Skip to content

[native] Handle scan filter with NaN limits during plan conversion #23289

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

Merged
amitkdutta merged 1 commit into from
Jul 24, 2024
Merged

[native] Handle scan filter with NaN limits during plan conversion #23289

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
84 changes: 43 additions & 41 deletions presto-native-execution/presto_cpp/main/types/PrestoToVeloxConnector.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -214,20 +214,13 @@ int64_t dateToInt64(
return value.value<int32_t>();
}

double toDouble(
const std::shared_ptr<protocol::Block>& block,
const VeloxExprConverter& exprConverter,
const TypePtr& type) {
auto variant = exprConverter.getConstantValue(type, *block);
return variant.value<double>();
}

float toFloat(
template <typename T>
T toFloatingPoint(
const std::shared_ptr<protocol::Block>& block,
const VeloxExprConverter& exprConverter,
const TypePtr& type) {
auto variant = exprConverter.getConstantValue(type, *block);
return variant.value<float>();
return variant.value<T>();
}

std::string toString(
Expand Down Expand Up @@ -393,47 +386,54 @@ std::unique_ptr<common::Filter> boolRangeToFilter(
VELOX_UNREACHABLE();
}

std::unique_ptr<common::DoubleRange> doubleRangeToFilter(
template <typename T>
std::unique_ptr<common::Filter> floatingPointRangeToFilter(
const protocol::Range& range,
bool nullAllowed,
const VeloxExprConverter& exprConverter,
const TypePtr& type) {
bool lowExclusive = range.low.bound == protocol::Bound::ABOVE;
bool lowUnbounded = range.low.valueBlock == nullptr && lowExclusive;
auto low = lowUnbounded ? std::numeric_limits<double>::lowest()
: toDouble(range.low.valueBlock, exprConverter, type);
auto low = lowUnbounded
? (-1.0 * std::numeric_limits<T>::infinity())
: toFloatingPoint<T>(range.low.valueBlock, exprConverter, type);

bool highExclusive = range.high.bound == protocol::Bound::BELOW;
bool highUnbounded = range.high.valueBlock == nullptr && highExclusive;
auto high = highUnbounded
? std::numeric_limits<double>::max()
: toDouble(range.high.valueBlock, exprConverter, type);
return std::make_unique<common::DoubleRange>(
low,
lowUnbounded,
lowExclusive,
high,
highUnbounded,
highExclusive,
nullAllowed);
}
? std::numeric_limits<T>::infinity()
: toFloatingPoint<T>(range.high.valueBlock, exprConverter, type);

std::unique_ptr<common::FloatRange> floatRangeToFilter(
const protocol::Range& range,
bool nullAllowed,
const VeloxExprConverter& exprConverter,
const TypePtr& type) {
bool lowExclusive = range.low.bound == protocol::Bound::ABOVE;
bool lowUnbounded = range.low.valueBlock == nullptr && lowExclusive;
auto low = lowUnbounded ? std::numeric_limits<float>::lowest()
: toFloat(range.low.valueBlock, exprConverter, type);
// Handle NaN cases as NaN is not supported as a limit in Velox Filters
if (!lowUnbounded && std::isnan(low)) {
if (lowExclusive) {
// x > NaN is always false as NaN is considered the largest value.
return std::make_unique<common::AlwaysFalse>();
}
// Equivalent to x > infinity as only NaN is greater than infinity
// Presto currently converts x >= NaN into the filter with domain
// [NaN, max), so ignoring the high value is fine.
low = std::numeric_limits<T>::infinity();
lowExclusive = true;
high = std::numeric_limits<T>::infinity();
highUnbounded = true;
highExclusive = false;
} else if (!highUnbounded && std::isnan(high)) {
high = std::numeric_limits<T>::infinity();
if (highExclusive) {
// equivalent to x in [low , infinity] or (low , infinity]
highExclusive = false;
} else {
if (lowUnbounded) {
// Anything <= NaN is true as NaN is the largest possible value.
return std::make_unique<common::AlwaysTrue>();
}
// Equivalent to x > low or x >=low
highUnbounded = true;
}
}

bool highExclusive = range.high.bound == protocol::Bound::BELOW;
bool highUnbounded = range.high.valueBlock == nullptr && highExclusive;
auto high = highUnbounded
? std::numeric_limits<float>::max()
: toFloat(range.high.valueBlock, exprConverter, type);
return std::make_unique<common::FloatRange>(
return std::make_unique<common::FloatingPointRange<T>>(
low,
lowUnbounded,
lowExclusive,
Expand Down Expand Up @@ -653,14 +653,16 @@ std::unique_ptr<common::Filter> toFilter(
case TypeKind::HUGEINT:
return hugeintRangeToFilter(range, nullAllowed, exprConverter, type);
case TypeKind::DOUBLE:
return doubleRangeToFilter(range, nullAllowed, exprConverter, type);
return floatingPointRangeToFilter<double>(
range, nullAllowed, exprConverter, type);
case TypeKind::VARCHAR:
case TypeKind::VARBINARY:
return varcharRangeToFilter(range, nullAllowed, exprConverter, type);
case TypeKind::BOOLEAN:
return boolRangeToFilter(range, nullAllowed, exprConverter, type);
case TypeKind::REAL:
return floatRangeToFilter(range, nullAllowed, exprConverter, type);
return floatingPointRangeToFilter<float>(
range, nullAllowed, exprConverter, type);
case TypeKind::TIMESTAMP:
return timestampRangeToFilter(range, nullAllowed, exprConverter, type);
default:
Expand Down