”.
This means that the contents of this tree are seen as part of the same
set, so commands like project-switch-to-buffer
(C-x p b
by default)
will only consider buffers in the current project (e.g. three notes that
are currently being visited).
Normally, a “project” is a directory tree whose root is under version
control. For our purposes, all you need is to navigate to the
denote-directory
(for the shell or via Dired) and use the command-line
to run this (requires the git
executable):
git init
From Dired, you can type M-!
which invokes dired-smart-shell-command
and then run the git call there.
The project can then be registered by invoking any project-related
command inside of it, such as project-find-file
(C-x p f
).
It is a good idea to keep your notes under version control, as that
gives you a history of changes for each file. We shall not delve into
the technicalities here, though suffice to note that Emacs’ built-in
version control framework or the exceptionally well-crafted magit
package will get the job done (VC can work with other backends besides
Git).
The package is available as denote
. Simply do:
M-x package-refresh-contents M-x package-install
And search for it.
GNU ELPA provides the latest stable release. Those who prefer to follow the development process in order to report bugs or suggest changes, can use the version of the package from the GNU-devel ELPA archive. Read: https://protesilaos.com/codelog/2022-05-13-emacs-elpa-devel/.
Assuming your Emacs files are found in ~/.emacs.d/
, execute the
following commands in a shell prompt:
cd ~/.emacs.d
# Create a directory for manually-installed packages
mkdir manual-packages
# Go to the new directory
cd manual-packages
# Clone this repo, naming it "denote"
git clone https://git.sr.ht/~protesilaos/denote denote
Finally, in your init.el
(or equivalent) evaluate this:
;; Make Elisp files in that directory available to the user.
(add-to-list 'load-path "~/.emacs.d/manual-packages/denote")
Everything is in place to set up the package.
(require 'denote)
;; Remember to check the doc strings of those variables.
(setq denote-directory (expand-file-name "~/Documents/notes/"))
(setq denote-known-keywords '("emacs" "philosophy" "politics" "economics"))
(setq denote-infer-keywords t)
(setq denote-sort-keywords t)
(setq denote-file-type nil) ; Org is the default, set others here
(setq denote-prompts '(title keywords))
;; We allow multi-word keywords by default. The author's personal
;; preference is for single-word keywords for a more rigid workflow.
(setq denote-allow-multi-word-keywords t)
(setq denote-date-format nil) ; read doc string
;; You will not need to `require' all those individually once the
;; package is available.
(require 'denote-retrieve)
(require 'denote-link)
;; By default, we fontify backlinks in their bespoke buffer.
(setq denote-link-fontify-backlinks t)
;; Also see `denote-link-backlinks-display-buffer-action' which is a bit
;; advanced.
;; If you use Markdown or plain text files (Org renders links as buttons
;; right away)
(add-hook 'find-file-hook #'denote-link-buttonize-buffer)
(require 'denote-dired)
(setq denote-dired-rename-expert nil)
;; We use different ways to specify a path for demo purposes.
(setq denote-dired-directories
(list denote-directory
(thread-last denote-directory (expand-file-name "attachments"))
(expand-file-name "~/Documents/books")))
;; Generic (great if you rename files Denote-style in lots of places):
;; (add-hook 'dired-mode-hook #'denote-dired-mode)
;;
;; OR if only want it in `denote-dired-directories':
(add-hook 'dired-mode-hook #'denote-dired-mode-in-directories)
;; Here is a custom, user-level command from one of the examples we
;; showed in this manual. We define it here and add it to a key binding
;; below.
(defun my-denote-journal ()
"Create an entry tagged 'journal', while prompting for a title."
(interactive)
(denote
(denote--title-prompt)
'("journal")))
;; Denote does not define any key bindings. This is for the user to
;; decide. For example:
(let ((map global-map))
(define-key map (kbd "C-c n j") #'my-denote-journal) ; our custom command
(define-key map (kbd "C-c n n") #'denote)
(define-key map (kbd "C-c n N") #'denote-type)
(define-key map (kbd "C-c n d") #'denote-date)
(define-key map (kbd "C-c n s") #'denote-subdirectory)
;; If you intend to use Denote with a variety of file types, it is
;; easier to bind the link-related commands to the `global-map', as
;; shown here. Otherwise follow the same pattern for `org-mode-map',
;; `markdown-mode-map', and/or `text-mode-map'.
(define-key map (kbd "C-c n i") #'denote-link) ; "insert" mnemonic
(define-key map (kbd "C-c n I") #'denote-link-add-links)
(define-key map (kbd "C-c n l") #'denote-link-find-file) ; "list" links
(define-key map (kbd "C-c n b") #'denote-link-backlinks)
;; Note that `denote-dired-rename-file' can work from any context, not
;; just Dired bufffers. That is why we bind it here to the
;; `global-map'.
(define-key map (kbd "C-c n r") #'denote-dired-rename-file))
(with-eval-after-load 'org-capture
(require 'denote-org-capture)
(setq denote-org-capture-specifiers "%l\n%i\n%?")
(add-to-list 'org-capture-templates
'("n" "New note (with denote.el)" plain
(file denote-last-path)
#'denote-org-capture
:no-save t
:immediate-finish nil
:kill-buffer t
:jump-to-captured t)))
Denote is a GNU ELPA package. As such, any significant change to the code requires copyright assignment to the Free Software Foundation (more below).
You do not need to be a programmer to contribute to this package. Sharing an idea or describing a workflow is equally helpful, as it teaches us something we may not know and might be able to cover either by extending Denote or expanding this manual (Things to do). If you prefer to write a blog post, make sure you share it with us: we can add a section herein referencing all such articles. Everyone gets acknowledged (Acknowledgements). There is no such thing as an “insignificant contribution”—they all matter.
- Package name (GNU ELPA):
denote
- Official manual: https://protesilaos.com/emacs/denote
- Change log: https://protesilaos.com/emacs/denote-changelog
- Git repo on SourceHut: https://git.sr.ht/~protesilaos/denote
- Mirrors:
- Mailing list: https://lists.sr.ht/~protesilaos/denote
If our public media are not suitable, you are welcome to contact me (Protesilaos) in private: https://protesilaos.com/contact.
Copyright assignment is a prerequisite to sharing code. It is a simple process. Check the request form below (please adapt it accordingly). You must write an email to the address mentioned in the form and then wait for the FSF to send you a legal agreement. Sign the document and file it back to them. This could all happen via email and take about a week. You are encouraged to go through this process. You only need to do it once. It will allow you to make contributions to Emacs in general.
Please email the following information to assign@gnu.org, and we will send you the assignment form for your past and future changes. Please use your full legal name (in ASCII characters) as the subject line of the message. REQUEST: SEND FORM FOR PAST AND FUTURE CHANGES [What is the name of the program or package you're contributing to?] GNU Emacs [Did you copy any files or text written by someone else in these changes? Even if that material is free software, we need to know about it.] Copied a few snippets from the same files I edited. Their author, Protesilaos Stavrou, has already assigned copyright to the Free Software Foundation. [Do you have an employer who might have a basis to claim to own your changes? Do you attend a school which might make such a claim?] [For the copyright registration, what country are you a citizen of?] [What year were you born?] [Please write your email address here.] [Please write your postal address here.] [Which files have you changed so far, and which new files have you written so far?]
Denote should work well for what is described in this manual. Though we can always do better. These are some of the tasks that are planned for the future and which you might want to help with (Contributing).
This is a non-exhaustive list and you are always welcome to either report or work on something else.
- [ ] Ensure integration between
denote:
links and Embark. - [ ] Add command that expands the identifier in links to a full name.
- [ ] Add command that rewrites full names in links, if they are invalid.
- [ ] Consider completion-at-point after
denote:
links. - [ ] Support mutually-exclusive sets of tags.
These are just ideas. We need to consider the pros and cons in each case and act accordingly.
What follows is a list of Emacs packages for note-taking. I (Protesilaos) have not used any of them, as I was manually applying my file-naming scheme beforehand and by the time those packages were available I was already hacking on the predecessor of Denote as a means of learning Emacs Lisp (a package which I called “Unassuming Sidenotes of Little Significance”, aka “USLS” which is pronounced as “U-S-L-S” or “useless”). As such, I cannot comment at length on the differences between Denote and each of those packages, beside what I gather from their documentation.
- org-roam
- The de facto standard in the Emacs milieu—and rightly so! It has a massive community, is featureful, and should be an excellent companion to anyone who is invested in the Org ecosystem and/or knows what “Roam” is (I don’t). It has been explained to me that Org Roam uses a database to store a cache about your notes. It otherwise uses standard Org files. The cache helps refer to the same node through aliases which can provide lots of options. Personally, I follow a single-topic-per-note approach, so anything beyond that is overkill. If the database is only for a cache, then maybe that has no downside, though I am careful with any kind of specialised program as it creates a dependency. If you ask me about database software in particular, I have no idea how to use one, let alone debug it or retrieve data from it if something goes awry (I could learn, but that is beside the point).
- zk (or zk.el)
- Reading its documentation makes me think that this is Denote’s sibling—the two projects have a lot of things in common, including the preference to rely on plain files and standard tools. The core difference is that Denote has a strict file-naming scheme. Other differences in available features are, in principle, matters of style or circumstance: both packages can have them. As its initials imply, ZK enables a zettelkasten-like workflow. It does not enforce it though, letting the user adapt the method to their needs and requirements.
- zettelkasten
- This is another one of Denote’s relatives, at least
insofar as the goal of simplicity is concerned. The major difference
is that according to its documentation “the name of the file that is
created is just a unique ID”. This is not consistent with our
file-naming scheme which is all about making sense of your files by
their name alone and being able to visually parse a listing of them
without any kind of specialised tool (e.g.
ls -l
orls -C
on the command-line from inside thedenote-directory
give you a human-readable set of files names, whilefind * -maxdepth 0 -type f
is another approach). - zetteldeft
- This is a zettelkasten note-taking system built on top
of the
deft
package. Deft provides a search interface to a directory, in this case the one holding the user’szetteldeft
notes. Denote has no such dependency and is not opinionated about how the user prefers to search/access their notes: use Dired, Grep, theconsult
package, or whatever else you already have set up for all things Emacs, not just your notes.
Searching through M-x list-packages
for “zettel” brings up more
matches. zetteldesk
is an extension to Org Roam and, as such, I
cannot possibly know what Org Roam truly misses and what the added-value
of this package is. neuron-mode
builds on top of an external program
called neuron
, which I have never used.
Searching for “note” gives us a few more results. notes-mode
has
precious little documentation and I cannot tell what it actually does
(as I said in my presentation for LibrePlanet 2022, inadequate docs are
a bug). side-notes
differs from what we try to do with Denote, as it
basically gives you the means to record your thoughts about some other
project you are working on and keep them on the side: so it and Denote
should not be mutually exclusive.
If I missed something, please let me know.
This section covers blog posts from the Emacs community on the matter of note-taking. They may reference some of the packages covered in the previous section or provide their custom code (Alternatives to Denote). The list is unsorted.
- José Antonio Ortega Ruiz (aka “jao”) explains a note-taking method that is simple like Denote but differs in other ways. An interesting approach overall: https://jao.io/blog/2022-06-19-simple-note-taking.html.
- Jethro Kuan (the main
org-roam
developer) explains their note-taking techniques: https://jethrokuan.github.io/org-roam-guide/. Good ideas all round, regardless of the package/code you choose to use.
[ Development note: help expand this list. ]
I (Protesilaos) answer some questions I have received or might get. It is assumed that you have read the rest of this manual: I will not go into the specifics of how Denote works.
I wrote Denote because I was using a variant of Denote’s file-naming
scheme before I was even an Emacs user (I switched to Emacs from
Tmux+Vim+CLI in the summer of 2019). I was originally inspired by
Jekyll, the static site generator, which I started using for my website
in 2016 (was on WordPress before). Jekyll’s files follow the
YYYY-MM-DD-TITLE.md
pattern. I liked its efficiency relative to the
unstructured mess I had before. Eventually, I started using that scheme
outside the confines of my website’s source code. Over time I refined
it and here we are.
Note-taking is something I take very seriously, as I am a prolific
writer (just check my website, which only reveals the tip of the
iceberg). As such, I need a program that does exactly what I want and
which I know how to extend. I originally tried to use Org capture
templates to create new files with a Denote-style file-naming scheme but
never managed to achieve it. Maybe because org-capture
has some
hard-coded assumptions or I simply am not competent enough to hack on
core Org facilities. Whatever the case, an alternative was in order.
The existence of PACKAGE is never a good reason for me not to conduct my own experiments for recreational, educational, or practical purposes. When the question arises of “why not contribute to PACKAGE instead?” the answer is that without me experimenting in the first place, I would lack the skills for such a task. Furthermore, contributing to another package does not guarantee I get what I want in terms of workflow.
Whether you should use Denote or not is another matter altogether: choose whatever you want.
I think Org is one of Emacs’ killer apps. I also believe it is not the right tool for every job. When I write notes, I want to focus on writing. Nothing more. I thus have no need for stuff like org-babel, scheduling to-do items, clocking time, and so on. The more “mental dependencies” you add to your workflow, the heavier the burden you carry and the less focused you are on the task at hand: there is always that temptation to tweak the markup, tinker with some syntactic construct, obsess about what ought to be irrelevant to writing as such.
In technical terms, I also am not fond of Org’s code base (I understand
why it is the way it is—just commenting on the fact). Ever tried to
read it? You will routinely find functions that are tens-to-hundreds of
lines long and have all sorts of special casing. As I am not a
programmer and only learnt to write Elisp through trial and error, I
have no confidence in my ability to make Org do what I want at that
level, hence denote
instead of org-denote
or something.
Perhaps the master programmer is one who can deal with complexity and keep adding to it. I am of the opposite view, as language—code included—is at its communicative best when it is clear and accessible.
Make no mistake: I use Org for the agenda and also to write technical documentation that needs to be exported to various formats, including this very manual.
My notes form part of my longer-term storage. I do not want to have to rely on a special program to be able to read them or filter them. Unix is universal, at least as far as I am concerned.
Denote streamlines some tasks and makes things easier in general, which is consistent with how Emacs provides a layer of interactivity on top of Unix. Still, Denote’s utilities can, in principle, be implemented as POSIX shell scripts (minus the Emacs-specific parts like fontification in Dired or the buttonization of links).
Portability matters. For example, in the future I might own a smartphone, so I prefer not to require Emacs, Org, or some other executable to access my files on the go.
Furthermore, I might want to share those files with someone. If I make Emacs a requirement, I am limiting my circle to a handful of relatively advanced users.
Please don’t misinterpret this: I am using Emacs full-time for my computing and maintain a growing list of packages for it. This is just me thinking long-term.
I have read that Org favours the latter method. If true, I strongly disagree with it because of the implicit dependency it introduces and the way it favours machine-friendliness over human-readability in terms of accessing information. Notes are long-term storage. I might want to access them on (i) some device with limited features, (ii) print on paper, (iii) share with another person who is not a tech wizard.
There are good arguments for few large files, but all either prioritize machine-friendliness or presuppose the use of sophisticated tools like Emacs+Org.
Good luck using less
on a generic TTY to read a file with a zillion
words, headings, sub-headings, sub-sub-headings, property drawers, and
other constructs! You will not get the otherwise wonderful folding of
headings the way you do in Emacs—do not take such features for
granted.
My point is that notes should be atomic to help the user—and potentially the user’s family, friends, acquaintances—make sense of them in a wide range of scenaria. The more program-agnostic your file is, the better for you and/or everyone else you might share your writings with.
Human-readability means that we optimize for what matters to us. If (a)
you are the only one who will ever read your notes, (b) always have
access to good software like Emacs+Org, (c) do not care about printing
on paper, then Denote’s model is not for you. Maybe you need to tweak
some org-capture
template to append a new entry to one mega file (I do
that for my Org agenda, by the way, as I explained before about using
the right tool for the job).
I have not tested it, but assume that yes, many files will slow down the
agenda. Consider looking into one of Denote’s alternatives, with
org-roam
being the obvious choice (Alternatives to Denote).
Or, if you want my opinion, decouple your longer-term storage from your ephemeral to-do list: Denote (and others) can be used for the former, while you let standard Org work splendidly for the latter—that is what I do, anyway.
Denote does not sort files and will not reinvent tools that handle such functionality. This is the job of the file manager or command-line executable that lists files.
I encourage you to read the manpage of the ls
executable. It will
help you in general, while it applies to Emacs as well via Dired. The
gist is that you can update the ls
flags that Dired uses on-the-fly:
type C-u M-x dired-sort-toggle-or-edit
(C-u s
by default) and append
--sort=time
at the prompt. To reverse the order, add the -r
flag.
The user option dired-listing-switches
sets your default preference.
Denote does not insert any meta data or heading pertaining to edits in the file. I am of the view that these either do not scale well or are not descriptive enough. Suppose you use a “lastmod” heading with a timestamp: which lines where edited and what did the change amount to?
This is where an external program can be helpful. Use a Version Control
System, such as Git, to keep track of all your notes. Every time you
add a new file, record the addition. Same for post-creation edits.
Your VCS will let you review the history of those changes. For
instance, Emacs’ built-in version control framework has a command that
produces a log of changes for the current file: M-x vc-print-log
,
bound to C-x v l
by default. From there one can access the
corresponding diff output (use M-x describe-mode
(C-h m
) in an
unfamiliar buffer to learn more about it). With Git in particular,
Emacs users have the option of the all-round excellent magit
package.
In short: let Denote (or equivalent) create notes and link between them, the file manager organise and provide access to files, search programs deal with searching and narrowing, and version control software handle the tracking of changes.
Denote uses Emacs’ Xref to find backlinks. Xref requires xargs
and
one of grep
or ripgrep
, depending on your configuration.
This is usually not an issue on *nix systems, but the necessary
executables are not available on Windows Emacs distributions. Please
ensure that you have both xargs
and either grep
or ripgrep
available within your PATH
environment variable.
If you have git
on Windows installed, then you may use the following
code (adjust the git’s installation path if necessary):
(setenv "PATH" (concat (getenv "PATH") ";" "C:\\Program Files\\Git\\usr\\bin"))
Denote is meant to be a collective effort. Every bit of help matters.
- Author/maintainer
- Protesilaos Stavrou.
- Contributions to code or the manual
- Benjamin Kästner, Damien Cassou, Jack Baty, Jean-Philippe Gagné Guay, Kaushal Modi, Stefan Monnier.
- Ideas and/or user feedback
- Abin Simon, Alan Schmitt, Alfredo Borrás, Benjamin Kästner, Colin McLear, Damien Cassou, Frank Ehmsen, Jack Baty, Kaushal Modi, M. Hadi Timachi, Peter Prevos, Shreyas Ragavan, Summer Emacs, Sven Seebeck, Ypot, pRot0ta1p.
Special thanks to Peter Povinec who helped refine the file-naming scheme, which is the cornerstone of this project.