fix(mysql): keep unknown prepare placeholders

[discord9]

I hereby agree to the terms of the GreptimeDB CLA.

Refer to a related PR or issue link (optional)

Fixes #8142.

Related history:

• fix: mysql prepare limit&offset param #5734 added a fix for inferring placeholder types in LIMIT.
• feat: add a fallback parameter type inference by reading cast type #7712 refactored the planner path and the LIMIT ? type inference regressed.
• fix(mysql): infer LIMIT placeholders in prepare #8149 restores the narrow LIMIT/OFFSET ? inference behavior.
• This PR fixes the broader MySQL prepared-statement protocol issue: placeholders must be advertised even when their concrete type is unknown.

What's changed and what's your intention?

This PR fixes MySQL prepared statements whose placeholders are present in the SQL but do not have concrete types inferred by the planner.

Before this change, the MySQL prepare response only advertised placeholders with concrete inferred types. If DataFusion returned None for a placeholder type, prepared_params() skipped it, so the client could see fewer parameters than the SQL actually contains and fail before execution with an argument count mismatch.

The changes are:

• Count real prepared-statement placeholders from the SQL AST and return one MySQL parameter column per counted placeholder.
• Use inferred concrete types when available and NULL metadata for unknown placeholder types.
• Cache SqlPlan::Plan only when all placeholders have concrete inferred types; otherwise keep the statement fallback path.
• Preserve the original raw SQL in the fallback path instead of relying on AST to_string() round-trips.
• Replace fallback parameters by collecting placeholder spans from the transformed AST and replacing those original-query ranges from right to left.
• Guard span replacement by verifying each span maps back to a real ? in the original SQL.
• Format string/binary parameters in the fallback path with mysql_common::Value::as_sql(false).
• Add tests for unknown placeholders, mixed known/unknown placeholders, comments/literals, UTF-8 and CRLF span handling, replacement length shifts, NULL, binary bytes, and LIMIT ?.

This does not change public APIs, schemas, or persisted data.

PR Checklist

Please convert it to a draft if some of the following conditions are not met.

• I have written the necessary rustdoc comments.
• I have added the necessary unit tests and integration tests.
• This PR requires documentation updates.
• API changes are backward compatible.
• Schema or data changes are backward compatible.

Tests:

• cargo fmt --check
• cargo check -p servers
• cargo test -p servers --lib mysql::handler::tests
• cargo test -p servers --features testing --test mod mysql::mysql_server_test::test_query_prepared

[gemini-code-assist]

Code Review

This pull request enhances the MySQL handler to correctly support prepared statements with untyped placeholders, ensuring they are properly reported to clients by using a default null datatype when inference fails. It also introduces a SQL string escaping utility for the statement fallback path and includes comprehensive unit and integration tests. Review feedback focuses on optimizing the string escaping logic to reduce allocations and improving variable naming to avoid shadowing and enhance code clarity.

[Copilot]

Pull request overview

This PR fixes MySQL prepared-statement behavior so that the server advertises all ? placeholders found in the SQL (even when DataFusion cannot infer a concrete parameter type), preventing client-side “expected N arguments, got M” errors. It also hardens the statement-fallback execution path by replacing placeholders using AST spans and emitting properly escaped SQL literals for bytes/strings.

Changes:

• Count placeholders from the parsed SQL AST and always return one MySQL parameter column per placeholder, using NULL type metadata when the type is unknown.
• Avoid caching SqlPlan::Plan unless all placeholders have inferred concrete types; otherwise keep the statement-fallback path and preserve raw SQL.
• Improve fallback parameter substitution by replacing ? ranges from right-to-left using AST spans, and escape bytes/strings via mysql_common::Value::as_sql(false); add extensive tests.

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/servers/tests/mysql/mysql_server_test.rs	Adds integration coverage for unknown placeholders, mixed known/unknown placeholders, comments/literals, UTF-8/CRLF spans, binary bytes, NULL, and LIMIT placeholders.
src/servers/src/mysql/helper.rs	Adds placeholder counting and placeholder span collection utilities used by prepare/execute.
src/servers/src/mysql/handler.rs	Updates prepare/execute logic to advertise unknown placeholders, gate plan caching on full typing, and replace fallback params by span; uses mysql_common for SQL literal escaping.
src/servers/Cargo.toml	Adds mysql_common dependency.
Cargo.lock	Locks mysql_common version resolution.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[killme2008]

It's better to add some black-box tests to https://github.com/GreptimeTeam/greptimedb-tests

@discord9

[killme2008]

LGTM