A standalone component of the Aphrodite project | discord

WIP -- while the base functionality is there, there is still work to be done to make this production ready. ETA beta release -- Nov 15.

SQLite is a foundation of offline, local-first and edge deployed software. Wouldn't it be great, however, if we could merge two or more SQLite databases together and not run into any conflicts?

This project implements CRDTs and CRRs in SQLite, allowing databases that share a common schema to merge their state together. Merges can happen between an arbitrary number of peers and all peers will eventually converge to the same state.

crsqlite works by adding metadata tables and triggers around your existing database schema. This means that you do not have to change your schema in order to get conflict resolution support -- with a few caveats around uniqueness constraints and foreign keys. See Schema Design for CRDTs & Eventual Consistency.

crsqlite exposes two APIs:

• A function extension (crsql_as_crr) to upgrade existing tables to "crrs" or "conflict free replicated relations"
  • select crsql_as_crr('table_name')
• And a virtual table (crsql_changes) to ask the database for changesets or to apply changesets from another database
  • select * from crsql_changes where version > x
  • insert into crsql_changes values ([patches receied from select on another peer])

Application code would use the function extension to enable crr support on tables. Networking code would use the crsql_changes virtual table to fetch and apply changes.

Usage looks like:

-- load the extension if it is not statically linked
.load crsqlite
.mode column
-- create tables as normal
create table foo (a primary key, b);
create table baz (a primary key, b, c, d);

-- update those tables to be crrs / crdts
select crsql_as_crr('foo');
select crsql_as_crr('baz');

-- insert some data / interact with tables as normal
insert into foo (a,b) values (1,2);
insert into baz (a,b,c,d) values ('a', 'woo', 'doo', 'daa');

-- ask for a record of what has changed
select * from crsql_changes;

table  pk   cid  val    version  site_id
-----  ---  ---  -----  -------  -------
foo    1    1    2      1
baz    'a'  1    'woo'  2
baz    'a'  2    'doo'  3
baz    'a'  3    'daa'  4

-- merge changes from a peer
insert into crsql_changes
  ("table", pk, cid, val, version, site_id)
  values
  ('foo', 5, 1, '''thing''', 5, X'7096E2D505314699A59C95FABA14ABB5');
insert into crsql_changes ("table", pk, cid, val, version, site_id)
  values
  ('baz', '''a''', 1, 123, 101, X'7096E2D505314699A59C95FABA14ABB5');

-- check that peer's changes were applied
select * from foo;
a  b
-  -----
1  2
5  thing

select * from baz;
a  b    c    d
-  ---  ---  ---
a  123  doo  daa

https://munin.uit.no/bitstream/handle/10037/22344/thesis.pdf?sequence=2

crsqlite improves upon [1] in the following ways --

• [1] stores two copies of all the data. crsqlite only keeps one by leveraging views and ISNTEAD OF triggers.
• [1] cannot compute deltas between databases without sending the full copy of each database to be compared. crsqlite only needs the logical clock (1 64bit int per peer) of a given database to determine what updates that database is missing.

https://hal.inria.fr/hal-02983557/document

crsqlite improves upon [2] in the following ways --

• [2] is implemented in a specific ORM. crsqlite runs at the db layer and allows existing applications to interface with the db as normal.
• [2] keeps a queue of all writes. This queue is drained when those writes are merged. This means that [2] can only sync changes to a single centralized node. crsqlite keeps a logical clock at each database. If a new database comes online it sends its logical clock to a peer. That peer can compute what changes are missing from the clock.

https://www.youtube.com/watch?v=DEcwa68f-jY

crsqlite improves upon [3] in the following ways --

• [3] requires retaining all history for all time (iiuc), crsqlite only needs the latest state
• [3] keeps a hloc per column, crsqlite only keeps an extra int per column and a clock per row.

[3] is better in the following way --

• crsqlite requires more work at the network layer to ensure ordered delivery and to deliver only the columns of a row that changed. [3] doesn't require any causal order to delivery and already identifies single column changes.

These projects helped improve my understanding of CRDTs on this journey --

• shelf
• tiny-merge
• Merkle-CRDT

crsqlite currently does not support:

1. Foreign key cosntraints. You can still have foreign keys (i.e. a column with an id of another row), but they can't be enforced by the db.
   1. TODO: discuss design alternatives and why this is actually not a bad thing when considering row level security.
2. Uniqueness constraints other than the primary key. The only enforceably unique column in a table should be the primary key. Other columns may be indices but they may not be unique.
   1. TODO: discuss this in much more detail.

Note: prior art [1] & [2] claim to support foreign key and uniqueness constraints. I believe their approach may be unsound and result in update loops and have not incoroprated it into crsqlite yet. If I'm wrong, I'll gladly fold their approach in.

Tables are modeled as GSets where each item has a causal length. You can call this a "CLSet". This allows us to keep all rows as well as track deletes so your application will not see deleted rows.

Rows are currently modeled as LWWW maps. I.e., each column in a row is a LWW Register.

Things to support in the future

• counter columns
• MVR (multi-value register) columns

Deltas between databases are calculated by each database keeping a version vector that represents the last time it synced with a given peer.

Every row and column in the database is associated with a lamport timestamp. This clock allows peers to ask one another for updates since the last time they communicated.

• Sharing & Privacy -- in a real-world collaborative scenario, you may not want to share your entire database with other peers. Thus, in addition to clock information, we must keep visibility information to use when computing deltas and doing replication.
• Byzantine fault tolerance -- crsqlite currently assumes friendly actors. We need to guard against malicious updates.
• Subselects -- peers may want to sync subsets of the database even if they have access to the entire thing. Compute deltas but only send those deltas that fall into the peer's provided query.

Say we have a databse schema called "Animal App." Alice, Bob and Billy all have local copies of "Animal App" on their devices. They start their day at a hostel with all of their devices synced. They then part ways, backpacking into the wilderness each with their own copy of the db.

As they see different (or maybe even the same) animals, they record their observations.

• Name
• Genus
• Species
• Habitat
• Diet
• Etc.

Some observations may even involve making updates to the prior day's (or week's) observations written by other members of the party.

At the end of the day, the group comes back together. They need to merge all of their work. crsqlite will allow Alice, Bob and Billy to merge their changes (without conflict) in a p2p fashion and converge to the same state.

Note that "without conflict" would be based on the rules of the selected CRDTs used within the schema.

Some example are --

• Tables might be grow only sets -- thus never losing an observation.
  • Or sets with a causal length so we can safely remove rows
• Table columns might be last write win (LWW) registers -- converging but throwing out earlier writes
• Table columns might be multi value (MV) registers -- keeping around all concurrent edits to a single column for users (or code) to pick and merge later.
• A column might be a counter CRDT which accumulates all observations from all parties

A description of the original design. Note that this design was only used for the prototype and we've evolved it for the production version --