Difference between JSON and JSONB in PostgreSQL

What Happens When Storing JSON

When you insert text into a column of type JSON, PostgreSQL performs the following steps:

1. Syntax Validation (Minimal Parsing)

   The most crucial step is validating the syntax. PostgreSQL must confirm that the input string is a well-formed JSON document according to the JSON standard (RFC 7159). This involves checking for things like:

   • Matching brackets ({} and []).
   • Correct placement of colons (:) and commas (,).
   • Properly quoted keys and strings.
   • Valid JSON primitive types (numbers, booleans, null).

   If this check fails, the insertion will be rejected. This minimal parsing is necessary to ensure the column actually contains JSON, not just arbitrary text.

2. Storage as Exact Text

   After validation, the key characteristic of JSON storage kicks in: the exact input text is stored verbatim.

   • NO canonicalization of data structure.
   • NO removal of whitespace.
   • NO reordering of keys.
   • NO removal of duplicate keys.

The stored value is simply a string of characters.

What Happens When Storing JSON

Below transformations are done when storing data in JSONB type

- canonicalization of data structure.
- removal of whitespace.
- reordering of keys.
- removal of duplicate keys.

Comparison of Transformations on Write

Feature	JSON (Text Storage)	JSONB (Binary Storage)
Validation	Minimal: Only checks syntax compliance.	Full: Checks syntax and semantic properties (e.g., number range, valid Unicode).
Whitespace	Preserved (stored as-is).	Removed (stripped during parsing).
Key Order	Preserved (stored as-is).	Canonicalized (may be sorted/reordered).
Duplicate Keys	Preserved (stored as-is).	Eliminated (only the last value is kept).
Data Type Mapping	None (values remain as part of the text string).	Performed (values are mapped to native PostgreSQL types like int, text, etc.).