JSON vs JSONB in PostgreSQL

PostgreSQL® offers two types of data types to handle JSON data:

• JSON stores the JSON as text, performing a validation on the correctness of the JSON syntax
• JSONB optimizes the JSON storage in a custom binary format. Therefore, on top of validating the correctness of the JSON format, time is spent to properly parse and store the content.

NOTE: more info is available in the PostgreSQL JSON types documentation page

What are the differences between JSON and JSONB

Storage

The main difference is how the data is stored: while JSON stores the data as text, JSONB optimises the information storage by creating a tree structure.

Read and Write performance

Both formats do a content validation to ensure that the input is a valid JSON document, but, while JSON only performs an overall check, with JSONB the data is then transformed from the string into the tree structure. This additional processing means than JSON is faster compared to JSONB when inserting data, due to the slimmer parsing process. On the other side, when retrieving data, specifically if only a subset of the document is selected, the JSONB performances are way better, since the binary tree can be parsed way faster compared to a scan of the full document and the selection of some subitems.

Content immutability

When using JSONB, the content of the JSON document could be modified:

• Unnecessary white spaces are removed: the white spaces outside the strings, will be removed.

E.g. in {"id": "123"} the 3 spaces between the : and the string 123 will be removed since unnecessary.

• Duplicated keys are merged: in case duplicated keys are present at the same level of a document, only the last value will be kept.

E.g. in {"id":"123", "name":"Ugo", "name":"Francesco"} only the "name":"Francesco" association will be kept. Nested structures and arrays will be stored correctly.

• Order of keys could be altered: the order of the keys can be altered when inserting and retrieving an JSON document.

E.g. the document {"id":"123", "name":"Ugo"} could be retrieved as {"name":"Ugo", "id":"123"}

Therefore if you absolutely need either to keep the white spaces, the duplicated keys or the order of the keys, you'll need to use JSON.

Indexing

Since its optimized data structure, JSONB provides advanced methods for indexing the entire document or only some parts. Read more in the dedicated PostgreSQL JSON indexing page.

Editing

With JSON PostgreSQL is storing the JSON document as text, therefore any edits to the document mean a full replacement of the content.
With JSONB, on the other hand, offers a way to just edit a particular entry in the tree, leaving the rest of the document unaltered. Read more in How to edit a JSON document in PostgreSQL?.

Summary

The following table summarizes the differences between JSON and JSONB data types in PostgreSQL.

Review all the JSON PostgreSQL use-cases listed in the main page