# name: test/sql/copy/per_thread_output.test
# description: test PER_THREAD_OUTPUT parameter for COPY
# group: [copy]

require parquet

statement ok
PRAGMA verify_parallelism;

statement ok
pragma threads=4;

statement ok
CREATE TABLE bigdata AS SELECT i AS col_a, i AS col_b FROM range(0,10000) tbl(i);

statement ok
COPY (FROM bigdata UNION ALL FROM bigdata) TO '__TEST_DIR__/per_thread_output' (FORMAT PARQUET, PER_THREAD_OUTPUT TRUE);

query I
SELECT COUNT(*) FROM PARQUET_SCAN('__TEST_DIR__/per_thread_output/*.parquet')
----
20000

# we now lazily create the files to prevent empty files from being created
# despite setting 4 threads, this may result in less than 4 files
# instead of checking we have file 0, 1, 2, and 3 like we did previously,
# we just check if we created more than one file that matches the glob pattern
query I
SELECT COUNT(*) > 1 f FROM GLOB('__TEST_DIR__/per_thread_output/data_*.parquet') ORDER BY f
----
true

# CSV also works
statement ok
COPY (FROM bigdata UNION ALL FROM bigdata) TO '__TEST_DIR__/per_thread_output_csv' (FORMAT CSV,  PER_THREAD_OUTPUT TRUE);

query I
SELECT COUNT(*) FROM read_csv('__TEST_DIR__/per_thread_output_csv/*.csv', columns={'col_a': 'INT', 'col_b' : 'INT'});
----
20000

query I
SELECT COUNT(*) >= 2 f FROM GLOB('__TEST_DIR__/per_thread_output_csv/data_*.csv') ORDER BY f
----
true

# with a trailing slash
statement ok
COPY (FROM bigdata) TO '__TEST_DIR__/per_thread_output2/' (FORMAT PARQUET, PER_THREAD_OUTPUT TRUE);

query I
SELECT COUNT(*) FROM PARQUET_SCAN('__TEST_DIR__/per_thread_output2/*.parquet')
----
10000


statement error
COPY (FROM bigdata) TO '__TEST_DIR__/per_thread_output2/' (FORMAT PARQUET, PER_THREAD_OUTPUT TRUE);
----
IO Error

# we have not added anything
query I
SELECT COUNT(*) FROM PARQUET_SCAN('__TEST_DIR__/per_thread_output2/*.parquet')
----
10000


statement error
COPY (FROM bigdata) TO '__TEST_DIR__/per_thread_output3' (FORMAT PARQUET, PER_THREAD_OUTPUT TRUE, USE_TMP_FILE TRUE);
----
Not implemented Error