ClickHouse: format aggregate functions with combinators as function calls (countMerge, sumIf, ...)

[MunMunMiao]

Problem

In the ClickHouse dialect, the name of an aggregate function with a combinator appended (sumIf, countMerge, uniqExactIfState, ...) is not in the list of functions, because the list comes from system.functions and ClickHouse builds these names from the aggregate function and the combinators. They are treated as plain identifiers, so there is a space before the parenthesis, unlike sum(x):

SELECT sumIf(x, y > 1), countMerge(s), uniqExactIfState(id), groupArrayArray(a) FROM t

Before (15.9.0, language: 'clickhouse'):

SELECT
  sumIf (x, y > 1),
  countMerge (s),
  uniqExactIfState (id),
  groupArrayArray (a)
FROM
  t

After:

SELECT
  sumIf(x, y > 1),
  countMerge(s),
  uniqExactIfState(id),
  groupArrayArray(a)
FROM
  t

Combinators are very common in materialized views and in queries on AggregatingMergeTree tables (...State when writing, ...Merge when reading).

Change

postProcess in the ClickHouse dialect already turns SET into a function name when a parenthesis follows. This does the same for an identifier that

• is directly followed by a parenthesis, and
• is a name from the function list with one or more combinators appended.

The combinators are the ones in the documentation. Requiring the base name to be a known function keeps the change narrow: an identifier like userState(x) is formatted as before.

Not in this PR

A parametric aggregate call quantile(0.9)(x) is still formatted as quantile(0.9) (x), also for functions that are in the list. That comes from how a second parenthesis after a function call is parsed, not from the dialect, so I left it out.

Tests

Added a test for the new names. pnpm exec jest passes for all dialects (5886 passed, 1 skipped), and tsc --noEmit, prettier --check and eslint are clean.

Fixes #990.

---

This change was prepared with Claude Code (an AI coding assistant) and checked by running the full test suite locally.

[Copilot]

🟡 Changes recommended

Reclassified functions do not honor the standard uppercase functionCase option.

2 open findings

• Canonicalize function text when reclassifying identifiers · New
• Keep Window functions documentation link with its test suite · New

What changed in this PR

Adds recognition of ClickHouse aggregate-function combinators so calls format without unwanted spaces.

Changes:

• Detects known functions with appended aggregate combinators.
• Adds representative combinator-formatting coverage.

File	Description
src/​languages/​clickhouse/​clickhouse.formatter.ts	Recognizes combinator-based aggregate calls.
test/​clickhouse.test.ts	Tests representative combinator names.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[MunMunMiao]

Closing for now: I'm going to handle the formatting on my side instead. Thanks for the library.