Skip to content

leostera/pachadb

Repository files navigation

PachaDB

Pacha is an embeddable, immutable, graph, edge database.

Pacha is an edge database. This means you can run it on edge providers like Cloduflare, and it'll scale all over the world without you doing much.

Pacha is a graph database. This means that the data in Pacha is stored as tiny little facts that form a huge graph of knowledge. Like this:

  • Obi-wan is the master of Anakin
  • Anakin is the master of Ahsoka

Obi-wan here is an entity. "Is the master of" is what we call a relation, and then Anakin and Ahsoka are values. Notice that any entity can be referenced as a value.

Pacha is an embeddable database and you can run it everywhere:

  • on the browser
  • on your backend with a file system or in-memory
  • and even as a standalone

Pacha uses Datalog instead of SQL for queries. Since the data in Pacha forms a Graph, Datalog is a much easier way to query for it. For example:

(?master is-master-of ?padawan) returns all the masters with all their padawans.

But we can make much more complex queries, like this one. Here are trying to find all the jedi masters with pupils that had encounters with the sith, where the sith was Darth Maul, ONLY if the pupil has already become a jedi master too:

(?who has-rank jedi-master
 ?who is-master-of ?padawan
 ?padawan fought-in ?fight
 ?fight enemy-was darth-maul
 ?padawan has-rank jedi-master)

The equivalent SQL for this would involve several joins, and a lot more writing! Something like this:

SELECT j.id, jp.id, sf.id from jedis as j  
 JOIN jedis_fights as jf ON jf.jedi_id = j.id
 JOIN jedis_masters as jm ON jm.master_id = j.id
 JOIN jedis_masters as jp ON jp.padawan_id = j.id
 JOIN siths_fights as sf ON sf.fight_id = jf.fight_id
 JOIN siths as s ON sf.sith_id = s.id
 WHERE j.has_rank = "jedi-master"
   AND e.has_name = "darth-maul"
   AND jp.has_rank = "jedi-master"

Pacha is an immutable database. This means that as you add information, nothing is lost! This allows you to query your data over time.

In the datalog queries we did before, we can specify an as-of marker to go back in time and see the results of this query then.