How to use specialized codecs for nested columns?

[schneefux]

I am trying to create a nested table with a Gorilla encoded column but I get a syntax error:

ClickHouse client version 20.5.2.7.
Connecting to localhost:9000 as user default.
Connected to ClickHouse server version 20.5.2 revision 54435.

clickhouse-db :) create table nested_test (id UInt8, Attrs Nested(Key String, Value UInt8 Codec(Gorilla))) ENGINE=MergeTree() ORDER BY id;

Syntax error: failed at position 74:
create table nested_test (id UInt8, Attrs Nested(Key String, Value UInt8 Codec(Gorilla))) ENGINE=MergeTree() ORDER BY id;
Expected one of: Comma, token

The following works, but I do not understand what it does:

create table nested_test (id UInt8, Attrs Nested(Key String, Value UInt8) Codec(Gorilla)) ENGINE=MergeTree() ORDER BY id;

The documentation does not specify how Array or Nested columns work with custom codecs.
Can you clarify? 😊

[den-crane]

CREATE TABLE nested_test
(
    `id` UInt8,
    `Attrs` Nested(    Key String,     Value UInt8) CODEC(Gorilla)
)
ENGINE = MergeTree()
ORDER BY id


DESCRIBE TABLE nested_test

┌─name────────┬─type──────────┬─default_type─┬─default_expression─┬─comment─┬─codec_expression─┬─ttl_expression─┐
│ id          │ UInt8         │              │                    │         │                  │                │
│ Attrs.Key   │ Array(String) │              │                    │         │ Gorilla          │                │
│ Attrs.Value │ Array(UInt8)  │              │                    │         │ Gorilla          │                │
└─────────────┴───────────────┴──────────────┴────────────────────┴─────────┴──────────────────┴────────────────┘

Nested is a syntax sugar for CREATE TABLE

WA (the same result as Nested):

CREATE TABLE nested_test
(
    `id` UInt8,
    `Attrs.Key` Array(String),
    `Attrs.Value` UInt8 CODEC(Gorilla)
)
ENGINE = MergeTree()
ORDER BY id

BTW Gorilla is good for Float types, try Delta for UInt8.

[darkleaf]

How does codec work with an array? Codec is applied to two neighboring values in the same array, isn’t it? Or is it applied to values with the same index in to neighboring arrays?

[alexey-milovidov]

See #12551 and #15089

[alexey-milovidov]

Codec is applied to two neighboring values in the same array, isn’t it?

Almost this. It is applied to the sequence of all elements of arrays if placed one after the other.
For example, if we have two rows with [1, 2, 3], [1, 2], the codec will be applied to the sequence 1, 2, 3, 1, 2.