should be it

external/duckdb/test/sql/sample/table_samples/basic_sample_tests.test

@@ -0,0 +1,88 @@
+# name: test/sql/sample/table_samples/basic_sample_tests.test
+# group: [table_samples]
+
+mode skip
+
+# currently require fixed vector size since the "randomness" of the sample depends on
+# the vector size. If the vector size decreases, the randomness of the sample decreases
+# This is especially noticeable for small tables and their samples
+require vector_size 2048
+
+statement ok
+PRAGMA enable_verification
+
+load __TEST_DIR__/test_samples_basic.db
+
+query I
+select count(*) from range(100000) using sample (10000);
+----
+10000
+
+query I
+select count(*) from range(100) using sample (10);
+----
+10
+
+query I
+select count(*) from range(205000) using sample (10000);
+----
+10000
+
+statement ok
+create table t1 as select range a from range(204800);
+
+statement ok
+select * from duckdb_table_sample('t1');
+
+statement ok
+create or replace table t1 as select range a from range(1000);
+
+query II
+select avg(a) > 200, avg(a) < 800 from duckdb_table_sample('t1');
+----
+true	true
+
+statement ok
+create or replace table t1 as select range a from range(204800);
+
+# average is not skewed
+query II
+select avg(a) > (0.2*204800), avg(a) < (0.8*204800) from duckdb_table_sample('t1');
+----
+true	true
+
+# about half the samples are below 102400 and half above
+query I
+select count(*) from duckdb_table_sample('t1') where a < 102400;
+----
+1069
+
+query I
+select count(*) from duckdb_table_sample('t1') where a > 102400;
+----
+979
+
+query I
+select count(*) from t1 using sample (200000);
+----
+200000
+
+statement ok
+create or replace table materialized_range as select * from range(100);
+
+statement ok
+create or replace table integers_1 as (select range b from materialized_range);
+
+query I
+select count(b) from duckdb_table_sample('integers_1') where b in (select * from materialized_range);
+----
+1
+
+# sample exists after restart
+restart
+
+query I
+select count(b) from duckdb_table_sample('integers_1') where b in (select * from materialized_range);
+----
+1
+

external/duckdb/test/sql/sample/table_samples/sample_stores_rows_from_later_on.test_slow

@@ -0,0 +1,40 @@
+# name: test/sql/sample/table_samples/sample_stores_rows_from_later_on.test_slow
+# description: Test sampling of larger relations
+# group: [table_samples]
+
+mode skip
+
+# required when testing table samples. See basic_sample_test.test
+require vector_size 2048
+
+load __TEST_DIR__/test_sample_conversion.db
+
+statement ok
+PRAGMA enable_verification
+
+statement ok
+create table t1 as select 1 a from range(200000);
+
+loop i 1 4805
+
+statement ok
+INSERT INTO t1 VALUES(${i} + 1);
+
+restart
+
+endloop
+
+query I
+select count(*) from duckdb_table_sample('t1');
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('t1') where a > 1;
+----
+48
+
+query I
+select count(*) from (select (floor(range/200000))::INT a from range(204800) using sample reservoir (1%)) t1 where a >= 1;
+----
+48

external/duckdb/test/sql/sample/table_samples/table_sample_converts_to_block_sample.test

@@ -0,0 +1,57 @@
+# name: test/sql/sample/table_samples/table_sample_converts_to_block_sample.test
+# description: Test sampling of larger relations
+# group: [table_samples]
+
+mode skip
+
+# required when testing table samples. See basic_sample_test.test
+require vector_size 2048
+
+# table samples first collect only 1% of the table, until the table has a cardinality of 2048.
+# then the sample stays at a fixed 2048 values.
+
+load __TEST_DIR__/test_sample_converts_after_load.db
+
+statement ok
+create table materialized_range as select 1 a from range(102400);
+
+# only 1% of 102400
+query I
+select count(*) from duckdb_table_sample('materialized_range');
+----
+1024
+
+restart
+
+statement ok
+insert into materialized_range select 2 a from range(102400);
+
+# collect another 1% of 102400
+query I
+select count(*) from duckdb_table_sample('materialized_range');
+----
+2048
+
+query II
+select a, count(*) from duckdb_table_sample('materialized_range') group by all order by a;
+----
+1	1024
+2	1024
+
+# insert another
+statement ok
+insert into materialized_range select 3 a from range(102400);
+
+# sample remains at 2048 values
+query I
+select count(*) from duckdb_table_sample('materialized_range');
+----
+2048
+
+# 2048 / 3 = 682. so each value should have at least >650
+query II
+select a, count(*) > 650 from duckdb_table_sample('materialized_range') group by all order by a;
+----
+1	1
+2	1
+3	1

external/duckdb/test/sql/sample/table_samples/table_sample_is_stored.test_slow

@@ -0,0 +1,148 @@
+# name: test/sql/sample/table_samples/table_sample_is_stored.test_slow
+# description: Test sampling of larger relations
+# group: [table_samples]
+
+mode skip
+
+# required when testing table samples. See basic_sample_test.test
+require vector_size 2048
+
+require icu
+
+load __TEST_DIR__/test_samples.db
+
+statement ok
+PRAGMA enable_verification
+
+statement ok
+create table materialized_range as select * from range(5000000);
+
+statement ok
+create table integers_1 as (select (range + 5) a, range b, get_current_time() as time from materialized_range);
+
+query II nosort result_1
+select a::INT, b from duckdb_table_sample('integers_1') order by all;
+----
+
+statement ok
+create table integers_2 as (select (range + 5) a, range b, get_current_time() as time from materialized_range);
+
+## samples should be the same given the same table and the same contents.
+query II nosort result_1
+select a::INT, b from duckdb_table_sample('integers_2') order by all;
+----
+
+statement ok
+create or replace table integers_1 as (select (range + 5) a, range b from materialized_range);
+
+statement ok
+create or replace table integers_2 as (select (range + 5) a, range b from materialized_range);
+
+# sample only has values in the table it was sampled from
+query I
+select count(*) from (select b from duckdb_table_sample('integers_1') intersect (select b from integers_1));
+----
+2048
+
+query I
+select count(*) from (select b from duckdb_table_sample('integers_2') intersect (select b from integers_2));
+----
+2048
+
+# sample exists after restart
+restart
+
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('integers_2');
+----
+2048
+
+
+query II
+select floor(b / 1000000) as interval, count(*) as frequency from duckdb_table_sample('integers_1') group by interval order by all;
+----
+0.0	453
+1.0	408
+2.0	406
+3.0	404
+4.0	377
+
+
+# adding another interval should subtract an equal number from the rest of the intervals
+statement ok
+insert into integers_1 (select (range + 5) a, range b from range(5000000,6000000));
+
+query II
+select floor(b / 1000000) as interval, count(*) as frequency from duckdb_table_sample('integers_1') group by interval order by all;
+----
+0.0	374
+1.0	334
+2.0	332
+3.0	334
+4.0	311
+5.0	363
+
+# If double the table count is appended, around half the sample should account for the new values.
+statement ok
+insert into integers_1 (select -1, -1 from range(6000000));
+
+query I
+select count(*) from integers_1;
+----
+12000000
+
+
+## about half of the samples should have the pair '-1', 1.
+# on latest storage test its something like 997
+query I
+select count(*) from duckdb_table_sample('integers_1') where a = -1 and b = -1;
+----
+914
+
+restart
+
+# updated sample is also newly serialized
+query I
+select count(*) from duckdb_table_sample('integers_1') where a = -1 and b = -1;
+----
+914
+
+# create a view on top of the sample
+statement ok
+create view sample_view as select * from duckdb_table_sample('integers_1');
+
+# update the sample
+statement ok
+insert into integers_1 (select -2, -2 from range(6000000));
+
+
+# 2048 / 3 = 682 (639 is good)
+query I
+select count(*) from sample_view where a = -2 and b = -2;
+----
+639
+
+restart
+
+query I
+select count(*)  from sample_view where a = -2 and b = -2;
+----
+639
+
+# currently have 18_000_000 values in the table.
+# to try and get 1 value in the sample, we should add
+# 18000000 / 2048 = 8789 values to see 1
+
+statement ok
+insert into integers_1 (select -3, -3 from range(7000));
+
+# 1 value makes it
+query I
+select count(*) from sample_view where a = -3 and b = -3;
+----
+1

external/duckdb/test/sql/sample/table_samples/test_sample_is_destroyed_on_updates.test

@@ -0,0 +1,125 @@
+# name: test/sql/sample/table_samples/test_sample_is_destroyed_on_updates.test
+# description: Test sampling of larger relations
+# group: [table_samples]
+
+mode skip
+
+# required when testing table samples. See basic_sample_test.test
+require vector_size 2048
+
+load __TEST_DIR__/test_sample_is_destroyed_on_update.db
+
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(102400);
+
+# no sample collected yet. There are only 5
+query I
+select count(*) from duckdb_table_sample('integers_1') order by all;
+----
+1024
+
+statement ok
+delete from integers_1 where a = 3;
+
+# sample no longer exists
+query I
+select count(*) from duckdb_table_sample('integers_1') order by all;
+----
+0
+
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(102400);
+
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+1024
+
+statement ok
+update integers_1 set a = 5 where a = 1;
+
+query II
+select * from duckdb_table_sample('integers_1');
+----
+
+# test adding columns destroys the sample.
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(204800);
+
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+2048
+
+statement ok
+Alter table integers_1 add column c DOUBLE;
+
+query III
+select * from duckdb_table_sample('integers_1');
+----
+
+
+# test altering types destroys the sample
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(102400);
+
+
+# don't have enough smaples yet.
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+1024
+
+statement ok
+Alter table integers_1 alter b TYPE VARCHAR
+
+query II
+select * from duckdb_table_sample('integers_1');
+----
+
+# test dropping a columns
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(102400);
+
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+1024
+
+statement ok
+Alter table integers_1 drop b;
+
+query I
+select * from duckdb_table_sample('integers_1');
+----
+
+# test sample is destroyed after a restart
+statement ok
+create or replace table integers_1 as select range a, range+1 b  from range(500);
+
+query I
+select count(*) from duckdb_table_sample('integers_1');
+----
+5
+
+statement ok
+Alter table integers_1 drop b;
+
+# sample is destroyed
+query I
+select * from duckdb_table_sample('integers_1');
+----
+
+restart
+
+statement ok
+insert into integers_1 select range a from range(500);
+
+# sample is still destroyed
+query I
+select * from duckdb_table_sample('integers_1');
+----
+
+
+
+

external/duckdb/test/sql/sample/table_samples/test_sample_types.test

@@ -0,0 +1,80 @@
+# name: test/sql/sample/table_samples/test_sample_types.test
+# description: Test sampling of larger relations
+# group: [table_samples]
+
+mode skip
+
+# test valid sampling types (for now only integral types)
+
+statement ok
+pragma enable_verification;
+
+statement ok
+create table string_samples as select range::Varchar a from range(204800);
+
+query I
+select count(*) from duckdb_table_sample('string_samples') where a is NULL;
+----
+2048
+
+statement ok
+create table struct_samples as select {'key1': 'quack-a-lack', 'key2': range} a from range(204800);
+
+query I
+select count(*) from duckdb_table_sample('struct_samples') where a is null;
+----
+2048
+
+statement ok
+create table blob_samples as select '\xAA\xAB\xAC'::BLOB a from range(204800);
+
+query I
+select count(*) from duckdb_table_sample('blob_samples') where a is NULL;
+----
+2048
+
+statement ok
+create table integral_samples as select range::BIGINT a, range::DOUBLE b, range::FLOAT c, range::HUGEINT d, INTERVAL 1 YEAR e from range(204800);
+
+query I
+select count(*) from duckdb_table_sample('integral_samples') where a NOT null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('integral_samples') where b NOT null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('integral_samples') where c NOT null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('integral_samples') where d NOT null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('integral_samples') where e IS null;
+----
+2048
+
+statement ok
+CREATE or replace TABLE t1 as select range a, [1, a, 2] b, a::VARCHAR || 'ducktastic' c, get_current_timestamp() d from range(1000000);
+
+query I
+select count(*) from duckdb_table_sample('t1') where b is null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('t1') where c is null;
+----
+2048
+
+query I
+select count(*) from duckdb_table_sample('t1') where d is null;
+----
+2048

external/duckdb/test/sql/sample/table_samples/test_table_sample_errors.test

@@ -0,0 +1,21 @@
+# name: test/sql/sample/table_samples/test_table_sample_errors.test
+# description: test table sampl[e errors
+# group: [table_samples]
+
+mode skip
+
+statement ok
+create table t1 as select range a from range(204800);
+
+statement ok
+create view v1 as select * from t1;
+
+statement error
+select * from duckdb_table_sample('v1');
+----
+<REGEX>:.*Invalid Catalog type.*
+
+statement error
+select * from duckdb_table_sample('a');
+----
+<REGEX>:.*Catalog Error:.*Table.*does not exist.*