Unconditionally pushdown varchar predicate to Clickhouse

[ssheikin]

ClickHouse collation is case-sensitive.
ClickHouse has same sort ordering as Trino.

Per https://clickhouse.com/docs/en/sql-reference/statements/show#show_columns
ClickHouse has no per-column collations
Clickhouse is UTF-8 encoded with byte-by-byte comparison.
https://clickhouse.com/docs/en/sql-reference/statements/select/order-by#collation-support
So exactly as trino.
https://github.com/airlift/slice/blob/2.2/src/main/java/io/airlift/slice/Slice.java#L1205
That’s why all operations on varchars may pushdown.

Description

Additional context and related issues

• More predicate pushdown in ClickHouse connector #7100
• Running branch for Clickhouse pushdowns #23460
• Improve Clickhouse pushdown. Attempt 1 #22921
• Test predicates pushdown to ClickHouse FLOAT type #23496
• Unconditionally pushdown varchar predicate to Clickhouse #23516
• Improve Clickhouse varchar equalities pushdown #23455
• Support Clickhouse varchar comparison pushdown #23558
• Improve IS NULL pushdown for ClickHouse complex expression #23459

Release notes

( ) This is not user-visible or is docs only, and no release notes are required.
( ) Release notes are required. Please propose a release note for me.
(x) Release notes are required, with the following suggested text:

# Section
* Pushdown varchar predicate to ClickHouse unconditionally. ({issue}`23516`)

[ebyhr]

ClickHouse supports collation at an index level. https://clickhouse.com/docs/en/sql-reference/statements/show#show-index
What happens if the pushed-down query uses the index?

[ssheikin]

@ebyhr IIUC DB does not use index if query condition does not match index parameters.
In case of ClickHouse, index collation is just an ordering of the values within index.

collation - The sorting of the column in the index: A if ascending, D if descending, NULL if unsorted. (Nullable(String))

if the pushed-down query uses the index

it means that ordering for index matched order requested by query and ClickHouse executes query faster.

[ebyhr]

Per https://clickhouse.com/docs/en/sql-reference/statements/show#show_columns
ClickHouse has no per-column collations

Checking SHOW COLUMNS docs is basically insufficient. We should check if the database supports collation when creating tables, starting the instance and so on.

Actually, ClickHouse supports specifying column collation for new tables:

CREATE TABLE test (x varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL) ENGINE = Memory;

It's just allowed at syntax level and it doesn't affect results as far as I tested locally, though.