The problem¶
A PostgreSQL row lives in a database page. A large value does not fit well in that row. PostgreSQL can compress the value and move it to a TOAST table.
The TOAST table stores the value in numbered chunks. The table row keeps a small pointer to those chunks.
A normal field read¶
PostgreSQL's jsonb operators work on the jsonb value. For a large value,
the database first restores the value from TOAST. If the value is compressed,
the database also decompresses it. The operator can then find the requested
field.
This behavior is general and correct. Its cost can be high when the query needs a small field from a large document.
The research question¶
An uncompressed jsonb value has a binary directory before its data. The
directory describes the positions of its keys and values. A reader can use
that directory to find the byte range of one value.
The project asks whether a reader can use this path:
Version 0.1.0 does not implement this path. It contains the installable extension shell and documents the question.
The trade-off¶
Uncompressed storage can make a partial read possible. It also uses more disk space and can write more data during an update. A complete evaluation must account for three costs:
- read time and bytes read;
- table and index size;
- update time and write-ahead log volume.
Version 0.1.0 contains no benchmark result or performance claim.