[Feedback] Page: /use-timescale/latest/compression/compression-methods/

[jedwards1211]

Is it easy to find the information you need?

No

Are the instructions clear?

No

How could we improve the Timescale documentation site?

The docs on "dictionary compression" don't make it clear if and how well JSONB would be compressed. Since the example only depicts associating a number with each unique value, it seems to mean that each distinct JSONB value would be associated with a different number, which wouldn't be viable in practice, although in reality according to other sources LZ compression is used, which is fine.

LZ is a "dictionary coder" so the documentation kind of hints at this, but it should just be fully explicit.

[github-actions]

Thank you for the report. We welcome documentation contributions!

• For information about how to propose a change, see the contributing guide in our GitHub repository.
• For information on style and word usage, see the style guide

[atovpeko]

@svenklemm could you help out clarifying the docs here, please?

[svenklemm]

There are 2 layers to this, when encountering jsonb we will try our own dictionary compression and if it does not perform well we will not compress and let postgres handle this with toast compression which by default is pglz but can be changed to something more modern with default_toast_compression.

2025-11-10 09:54:04 sven@tsdb(schemata)[2429931]=> show default_toast_compression ;
 default_toast_compression
---------------------------
 lz4
(1 row)