Project Mentat is a persistent, embedded knowledge base. It draws heavily on [DataScript](https://github.com/tonsky/datascript) and [Datomic](http://datomic.com).
This project was started by Mozilla, but [is no longer being developed or actively maintained by them](https://mail.mozilla.org/pipermail/firefox-dev/2018-September/006780.html). [Their repository](https://github.com/mozilla/mentat) was marked read-only, [this fork](https://github.com/qpdb/mentat) is an attempt to revive and continue that interesting work. We owe the team at Mozilla more than words can express for inspiring us all and for this project in particular.
By abstracting away the storage schema, and by exposing change listeners outside the database (not via triggers), we hope to make domain schemas stable, and allow both the data store itself and embedding applications to use better architectures, meeting performance goals in a way that allows future evolution.
In a SQL database, the same schema definition defines everything from high-level domain relationships through to numeric field sizes in the same smear of keywords. It's difficult for someone unfamiliar with the domain to determine from such a schema what's a domain fact and what's an implementation concession — are all part numbers always 16 characters long, or are we trying to save space? — or, indeed, whether a missing constraint is deliberate or a bug.
The developer must think about foreign key constraints, compound uniqueness, and nullability. They must consider indexing, synchronizing, and stable identifiers. Most developers simply don't do enough work in SQL to get all of these things right. Storage thus becomes the specialty of a few individuals.
They both have limitations — the Mentat schema allows only for an open world (it's possible to declare friendships with people whose email isn't known), and requires validation code to enforce email string correctness — but we think that even such a tiny SQL example is harder to understand and obscures important domain decisions.
- Queries are intimately tied to structural storage choices. That not only hides the declarative domain-level meaning of the query — it's hard to tell what a query is trying to do when it's a 100-line mess of subqueries and `LEFT OUTER JOIN`s — but it also means a simple structural schema change requires auditing _every query_ for correctness.
- Developers often capture less event-shaped than they perhaps should, simply because their initial requirements don't warrant it. It's quite common to later want to [know when a fact was recorded](https://bugzilla.mozilla.org/show_bug.cgi?id=1341939), or _in which order_ two facts were recorded (particularly for migrations), or on which device an event took place… or even that a fact was _ever_ recorded and then deleted.
- Common queries are hard. Storing values only once, upserts, complicated joins, and group-wise maxima are all difficult for non-expert developers to get right.
- It's hard to evolve storage schemas. Writing a robust SQL schema migration is hard, particularly if a bad migration has ever escaped into the wild! Teams learn to fear and avoid schema changes, and eventually they ship a table called `metadata`, with three `TEXT` columns, so they never have to write a migration again. That decision pushes storage complexity into application code. (Or they start storing unversioned JSON blobs in the database…)
- It's hard to share storage with another component, let alone share _data_ with another component. Conway's Law applies: your software system will often grow to have one database per team.
- It's hard to build efficient storage and querying architectures. Materialized views require knowledge of triggers, or the implementation of bottleneck APIs. _Ad hoc_ caches are often wrong, are almost never formally designed (do you want a write-back, write-through, or write-around cache? Do you know the difference?), and often aren't reusable. The average developer, faced with a SQL database, has little choice but to build a simple table that tries to meet every need.
Mentat is not interested in that. Instead, it's focused on persistence and performance, with very little interest in immutable databases/databases as values or throwaway use.
One might say that Mentat's question is: "What if a database could store arbitrary relations, for arbitrary consumers, without them having to coordinate an up-front storage-level schema?"
Consider this a practical approach to facts, to knowledge its storage and access, much like SQLite is a practical RDBMS.
Another possible question would be: "What if we could bake some of the concepts of [CQRS and event sourcing](http://www.baeldung.com/cqrs-event-sourced-architecture-resources) into a persistent relational store, such that the transaction log itself were of value to queries?"
Some thought has been given to how databases as values — long-term references to a snapshot of the store at an instant in time — could work in this model. It's not impossible; it simply has different performance characteristics.
Datomic is a server-side, enterprise-grade data storage system. Datomic has a beautiful conceptual model. It's intended to be backed by a storage cluster, in which it keeps index chunks forever. Index chunks are replicated to peers, allowing it to run queries at the edges. Writes are serialized through a transactor.
Many of these design decisions are inapplicable to deployed desktop software; indeed, the use of multiple JVM processes makes Datomic's use in a small desktop app, or a mobile device, prohibitive.
SQLite is a traditional SQL database in most respects: schemas conflate semantic, structural, and datatype concerns, as described above; the main interface with the database is human-first textual queries; sparse and graph-structured data are 'unnatural', if not always inefficient; experimenting with and evolving data models are error-prone and complicated activities; and so on.
Mentat aims to offer many of the advantages of SQLite — single-file use, embeddability, and good performance —while building a more relaxed, reusable, and expressive data model on top.
For most `cargo` commands you can pass the `-p` argument to run the command just on that package. So, `cargo build -p mentat_query_algebrizer` will build just the "query-algebrizer" folder.
We use multiple sub-crates for Mentat for four reasons:
1. To improve incremental build times.
2. To encourage encapsulation; writing `extern crate` feels worse than just `use mod`.
3. To simplify the creation of targets that don't use certain features: _e.g._, a build with no syncing, or with no query system.
4. To allow for reuse (_e.g._, the EDN parser is essentially a separate library).
So what are they?
### Building blocks
#### `edn`
Our EDN parser. It uses `rust-peg` to parse [EDN](https://github.com/edn-format/edn), which is Clojure/Datomic's richer alternative to JSON. `edn`'s dependencies are all either for representing rich values (`chrono`, `uuid`, `ordered-float`) or for parsing (`serde`, `peg`).
This is the lowest-level Mentat crate. It collects together the following things:
- Fundamental domain-specific data structures like `ValueType` and `TypedValue`.
- Fundamental SQL-related linkages like `SQLValueType`. These encode the mapping between Mentat's types and values and their representation in our SQLite format.
- Common utilities (some in the `util` module, and others that should be moved there or broken out) like `Either`, `InternSet`, and `RcCounter`.
- Reusable lazy namespaced keywords (_e.g._, `DB_TYPE_DOUBLE`) that are used by `mentat_db` and EDN serialization of core structs.
### Types
#### `mentat_query`
This crate defines the structs and enums that are the output of the query parser and used by the translator and algebrizer. `SrcVar`, `NonIntegerConstant`, `FnArg`… these all live here.
#### `mentat_query_sql`
Similarly, this crate defines an abstract representation of a SQL query as understood by Mentat. This bridges between Mentat's types (_e.g._, `TypedValue`) and SQL concepts (`ColumnOrExpression`, `GroupBy`). It's produced by the algebrizer and consumed by the translator.
### Query processing
#### `mentat_query_algebrizer`
This is the biggest piece of the query engine. It takes a parsed query, which at this point is _independent of a database_, and combines it with the current state of the schema and data. This involves translating keywords into attributes, abstract values into concrete values with a known type, and producing an `AlgebraicQuery`, which is a representation of how a query's Datalog semantics can be satisfied as SQL table joins and constraints over Mentat's SQL schema. An algebrized query is tightly coupled with both the disk schema and the vocabulary present in the store when the work is done.
#### `mentat_query_projector`
A Datalog query _projects_ some of the variables in the query into data structures in the output. This crate takes an algebrized query and a projection list and figures out how to get values out of the running SQL query and into the right format for the consumer.
#### `mentat_query_translator`
This crate works with all of the above to turn the output of the algebrizer and projector into the data structures defined in `mentat_query_sql`.
#### `mentat_sql`
This simple crate turns those data structures into SQL text and bindings that can later be executed by `rusqlite`.
### The data layer: `mentat_db`
This is a big one: it implements the core storage logic on top of SQLite. This crate is responsible for bootstrapping new databases, transacting new data, maintaining the attribute cache, and building and updating in-memory representations of the storage schema.
### The main crate
The top-level main crate of Mentat assembles these component crates into something useful. It wraps up a connection to a database file and the associated metadata into a `Store`, and encapsulates an in-progress transaction (`InProgress`). It provides modules for programmatically writing (`entity_builder.rs`) and managing vocabulary (`vocabulary.rs`).
Sync code lives, for [referential reasons](https://engineering.linkedin.com/distributed-systems/log-what-every-software-engineer-should-know-about-real-time-datas-unifying), in a crate named `tolstoy`. This code is a work in progress; current state is a proof-of-concept implementation which largely relies on the internal transactor to make progress in most cases and comes with a basic support for timelines. See [Tolstoy's documentation](https://github.com/mozilla/mentat/tree/master/tolstoy/README.md) for details.
Mentat uses partial indices, which are available in SQLite 3.8.0 and higher. It relies on correlation between aggregate and non-aggregate columns in the output, which was added in SQLite 3.7.11.
