π

Logseq from an Org-mode Point of View

Show Sidebar

As my wife is not happy to start using GNU Emacs with Org-mode as I recommended to her, I was looking for an alternative. The general requirements are those of a typical student, collecting all kinds of university- and domain-related knowledge and - I presume it will take the usual direction - contact management, information on things, household stuff, appointments, and so forth.

This article is a brief report on my personal experience from that journey. It's not a tutorial. It's not describing the full set of features of mentioned tools. This is just my remarks from an Org-mode point of view, mentioning things that caught my eye.

Many Tools to Consider

There are many knowledge management and note taking tools to choose from: personal wikis, general purpose wiki tools, note-taking software, and so forth.

I can't test all of them. I can't even take a look on a substantial part of the listed solutions. They are too many.

Obsidian is a current hype, just like Evernote and others were for a certain period. I could never use solutions like that since they are proprietary solutions which come with a large vendor lock-in I'll avoid at all cost.

A knowledge-management tool is not a hype tool I'll use for a few years until I switch to the next hyped solution. Such a tool should be a trusty companion for decades, if not until the end of my (digital) life.

So scratch Obsidian as one of the many proprietary solutions.

Joplin

The first tool I took a closer look was Joplin. As with most solutions nowadays, it also uses the Markdown lightweight markup language (LML) which is the most well-known LML but by far not the best IMHO.

Whatever, let's try it.

My first impression was that it's actually not that bad. Joplin has a manually curated hierarchy of "notebooks". Each notebook has one to many "notes" and "tasks". Within a notebook, they are listed either in a custom order or they are ordered by updated timestamp. Notes can contain Markdown list items with checkboxes.

This tool is note/todo-oriented as they seem to be the usual entity to work with.

You either see the Markdown source, or a side-by-side view of Markdown and its rendered result or only the rendered result, which is read-only.

Tags can be added via menu or shortcut. They are not visible n the note/todo views. However, they are listed within the dynamic list of tags. When visited, this dynamic view lists all notes and todos that were associated with that tag in a temporary notebook view.

You define a directory to store your data. In this directory, you see files like 834bd20fb2d8432e8b80d87865c1b75d.md. If you re-use the Markdown files with other tools, you first need to find out which file contains what note or task. Tool configuration files are stored within directories in ~/.config/joplin* on non-Windows systems.

OK, so far so good. Let's take a look at another tool.

Logseq and Its Markup

I got a recommendation to try Logseq via Mastodon.

The name "Logseq" is an abbreviation of "logbook sequence" and it's pronounced "log-seek" (and not "log-segg").

Visiting the project's website with activated NoScript, you don't see anything at all. If you allow their website to execute arbitrary JavaScript code, you'll notice that they point to their community on Discord and a forum. Instantly, my article kicks in why such forums are a bad idea. But hey, that's the current normal, I guess.

Right after playing around with the locally installed Logseq for a few minutes, it was very clear to me that Joplin is also off the list of candidates. Logseq is so much better.

First of all, Logseq is also using Markdown as the default LML. However, you can switch to Org-mode syntax, for which I coined the term Orgdown for very good reasons. The Orgdown web page also features a page where you can see a number of software tools that support Orgdown to some extend. The list is already longer than you'd probably expect!

You can clearly see on many occasions (e.g., the documentation) that Markdown is the main LML for Logseq. So far, I did not stumble on many things that were not possible with Orgdown except one Markdown table plugin. Switching to the much better Org-mode syntax is a good idea from my perspective. It's much easier to learn, type and except from a few Logseq-specific extensions, it's also very consistent in comparison to Markdown.

Simple text formatting follows the same principle as Orgdown: 

 /italic/ *bold* _underline_ +strikethrough+ ^^highlighted^^

You have noticed: the "highlighted" syntax was added by Logseq. It's where the Markdown inconsistency is visible: here, it takes two characters for starting the markup and another two for ending. Sigh.

General Principles of Logseq

You need to learn a few concepts in order to understand Logseq basics.

By selecting a local directory which will then hold your data files, you start your first "graph". You can have multiple graphs in parallel but they are not connected at all: you can't link one note of one graph to another node in a different graph. I created a Syncthing share with my wife's computer and started her graph in Logseq.

Each graph comes with certain settings. For example, you can have one or many graphs using Markdown and others with Orgdown default syntax. Unfortunately, when you switch your default LML, the existing pages are not converted. However, you can have Markdown and Orgdown mixed in a graph. Those pages can also link to each other.

Within a graph, you create "pages" which are not part of a hierarchy by default. Each page does contain a hierarchy of "blocks".

In contrast to Joplin, which is a note/todo-oriented tool, and Org-mode which is a file/heading-oriented tool, Logseq is a block-oriented tool. That means that you can add meta-data to each block and you can link to particular blocks within a page. The smallest possible entity for links and such are blocks.

A screenshot from Logseq that shows some features.
Logseq showing an example page. (click for a larger version)

On the file system, the same file looks like that: 

 * This is an *example* /block/.
 * This is another block. It features the tag #mytag which also links to [[mytag]].
 * If you like to have spaces in page titles, you need to link them like #[[page title]] or [[page title]].
 ** Blocks can have sub-blocks.
 *** You can indent them as you wish.
 *** TODO this is a scheduled task
 SCHEDULED: <2024-01-28 Sun>
 *** Strange thing: headings can occur on each level.
 *** This is a h1 heading
 :PROPERTIES:
 :heading: 1
 :END:
 *** This is no the same indentation level as the h1 heading.
 **** Here is a particular famous block. 🤩
 :PROPERTIES:
 :id: 65b67977-9737-42d8-9bbe-045f5e0a6d68
 :END:
 * This is a h2 heading
 :PROPERTIES:
 :heading: 2
 :END:
 * Here is a link to the famous block: ((65b67977-9737-42d8-9bbe-045f5e0a6d68))
 ** DONE link the famous block
 :PROPERTIES:
 :id: 65b67ef7-aee6-40c4-936e-aa78ada5d748
 :END:
 SCHEDULED: <2024-01-28 Sun>
 * You can also embed blocks or whole pages! Here, I embed the famous block:
 ** {{embed ((65b67977-9737-42d8-9bbe-045f5e0a6d68))}}
 *

As you can see, the Orgdown headings are used for each block. This way, you can add properties and other meta-data to each block. In this example, it's easy to spot that the heading level information is a special property as well as defined IDs for references. This is somewhat unusual for users of Org-mode where syntax elements like paragraphs don't "need" a heading each. This could cause issues when you're planning to directly use Orgdown files generated with Emacs Org-mode (or similar) in Logseq or vice-versa. Therefore, Logseq Org-mode syntax is not used in the same way as outside of Logseq.

By default, you always see the rendered version of the markup in Logseq. If you put your cursor within a block, you then see its "source" markup you can modify.

If you don't need indentation and you just want to write just like within a word processor or similar, you can toggle the "document mode" by clicking outside of any block (removing the current block edit mode) and type t d. This might be very interesting for many use-cases that don't need the full visual clutter of Logseq.

Todo keywords, SCHEDULED or DEADLINE works just like with Org-mode but on a block-level, not heading-level.

If you desperately need a hierarchy of pages, Logseq offers the concept of "Namespaces". If you create a page like "foo" and then a page like "foo/bar", the "bar" page is now a sub-page of "foo".

As with many Zettelkasten solutions, you also have a "Graph view" which visualizes your pages and their links. I'm not sure if I would use that often. Be warned: this view is very CPU intense. I've left this view open in background and my notebook battery was drained very fast. Therefore, use it for jumping around but don't stay within that view.

In Org-mode, you can have file-level meta-data such as file tags placed before the first heading. In Logseq, you need to add such stuff like properties in the so called "Frontmatter" which is the first block of a page. This way, you can define file tags like #+tags: foo, bar or #+ALIAS: page1, page2 which looks interesting.

One of the best things of Logseq is the effort-less linking of pages or blocks. You just type [[ followed by some search keywords. Within the search results, you choose your desired target, confirm with Return and now you've got a bi-directional link from the current position to the other page and vice versa. Linking particular blocks can be done with ((.

There is much functionality available when using the / commands. This way, you can add datestamps, timestamps, scheduled or deadlines, and much more. Most plugin functionality are accessible from here. Similarly, the < advanced commands offers access to various blocks such as quotes, center, source code. The shortcut C-k is one of the most important ones for Logseq: you can jump to any page, create new pages and much more. You can easily notice that the programmers have a very big heart for people who prefer to use the keyboard in order to be efficient and quick.

Related to links, you need to know that 

 [[foo]]

is a link to the page whose name is "foo". Alternatively, each page name is also a tag. So you can also reference the "foo" page by typing #foo. There is no difference between a tag and a link to a page except the visual representation of the link. If you would like to link to a page name that contains at least a space character, you'd need to type: 

 [[foo bar]] or #[[foo bar]]

It is very important to know that each link in Logseq is automatically a bi-directional link. So if you link from "John Doe" to the block of an event, this event also has a back-link to "John Doe". With Org-mode, you need additional packages such as org-super-links in order to get that feature. To me, this is one of the most important properties of a knowledge-management system. I think that most people who want to try a Zettelkasten system actually need bi-directional links only.

Something that even Org-mode does not offer by default are so-called embeds. If you write: 

 {{{embed [[page name]]}}}

... you are not only referencing to "page name" but also embed its content to the current position. If you just want to embed a block, you can use the 

 {{{embed ((block name))}}}

syntax instead.

And yes, three curly brackets this time, not two. (Consistency!)

Blocks can be collapsed and expanded just like with Org-mode. I did not find out a way to collapse and expand all blocks at a certain hierarchy level just like the TAB folding is working in Org-mode.

Logseq has a very capable query feature (and a builder) which offers many possibilities. You can use boolean operators to query for non-trivial stuff like: 

 {{query (and [[page1]] [[page2]] (not [[page3]] ) ) }}

Two curly brackets? → consistency!

You can base queries on properties, todo keywords, date ranges, and much more. This seems to be very powerful and allows for great re-use of content or generating summaries of some sort.

Furthermore, Logseq has a flexible template concept. You can turn any sub-hierarchy of blocks into a template. I didn't find out how to query values from the user when she's applying a template. If you do have any idea on that, drop me a line.

Logseq Plugins

Logseq comes with an easy to reach market place for plugins. You can choose from a large list of plugins. They offer great functionality. So far, I've installed following plugins:

There are also themes that can be installed just like plugins. I tested the "Quattro Theme".

Calculator

Logseq doesn't come with babel or tables using calc.

However, there is at least a Calculator which offers basic mathematics operations including assigning variables and such.

Links to Local Files

One things I was not happy with is the use-case of linking to local files.

It seems to be the case that with actions like drag and drop of, e.g., a PDF file, this file always gets copied to the data directory of the current graph. Therefore, you double the disk space and if you modify that file, you only modify the "copy" that Logseq is using and not the original file.

Of course, you are able to link to local files without copying them. Unfortunately, you always need to know the full absolute path because there is no file-picker or similar. This is nothing I'd use that way. Especially when I do have the ultimate way of linking local files in Org-mode.

Missing Things of Logseq in Comparison to Emacs Org-Mode

Just the most painful missing things from my perspective:

From my personal setup:

Nice Logseq Features That Are Missing in Emacs Org-Mode

Related: https://logseqtemplates.com/

Nice Logseq Features That Are Similar to Emacs Org-Mode

Orgdown Compatibility

As with many tools, I took a closer look on how well Orgdown syntax is supported within Logseq. With Org-mode syntax being one of the two options for page content, its Orgdown support is fairly good. I got 86 percent of syntax support for OD1.

I found some issues with lists, code, horizontal bars, tables, and similar.

Re-using Orgdown Files

Being curious, how Logseq reacts when I throw in a fairly large Orgdown files of mine which were created within my GNU Emacs Org-mode setup over the period of twelve years. I copied my current notes.org in the Logseq directory that holds the pages.

This is a file with 215523 lines holding 10640 Orgdown headings (1320 tasks and 9320 non-task headings), many tags, internal and external links.

The good news is that Logseq did somehow process its content and it did not modify the file content while doing so.

The number of pages exploded: for each Orgdown tag used, Logseq created a page in its database but not in the file system. Those "tag pages" contain all the links to the headings that are tagged using those tags.

Well, that was somehow expected.

I could not find "notes" within the page search results. Unfortunately, this is a bad term to search for as it is contained many times within the file.

Many if not all notes.org headings are somehow associated with my "Emacs Survey 2022" heading. This is a rather small and "unimportant" heading in the fourth layer of headings within notes.org. For some reason, Logseq got confused and I can't get any view of my large files to narrow down.

Searching for a node like "Windows 11", shows me the corresponding target below "Emacs Survey 2022". When selected, I always see the same node: "Emacs Survey 2022". Furthermore, many task headings are listed in a table that are not part of that sub-hierarchy. Other content from totally different sub-hierarchies is shown here as well.

This way, I could not test internal ID-links, jumping around, navigating within a large page file and so forth.

This import test did not work to my satisfaction at all.

Tons of More Stuff to Explore

You can dig in much deeper into Logseq with topics like advanced queries, Journal pages (I don't use), Flashcards, large number and freely changeable keyboard shortcuts, the whole universe of plugins, Zotero-integration (unfortunately cloud only!), iOS/Android app (Android app is not in Play Store yet), ... and so forth.

You can find many tutorial videos on YouTube and less on the WWW, unfortunately.

However, Logseq can't and will never be as flexible as the original: Org-mode. So there is zero chance that I would actually move my Orgdown data to Logseq. Not only because of the failed naïve import test using the large notes.org of mine.

Some great Logseq features, I wish I could use with Org-mode as well.

So if you do have a situation where GNU Emacs is no option for you (you should have really good arguments for that!), Logseq is a very good approximation as long as you don't plan to import large Orgdown files that were generated with the original Org-mode.

My wife will start with Logseq. Let's hope that digital note-taking will be a valuable companion also for her.

If I did get something wrong, please do drop me a line so that I can fix this here.

Links

DB Version

Update 2024-05-01: Good news and bad news:

They're working on a database version in parallel that will provide better scalability, performance and realtime collaboration (RTC). They'll charge for RTC.

Unfortunately, this seems to be the end for orgdown markup as they are implementing Markdown only now and a conversion feature later on.

Therefore, logseq is not an option for me any more and I'll need to think about a migration strategy for my wife.

Update 2025-08-01: the new DB variant (in alpha) looks as bad as anticipated. Therefore, I can not endorse use of logseq any more as I think that in long term, the non-DB version will be gone. See also my initial comment on Mastodon.

It seems to me that logseq wants to become an Obsidian ripoff with the same lock-in effects associated and limitations to the markup syntax used.

I will need to migrate the setup of my wife to something different. Emacs Org-mode does seem to be the only valid long-term choice around that fulfulls my requirements.

Related articles that link to this one:

Comment via email (persistent) or via Disqus (ephemeral) comments below: