# Importing from Logseq

> Bring a Logseq graph into nooklet, from either the classic file-based Logseq or the newer database (DB) version, and what carries over from each.

Source: https://nooklet.danielalder.cz/docs/importing-from-logseq

nooklet imports both kinds of Logseq graph with the same command. It looks at the folder you give
it, says which kind it found, and reads it accordingly.

```sh
pnpm nooklet import <folder> --data <nooklet data dir>
```

The import copies; it never writes to the Logseq folder. Run it with Logseq closed or open.

## Which folder to point at

| Logseq version | How to recognise it | Point the importer at |
|---|---|---|
| **File graph** (classic Logseq) | a folder with `pages/`, `journals/`, `logseq/config.edn` | that folder |
| **DB version** | a folder (usually `~/logseq/graphs/<name>/`) with `db.sqlite`, `assets/` and `mirror/markdown/` | that folder, the one holding `db.sqlite` |

For the DB version, pages come from its **Markdown Mirror**, so turn the mirror on first: in Logseq
(desktop), Settings → Markdown Mirror, then "Regenerate full mirror". The importer reads
`db.sqlite` for what the mirror leaves out. It reads a copy, so a running Logseq holding the
database is fine. Pointing at `mirror/` or `mirror/markdown/` also works; the importer finds the
root itself. A DB graph with no mirror is refused with a message saying how to turn it on.

## What carries over

| | File graph | DB version |
|---|---|---|
| Pages, journals, nesting, block order | yes | yes, from the mirror |
| Block refs `((…))` and page refs `[[…]]` | yes | yes. Refs the mirror leaves as raw ids are resolved from the database |
| Tasks (TODO/DOING/DONE…) | yes | yes |
| SCHEDULED / DEADLINE | yes | yes, the exact date and time from the database |
| Pasted images and files | yes, copied from `assets/` | yes. The database says which picture each line stands for |
| Page and block properties | yes | yes. A property whose value is a set of blocks stays as child blocks |
| Favourites | yes, from `config.edn` | yes, from the database |
| Tags | yes | yes, as `#tag` in the text |
| Journal title format | suggested in Settings | suggested in Settings |

## What does not carry over yet

| | File graph | DB version |
|---|---|---|
| Whiteboards | no | no |
| Built-in pages (property and class pages) | n/a | no; they are Logseq's own, not your notes |
| Property *types* and class schemas | n/a | no; values come over as text |
| Embeds of blocks | as written | expanded inline by the mirror, so they arrive as copies |
| Pre-2022 "dot" namespace file names (`a.b.md`) | no | n/a |

The import prints a summary. For a DB-version graph it includes how many images were matched,
how many lines were ambiguous and left as text (with a warning for each), and how many refs could
not be resolved. Ambiguous means the importer could not pin a line to one picture by position, and
the title was not unique. It never guesses.

Running the import again skips pages that already exist.