bug: panic when adding to a table a column as the primary key

[lmatz]

Describe the bug

dev=> create table t(v1 int, v2 int);
CREATE_TABLE

dev=> drop table tmp2;
DROP_TABLE

dev=> create table tmp2 as select * except (v1) from t;
INSERT 0 0

dev=> alter table tmp2 add column employee_id int primary key;
ERROR:  Panicked when handling the request: called `Result::unwrap()` on an `Err` value: []
This is a bug. We would appreciate a bug report at:
  https://github.com/risingwavelabs/risingwave/issues/new?labels=type%2Fbug&template=bug_report.yml
SSL connection has been closed unexpectedly

Error message/log

2024-11-29 10:56:04.860	
thread 'rw-standalone-frontend' panicked at src/frontend/src/handler/alter_table_column.rs:322:10:
2024-11-29 10:56:04.860	
called `Result::unwrap()` on an `Err` value: []
2024-11-29 10:56:04.860	
stack backtrace:
2024-11-29 10:56:04.978	
   0:     0xaaaad2dc7c74 - std::backtrace_rs::backtrace::libunwind::trace::h22abe6941b6c364e
2024-11-29 10:56:04.978	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/std/src/../../backtrace/src/backtrace/libunwind.rs:116:5
2024-11-29 10:56:04.978	
   1:     0xaaaad2dc7c74 - std::backtrace_rs::backtrace::trace_unsynchronized::h09825ba824b25c03
2024-11-29 10:56:04.978	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/std/src/../../backtrace/src/backtrace/mod.rs:66:5
2024-11-29 10:56:04.978	
   2:     0xaaaad2dc7c74 - std::sys_common::backtrace::_print_fmt::h8edb2ed98e6d043e
2024-11-29 10:56:04.978	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/std/src/sys_common/backtrace.rs:68:5
2024-11-29 10:56:04.978	
   3:     0xaaaad2dc7c74 - <std::sys_common::backtrace::_print::DisplayBacktrace as core::fmt::Display>::fmt::h59fe19f00e808c0f
2024-11-29 10:56:04.978	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/std/src/sys_common/backtrace.rs:44:22
2024-11-29 10:56:04.979	
   4:     0xaaaac82db660 - core::fmt::rt::Argument::fmt::h97fcedd5c989dd8e
2024-11-29 10:56:04.979	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/core/src/fmt/rt.rs:165:63
2024-11-29 10:56:04.979	
   5:     0xaaaac82db660 - core::fmt::write::h2ab4a0f6cb1c47e2
2024-11-29 10:56:04.979	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/core/src/fmt/mod.rs:1168:21
2024-11-29 10:56:04.979	
   6:     0xaaaad2dc135c - std::io::Write::write_fmt::h92c5bdbd600095f4
2024-11-29 10:56:04.979	
                               at /rustc/72fdf913c53dd0e75313ba83e4aa80df3f6e2871/library/std/src/io/mod.rs:1835:15
2024-11-29 10:56:04.979	
   7:     0xaaaad2dc7abc - std::sys_common::backtrace::_print::ha084ca3dc112f7ce

To Reproduce

Please use the SQL statements above

Expected behavior

No response

How did you deploy RisingWave?

No response

The version of RisingWave

No response

Additional context

This pattern is useful when we want to create a table with a primary key but without manually specifying the schema.

[BugenZhao]

I believe this issue arises because you're performing a schema change on a CREATE TABLE AS (CTAS). The table's definition isn't properly maintained, which is necessary for replacing the table plan.

Similar issue: #15334 (comment). Covered by #17472 (comment).

Adding a primary key to a normal table should be rejected gracefully.

[lmatz]

Always reject or reject under some conditions?

The doc includes "primary key" clause in the syntax:
https://docs.risingwave.com/sql/commands/sql-alter-table#clauses
but does not mention under what condition it can be used

[BugenZhao]

Always reject or reject under some conditions?

Always rejected. Primary key affects the physical distribution of the table data thus cannot be changed with instant schema change.