81 lines
1.8 KiB
Markdown
81 lines
1.8 KiB
Markdown
|
# Query Engines for Non-SQL Databases
|
||
|
|
|
||
|
|
A reference for why query engines are broader than SQL.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Short answer
|
||
|
|
|
||
|
|
Yes, a query engine can be built for non-SQL databases.
|
||
|
|
|
||
|
|
SQL is only one possible query language. The broader pattern is:
|
||
|
|
|
||
|
|
- a data model exists
|
||
|
|
- users need a declarative or structured way to ask for data
|
||
|
|
- the system needs planning and execution machinery to answer those requests
|
||
|
|
|
||
|
|
So query engines are not inherently relational.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## What changes
|
||
|
|
|
||
|
|
What changes is the underlying algebra and operator set.
|
||
|
|
|
||
|
|
For example:
|
||
|
|
|
||
|
|
- relational engines center on tables, joins, filters, and aggregates
|
||
|
|
- graph engines center on nodes, edges, traversals, and pattern matching
|
||
|
|
- document engines center on nested objects, arrays, and field-path predicates
|
||
|
|
- rule engines center on facts, unification, recursion, and fixpoint evaluation
|
||
|
|
|
||
|
|
The architecture may still look familiar, but the internal operators differ.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Examples
|
||
|
|
|
||
|
|
You can meaningfully talk about query engines for:
|
||
|
|
|
||
|
|
- document databases such as MongoDB
|
||
|
|
- search systems such as Lucene or Vespa
|
||
|
|
- vector databases such as Qdrant or Weaviate
|
||
|
|
- graph databases
|
||
|
|
- Datalog or rule engines
|
||
|
|
|
||
|
|
What makes them query engines is not SQL syntax. It is that they accept structured requests and execute them efficiently over some data model.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## What stays the same
|
||
|
|
|
||
|
|
Even outside SQL, many systems still have:
|
||
|
|
|
||
|
|
1. a query language or API
|
||
|
|
2. an internal representation of the request
|
||
|
|
3. optimization or rewrite steps
|
||
|
|
4. execution against indexes or stored data
|
||
|
|
|
||
|
|
So the concept generalizes cleanly beyond relational databases.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Practical mental model
|
||
|
|
|
||
|
|
SQL engines optimize relational algebra.
|
||
|
|
|
||
|
|
Non-SQL engines optimize some other access model:
|
||
|
|
|
||
|
|
- graph traversal
|
||
|
|
- text retrieval
|
||
|
|
- vector similarity
|
||
|
|
- logical derivation
|
||
|
|
|
||
|
|
That is the main difference.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Changelog
|
||
|
|
|
||
|
|
* **April 2, 2026** -- First version created.
|