pimsync-migration.7ΒΆ

pimsync-migration(7)
pimsync-migration(7) Miscellaneous Information Manual pimsync-migration(7)

pimsync-migration - migration guide from vdirsyncer to pimsync

This migration guide covers changes to be kept in mind when migrating from vdirsyncer to pimsync. It is geared towards users with a working configuration of the previous implementation.

The following features are implemented by vdirsyncer 0.19.3, but are not currently not available in pimsync:

  • HTTP Digest Auth. This needs to be implemented as a tower-service middleware. Should not be too complicated.
  • partial_sync configuration parameter. This probably won't come back in the same way. A mirror is the most likely successor. With this directive set, pimsync(1) would make B a mirror of A. In this mode, exactly one storage MUST be defined a read_only.
  • Encodings other than UTF-8. There are no concrete plans for any specific encoding at this time.
  • post_hook configuration parameter. Future versions might stream events, so external handlers can react to them instead.
  • fileignoreext configuration parameter. There are no plans to implement this feature at this time.
  • start_date, end_date and item_types configuration parameters for CalDav storages. These are planned, but no specific time frame.
  • The singlefile storage. There are no concrete plans at this time. The logic for extracting WebCal should make implementing this (in read-only mode) trivial.
  • All metadata is synchronised. Selecting individual fields is not currently possible.
  • The readonly configuration directive is not yet implemented.

The configuration format has changed. The general approach is the same but the syntax has been greatly simplified for humans to write. Technically, the format is the scfg format. It is documented in detail in pimsync.conf(5).

The following:

[pair.contacts]
a = "contacts_local"
b = "contacts_fastmail"
collections = "all"
conflict_resolution = ["nvim", "-d"]

Becomes:

pair contacts {
	storage_a contacts_local
	storage_b contacts_fastmail
	collections all
	conflict_resolution nvim -d
}

And the following:

[storage.contacts_local]
type = "filesystem"
path = "~/.local/share/contacts/"
fileext = "vcf"

Becomes this:

storage contacts_local {
	type vdir/vcard
	path ~/.local/share/contacts/
	fileext vcf
}

The prompt mechanism for fetching passwords has been dropped.

A fetch definition with a command would previously look like this:

password.fetch = ["command", "hiq", "-dFpassword", "proto=carddavs", "username=…"]

Now a block is provided to password with a single cmd directive specifying the command and its arguments.

	password {
		cmd hiq -dFpassword proto=carddavs username=…
	}

A fetch definition with a shell now uses a similar syntax:

	password {
		shell pass show communication/migadu.com | head -1
	}

Use an external prompter instead. For a GUI prompter, consider using:

password {
	cmd zenity --password
}

pimsync(1) will request passwords in sequence, so no special considerations are required to prevent concurrent prompts.

Discovering collections ahead of time is no longer required. Collections are discovered automatically if necessary.

The discover command merely prints discovered collections as a convenience for manually configuring collections. It does not affect pimsync's internal state. See pimsync(1) for details.

If a pair is configured to synchronise all collections from a and a new collection is found on storage A, then the collection will automatically be created on storage B.

For scenarios where automatic creation of collections is undesirable, individual collection should be explicitly specified instead.

It is now possible to execute a dry run, which only prints the tentative plan without executing it. The output is somewhat terse at this stage and not in its final form.

This can be used to audit new configurations and ensure that the planned actions make sense.

Filesystem collections have been renamed to vdir. They must now specify what type of items they contains.

A directory with icalendar files should specify:

type = "vdir/icalendar"

A directory with vcard files should specify:

type vdir/vcard

The vdir storage saves files as UTF-8. If another encoding is required for some scenario, please open an issue at:

https://todo.sr.ht/~whynothugo/pimsync

The fileext field for filesystem storages required a leading dot. This is no longer the case; the dot is not considered part of the extensions and should be omitted. Omit the leading dot when porting configurations.

To synchronise all collections from both sides, use:

collections all

To synchronise all collections present on a specific storage, use:

collections from b

Individual collections can be specified by id or by href:

  • The id is the name from the discovery process.
  • The href is the full path inside the storage.

Generally, using an id is recommended. Using href is intended for situations where discovery is not possible or where multiple collections have the same id.

To specify a single collection by id, use:

collection c037725e-e4fd-4b3e-b73d-d5e27d5a90a9

The above will find synchronise collections with the given id between both storages.

Two different collections can mapped on each side:

pair my_pair {
	# other fields here
	# first collection
	collection {
			alias personal
			href_a /calendars/hugo/c037725e-e4fd-4b3e-b73d-d5e27d5a90a9/
			href_b /personal/
	}
	# another collection
	collection {
			alias work
			id_a work
			href_b /calendars/work/
	}
}

Specifying collection null is no longer allowed; configuration should point to an explicitly collection instead.

The type of the http storage is now webcal, to better reflect its nature; other storages also use HTTP internally.

Due to its nature, the WebCal storage has a single collection. However, this collection has no id. A collection id must be provided for it explicitly. This simplifies configurations where it must be synchronised to another calendar with the same collection id.

storage calendars_lessons {
	type http
	collection_id lessons
	url {
		cmd hiq -dFurl proto=webcal alias=lessons
	}
}

Vdirsyncer keeps a status database in order to distinguish whether an item has changed on one side or both. Pimsync does not import this status database in any way. If you are migrating and have existing remote and local storages, makes sure that they are in sync before switching over. After the first run, pimsync will build its own status database.

Migrating with out-of-sync storages will result in conflicts during the first run.

Please report issues at:

https://todo.sr.ht/~whynothugo/vdirsyncer-rs

If you have a working vdirsyncer configuration, and this page is missing details necessary when upgrading, please reach out so this documentation can be improved.

2025-03-20