RDA is a text-encoding format that extends CSV with two core capabilities: inline delimiter definitions and recursive nested structures.
While traditional CSV assumes an externally agreed-upon delimiter (typically a comma) and a flat table, RDA declares its encoding characters directly in its inline header and allows any field to contain either a string value or a nested RDA structure.
RDA retains CSV's lightweight, delimiter-separated design while overcoming two of its major limitations.
Rather than relying on out-of-band agreements or file extensions to know how a file is formatted, an RDA-encoded string declares its encoding characters (delimiters and escape character) in its header, which are used to encode the payload data that follows the header:
|,\|Steel Bolts,500,12.4|Copper Wire,120,8.1
In this example, the leading header (|,\) explicitly defines:
|as the record delimiter,as the field delimiter\as the escape character
A parser needs only the RDA string itself to interpret the payload according to these explicit definitions.
Standard CSV is inherently two-dimensional. In RDA, any field can contain a fully nested RDA structure, enabling rich hierarchies within a uniform syntax.
For example, to attach a structured Destination (street address, suburb, postcode) to a specific record, a third delimiter (;) is declared for the next structural level down:
|,;\|Steel Bolts,500,12.4|Copper Wire,120,8.1|Packing Foam,40,3.6,45 Dock Rd;Fremantle;6160
Here, the header specifies three positional delimiters—| for records, , for top-level fields, and ; for second-level nested fields—followed by the escape character \.
Notice that only the final record contains a nested Destination value (45 Dock Rd;Fremantle;6160); earlier records remain simple and unburdened by empty fields or extra syntax. CSV cannot express this kind of optional depth without resorting to application-specific hacks or string escaping.
Because RDA natively models nested hierarchies, it serves as a lightweight alternative to JSON and XML in data interchange scenarios.
JSON and XML embed structural metadata directly into every single object. RDA allows you to separate schema definitions from the data payload when appropriate (or include an optional header row, similar to CSV). While this trade-off reduces self-describing verbosity, it mostly eliminates redundancy over the wire, and more importantly, it allows flexibilities such as having multiple versions of records in the same container.
See the dynamic schema-version handling demo for examples.
RDA uses positional delimiters in the encoding and can separately transmit schema and data, rather than repeating key names (JSON) or opening/closing tags (XML) on every record. As a result, RDA payloads are significantly smaller—especially for uniform or semi-uniform arrays of data.
See the size comparison demo for examples.
Despite its advantageous features, RDA isn't built to replace CSV, XML, or JSON as data containers.
RDA is designed for loosely coupled, late-binding data exchange — scenarios where data structures evolve over time across disparate systems, and where hardcoding schemas into compiled code or embedding heavy metadata in payload is inflexible.
While RDA defines the foundational wire format, higher-level capabilities—such as schema evolution, late binding, dynamic object mapping, and composable components—are handled by the surrounding ecosystem.
- Check out Charian for a lightweight reference parser/encoder implementation featuring working code and practical examples to experiment with RDA's benefits in your own projects.
- Explore Snappable to see how modular, late-bound components dynamically work and collaborate inside a generic component-based computing framework.
* RDA is released under the MIT License.
** Recursive Delimited Array and RDA are trademarks of Foldda Pty Ltd.