Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Cherry pick #44067 to 22.3: Fix undefined behavior in quantiles function #44140

Merged
merged 8 commits into from
Dec 13, 2022
Merged

Cherry pick #44067 to 22.3: Fix undefined behavior in quantiles function #44140

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
6f69200
Fix undefined behavior in `quantiles` function
alexey-milovidov Dec 9, 2022
46f7c94
Use UTC time zone in test
davenger Dec 9, 2022
a46699a
Add more tests
alexey-milovidov Dec 11, 2022
552eed9
Merge branch 'better-signal-messages' into fix-quantiles-msan
alexey-milovidov Dec 11, 2022
9cb0955
Update tests
alexey-milovidov Dec 11, 2022
0acadc6
Fix one more UB
alexey-milovidov Dec 11, 2022
bca1829
Merge pull request #44067 from ClickHouse/fix-quantiles-msan
alexey-milovidov Dec 12, 2022
1aca065
Merge branch 'backport/22.3/44067' into cherrypick/22.3/bca18298a14bb…
alexey-milovidov Dec 13, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions src/AggregateFunctions/AggregateFunctionQuantile.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -167,15 +167,15 @@ class AggregateFunctionQuantile final : public IAggregateFunctionDataHelper<Data
{
auto & data_to = assert_cast<ColumnVector<FloatReturnType> &>(arr_to.getData()).getData();
size_t old_size = data_to.size();
data_to.resize(data_to.size() + size);
data_to.resize(old_size + size);

data.getManyFloat(levels.levels.data(), levels.permutation.data(), size, data_to.data() + old_size);
}
else
{
auto & data_to = static_cast<ColVecType &>(arr_to.getData()).getData();
size_t old_size = data_to.size();
data_to.resize(data_to.size() + size);
data_to.resize(old_size + size);

data.getMany(levels.levels.data(), levels.permutation.data(), size, data_to.data() + old_size);
}
Expand Down
24 changes: 22 additions & 2 deletions src/AggregateFunctions/QuantileReservoirSampler.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,15 +55,35 @@ struct QuantileReservoirSampler
/// Get the value of the `level` quantile. The level must be between 0 and 1.
Value get(Float64 level)
{
return Value(data.quantileInterpolated(level));
if (data.empty())
return {};

if constexpr (is_decimal<Value>)
return Value(static_cast<typename Value::NativeType>(data.quantileInterpolated(level)));
else
return static_cast<Value>(data.quantileInterpolated(level));
}

/// Get the `size` values of `levels` quantiles. Write `size` results starting with `result` address.
/// indices - an array of index levels such that the corresponding elements will go in ascending order.
void getMany(const Float64 * levels, const size_t * indices, size_t size, Value * result)
{
bool is_empty = data.empty();

for (size_t i = 0; i < size; ++i)
result[indices[i]] = Value(data.quantileInterpolated(levels[indices[i]]));
{
if (is_empty)
{
result[i] = Value{};
}
else
{
if constexpr (is_decimal<Value>)
result[indices[i]] = Value(static_cast<typename Value::NativeType>(data.quantileInterpolated(levels[indices[i]])));
else
result[indices[i]] = Value(data.quantileInterpolated(levels[indices[i]]));
}
}
}

/// The same, but in the case of an empty state, NaN is returned.
Expand Down
24 changes: 22 additions & 2 deletions src/AggregateFunctions/QuantileReservoirSamplerDeterministic.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,15 +55,35 @@ struct QuantileReservoirSamplerDeterministic
/// Get the value of the `level` quantile. The level must be between 0 and 1.
Value get(Float64 level)
{
return data.quantileInterpolated(level);
if (data.empty())
return {};

if constexpr (is_decimal<Value>)
return static_cast<typename Value::NativeType>(data.quantileInterpolated(level));
else
return static_cast<Value>(data.quantileInterpolated(level));
}

/// Get the `size` values of `levels` quantiles. Write `size` results starting with `result` address.
/// indices - an array of index levels such that the corresponding elements will go in ascending order.
void getMany(const Float64 * levels, const size_t * indices, size_t size, Value * result)
{
bool is_empty = data.empty();

for (size_t i = 0; i < size; ++i)
result[indices[i]] = data.quantileInterpolated(levels[indices[i]]);
{
if (is_empty)
{
result[i] = Value{};
}
else
{
if constexpr (is_decimal<Value>)
result[indices[i]] = static_cast<typename Value::NativeType>(data.quantileInterpolated(levels[indices[i]]));
else
result[indices[i]] = static_cast<Value>(data.quantileInterpolated(levels[indices[i]]));
}
}
}

/// The same, but in the case of an empty state, NaN is returned.
Expand Down
5 changes: 5 additions & 0 deletions src/AggregateFunctions/ReservoirSampler.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,11 @@ class ReservoirSampler
return total_values;
}

bool empty() const
{
return samples.empty();
}

T quantileNearest(double level)
{
if (samples.empty())
Expand Down
5 changes: 5 additions & 0 deletions src/AggregateFunctions/ReservoirSamplerDeterministic.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,11 @@ class ReservoirSamplerDeterministic
return total_values;
}

bool empty() const
{
return samples.empty();
}

T quantileNearest(double level)
{
if (samples.empty())
Expand Down
25 changes: 25 additions & 0 deletions tests/queries/0_stateless/02499_quantile_nan_ubsan_msan.reference
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,25 @@
['1970-01-01 00:00:00']
['1970-01-01 00:00:00']

['1970-01-01 00:00:00']
['1970-01-01 00:00:00']
['1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00']

['1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00']
['1970-01-01 00:00:00']
['1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00']
[18446744073709552000]
['1970-01-01 00:00:00']
['1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00','1970-01-01 00:00:00']
[1.157920892373162e77]
[nan]
1970-01-01 00:00:00
1970-01-01 00:00:00

1970-01-01 00:00:00
1970-01-01 00:00:00
1970-01-01 00:00:00
18446744073709552000
1970-01-01 00:00:00
1.157920892373162e77
nan
22 changes: 22 additions & 0 deletions tests/queries/0_stateless/02499_quantile_nan_ubsan_msan.sql
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
SELECT quantiles(0.5)(now()::DateTime('UTC')) WHERE 0;
SELECT quantiles(0.5)(now()::DateTime('UTC')) WHERE 0 WITH TOTALS;
SELECT arrayReduce('quantiles(0.5)', []::Array(DateTime('UTC')));
SELECT quantiles(0.5, 1.1754943508222875e-38, 0.0001, -0., 0.0001, -0., 0.0001, 0., 0.5)(now()::DateTime('UTC')) WHERE 0 WITH TOTALS;

SELECT DISTINCT arrayReduce('quantiles(0.5)', materialize([]::Array(DateTime('UTC')))) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantiles(0, 0.5, 0.9, 1)', materialize([]::Array(DateTime('UTC')))) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantiles(0.5)', [0xFFFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFE]) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantilesDeterministic(0.5)', materialize([]::Array(DateTime('UTC'))), []::Array(UInt64)) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantilesDeterministic(0, 0.5, 0.9, 1)', materialize([]::Array(DateTime('UTC'))), []::Array(UInt64)) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantiles(0.5)', [CAST(-1, 'UInt256'), CAST(-2, 'UInt256')]) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantiles(0.5)', []::Array(Float64)) FROM numbers(1000) LIMIT 10;

SELECT quantile(0.5)(now()::DateTime('UTC')) WHERE 0;
SELECT quantile(0.5)(now()::DateTime('UTC')) WHERE 0 WITH TOTALS;
SELECT arrayReduce('quantile(0.5)', []::Array(DateTime('UTC')));

SELECT DISTINCT arrayReduce('quantile(0.5)', materialize([]::Array(DateTime('UTC')))) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantile(0.5)', [0xFFFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFE]) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantileDeterministic(0.5)', materialize([]::Array(DateTime('UTC'))), []::Array(UInt64)) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantile(0.5)', [CAST(-1, 'UInt256'), CAST(-2, 'UInt256')]) FROM numbers(1000) LIMIT 10;
SELECT DISTINCT arrayReduce('quantile(0.5)', []::Array(Float64)) FROM numbers(1000) LIMIT 10;