Fix empty table insert requirement

[smaheshwar-pltr]

Closes #631

[smaheshwar-pltr]

@Tmonster maybe, mind taking a look?

[Tmonster]

Hi @smaheshwar-pltr,

Thanks for the PR! Looks good to me. The failing test can be fixed with an order by all statement.
Also can you add a similar test for polaris with the following code?

# name: test/sql/local/irc/test_polaris_concurrent_insert.test
# description: Test concurrent inserts to a newly created table with no snapshots
# group: [insert]

require-env POLARIS_CLIENT_ID

require-env POLARIS_CLIENT_SECRET

require-env POLARIS_SERVER_AVAILABLE

require avro

require parquet

require httpfs

require iceberg

require aws

# Do not ignore 'HTTP' error messages!
set ignore_error_messages

statement ok
create secret polaris_secret (
	TYPE ICEBERG,
	CLIENT_ID '${POLARIS_CLIENT_ID}',
	CLIENT_SECRET '${POLARIS_CLIENT_SECRET}',
	ENDPOINT 'http://0.0.0.0:8181/api/catalog'
);

statement ok
attach 'quickstart_catalog' as my_datalake (
	type ICEBERG,
	ENDPOINT 'http://0.0.0.0:8181/api/catalog'
);

statement ok
DROP TABLE IF EXISTS my_datalake.default.concurrent_new_table_test;

statement ok
CREATE TABLE my_datalake.default.concurrent_new_table_test (id INTEGER, name VARCHAR);

# con1 starts a transaction and inserts
statement ok con1
begin

statement ok con1
INSERT INTO my_datalake.default.concurrent_new_table_test VALUES (1, 'con1');

# con2 starts a transaction and inserts
statement ok con2
begin

statement ok con2
INSERT INTO my_datalake.default.concurrent_new_table_test VALUES (2, 'con2');

# con2 commits first
statement ok con2
commit

# con1 should fail because con2 already added a snapshot
statement error con1
commit
----
<REGEX>:.*Requirement failed: branch main was created concurrently.*

# only con2's data is in the table
query II
SELECT * FROM my_datalake.default.concurrent_new_table_test ORDER BY id;
----
2	con2

statement ok
DROP TABLE my_datalake.default.concurrent_new_table_test;


# attach with support stage create as false
# polaris is currently file backed in tests, so folders are not automatically
# created when creating a table with stage_create=true
statement ok
attach or replace 'quickstart_catalog' as my_datalake (
	type ICEBERG,
	ENDPOINT 'http://0.0.0.0:8181/api/catalog',
	SUPPORT_NESTED_NAMESPACES true,
	support_stage_create false
);

statement ok
DROP TABLE IF EXISTS my_datalake.default.concurrent_new_table_test;

statement ok
CREATE TABLE my_datalake.default.concurrent_new_table_test (id INTEGER, name VARCHAR);

# con1 starts a transaction and inserts
statement ok con1
begin

statement ok con1
INSERT INTO my_datalake.default.concurrent_new_table_test VALUES (1, 'con1');

# con2 starts a transaction and inserts
statement ok con2
begin

statement ok con2
INSERT INTO my_datalake.default.concurrent_new_table_test VALUES (2, 'con2');

# con2 commits first
statement ok con2
commit

# con1 should fail because con2 already added a snapshot
statement error con1
commit
----
<REGEX>:.*Requirement failed: branch main was created concurrently.*

# only con2's data is in the table
query II
SELECT * FROM my_datalake.default.concurrent_new_table_test ORDER BY id;
----
2	con2

statement ok
DROP TABLE my_datalake.default.concurrent_new_table_test;

I've tested this and it passes locally on you branch.

And then for the spark test, can you also add a case where the catalog doesn't support stage create?