part of the ontonomy software suite

AGPL-3.0. Anton Davydov.

csvs

A plain-text relational database.

A csvs dataset is a directory of two-column CSV files called tablets. The schema tablet _-_.csv names the relationships between collections. Data tablets hold the values.

_-_.csv              name-age.csv       name-city.csv
name,age             john,35            john,Bath
name,city            jane,36            john,London

John is 35, Jane is 36, John lives in Bath and London.

Collections only exist in relationship with another. A tablet filename is two collection names joined by a hyphen. Values are unconstrained UTF-8. A key can have multiple values (multiple records with the same key). There are no headers inside the file; the filename is the header.

Large text lives in the prose store (@/), addressed by value.

The spec has the full grammar and rules.

Implementations

• js/ JavaScript (Node.js and browser)
• rs/ Rust (library + CLI)
• c/ C (library, not at feature parity)

Query

A query is a JSON object naming a collection and optionally constraining its values:

{ "_": "name" }

returns every name record with its age and city.

{ "_": "name", "name": "john" }

returns only John's record. Values are matched as regular expressions.