Logseq

The logseq preset converts between Logseq md and Logseq org. Set on one side only, it converts between a Logseq dialect and Vanilla (below).

Page properties

A page's first block of key:: value lines ↔ leading #+key: value lines.

• The block is plain text. An empty key:: counts, and its keys must be names org reads as keyword names. It may sit below a file-level drawer or a mode line.
• A keyword no key:: line can hold, such as #+CAPTION[short]:, travels as morg_keywords instead.
• Values travel verbatim, frontmatter values as written (01234 stays 01234).
• Keys are lower-cased on the way to Markdown: Logseq reads only lower case.

Flat frontmatter entries whose keys are lower case, digits, _ and - become such keywords too; other keys would not come back as written. A sequence is written as Logseq writes it (tags: [a, b] → #+tags: a, b). So YAML comes back as a key:: block after one round trip.

These stay in the frontmatter block:

• what a keyword line cannot hold: nested maps, line breaks, an item with a comma, a key that is no keyword name
• flow-style or aliased frontmatter, as a whole
• keys that act in Emacs or in export, as frontmatter they were passive data: todo, startup, options, include, setupfile, call, begin and end (which open and close a dynamic block), bibliography, cite_export, toc, index, raw export lines such as html, latex and markdown, html_*, infojs_opt, lco and the other options of org's own exporters, of ox-hugo and of org-re-reveal

The list is a denylist: other third-party exporters' keys and export metadata such as title or author do map. tags maps too: it holds Logseq's page tags.

A key:: block maps whole, acting keys included, since that is how an org page's own #+startup: travels through Markdown (ADR 0005). Its acting keys act in the org page: include:: x becomes #+include: x.

Outline

A page converts block by block, each block's content as one fragment, as Logseq parses it.

• A block is a headline of stars, a space and the content ↔ a - bullet indented one tab per level, continuation lines two spaces inside it. An empty block is the bare stars.
• :heading: N ↔ - ## ….
• A heading outside the bullets is a top-level block: Logseq writes a page's first block so if it is a heading. Other content before the first bullet is page content.

Block titles

A code block, table or rule that starts on the headline line converts as a whole. Otherwise a block's title is inline text: a title that looks like org line syntax (: a, ** a, - a, 1. a, # a) stays text, escaped in Markdown (\*\* a, 1\. a) as mldoc reads it. A block that starts with a comment has it below an empty title (* then # …), where it stays a comment.

Known limitation: Logseq reads an org title : a (or :a: b) as an empty title and a fixed-width line. morg reads it as text, as Emacs does.

Drawers and planning

• A block's property drawer ↔ key:: value lines where it was. If the content starts with it, the lines go on the headline line, as Logseq writes a query block.
• Planning lines and other drawers (:LOGBOOK:) stay as written, but for a repeated task's state log lines: bulleted - State … in org, * State … in Markdown, as Logseq writes them.
• An empty drawer is dropped.

Org blocks

An org block in a Markdown block (#+BEGIN_SRC…#+END_SRC, which Logseq reads there too) goes to org as written, as org → md writes them. Exceptions:

• A quote block's Markdown content converts. Logseq's <quote command writes one.
• A special block (#+BEGIN_TIP) stays org text, which Logseq shows as a box, not a GFM alert, which it shows as a quote.

A block's begin and end lines are written in upper case on either side (#+BEGIN_SRC sh), as Logseq writes them; Vanilla org keeps Emacs's lower case (#+begin_src).

Links and inline syntax

• A plain https://… link ↔ a bare url; a bracketed [[url]] ↔ <url>.
• An email address stays text, written bare.
• A #tag at the start of a line stays unescaped: mldoc reads \# as plain text, not a tag.
• Task markers and [#A] priorities stay text.
• Page refs [[page]] and labeled [label]([[page]]) ↔ org fuzzy links [[page][label]].
• Block refs [label](((uuid))) ↔ [[((uuid))][label]].
• ^^highlight^^ markup and hiccup ([:div …]) survive verbatim, unescaped in Markdown.

Logseq on one side only

--input-preset / --output-preset set logseq on one side only; the other side is Vanilla (ADR 0006).

Vanilla org

Vanilla org is read and written as Logseq org, which it nearly is: Logseq md ↔ Vanilla org converts as Logseq md ↔ Logseq org does, but for what Emacs reads otherwise than Logseq:

• An empty block keeps a space after its stars. Logseq org writes bare stars, which Emacs reads as text.
• A block whose first line starts an element that runs on below it (a src or other block, a table, a list, its property drawer) has that line below an empty headline. Logseq org puts it on the stars' line, where Emacs would read it as the title.
• A page whose blocks use a task marker Emacs does not know (NOW, LATER, DOING, WAIT, WAITING, IN-PROGRESS, STARTED, CANCELED, CANCELLED), and which its own #+TODO: lines do not declare, gets #+TODO: TODO NOW LATER DOING WAIT WAITING IN-PROGRESS STARTED | DONE CANCELED CANCELLED after its leading keywords in Vanilla org, and loses that line in Logseq org. Logseq reads no such line: a page's own one stays (a page property to Logseq), with a warning naming the markers Logseq shows as text.
• A block's :collapsed: true property is :VISIBILITY: folded in Vanilla org, by which Emacs folds the headline when it opens the file. Another VISIBILITY (children, content, all) stays, with a warning: Logseq has none.
• An image's size ([[image.png]]{:height 200, :width 300}, in Logseq md ![alt](image.png){:width 300}) is its #+ATTR_HTML: :height 200 :width 300 line in Vanilla org, in the order written, and back. This holds for an image line below a block's title that is a paragraph of its own. A block's title is a headline, which org does not size, so a title image keeps its size map, as it does in Vanilla md, which carries Logseq's syntax.

Translated within org (Logseq org ↔ Vanilla org), only these change. Other lines stay as written, line endings become LF and an empty page stays empty. A Vanilla headline whose title starts a list running on below it (* 3. Why? with its body indented three spaces) stays as it is on the way to Logseq org, but comes back with the title below empty stars.

What Logseq reads otherwise than Emacs stays as written, with a warning per kind; org blocks are skipped:

• a [[*heading]] or [[#custom-id]] link (labeled or not) and an [[id:…]] link without a label, which Logseq reads as refs to pages of that name
• a <<<radio>>> target, which Logseq misreads
• a #+KEY: line below the first headline (#+NAME:, #+RESULTS:), which Logseq takes for a page property

Converted to Logseq md, a [[*heading]] link (labeled or not) becomes a page ref, with that warning; the other kinds are Markdown links or text there.

Vanilla md

Vanilla md has a shape of its own. Logseq org ↔ Vanilla md:

A block's title is inline text here too, but for a table or a rule on the headline line (Block titles).

Vanilla md is read as Logseq reads a file it did not write:

• A heading is a block, nesting by heading level; the text below it, up to a list, is its content.
• Other text after a list is a block of its own.
• A list on an item's first line is its content: Markdown has no item whose content opens with a list and that has children besides.
• A reference link becomes inline before the page splits, and its definition goes.
• A footnote's definition moves to the block of its first reference.

With Obsidian on the other side, a page ref and a wikilink map onto each other ([[Page][label]] ↔ [[Page|label]]).

Within Markdown

Translated within Markdown (Logseq md ↔ Vanilla md), the outline maps as in the table. A block's content stays as written, but for its org blocks: a source or example block is fenced, a quote quoted, a query a query code block (and back: a query code block is a query block); others stay.

• Page properties are flat frontmatter keys, every key: no org is in the way to act on one. From Vanilla md, flat entries are page properties and the rest stays frontmatter.
• A rule after a Vanilla list is a block of ---.
• A block's content loses the indentation all its lines share, and text after a list is read from its own column. A tab indents to the next tab stop (4); what of it lies right of the column cut stays as spaces.
• A style option rewrites the markers it names (bullets, emphasis, strong, fences, rules) and nothing else. Logseq md keeps the - its blocks need, with a warning, and a fence stays if its code holds a run of the new marker.
• The Markdown options apply as in a conversion (ADR 0007): a definition list is written in the Spelling they name, and <dl>, <u>, <sup> and <sub> are read where interpretHtml asks. Each block holding one converts on its own; the rest stays as written. An image's #+ATTR_HTML: size line above it is written into its <img> where images = "html".

Known limitation: underline and scripts in their Markdown spelling (\_x\_, ^{x}) stay as written under "html" too, where a conversion writes <u>, <sup>, <sub>: they are org text inside Markdown text, which a user's own escapes look like.