Tanya-W commented on code in PR #13035:
URL: https://github.com/apache/doris/pull/13035#discussion_r1057245023
##########
be/src/vec/exec/scan/vscan_node.cpp:
##########
@@ -868,6 +880,193 @@ Status
VScanNode::_normalize_noneq_binary_predicate(VExpr* expr, VExprContext* e
return Status::OK();
}
+Status VScanNode::_normalize_compound_predicate(
+ vectorized::VExpr* expr, VExprContext* expr_ctx, PushDownType* pdt,
+ std::vector<ColumnValueRangeType>* column_value_rangs,
+ const std::function<bool(const std::vector<VExpr*>&, const VSlotRef**,
VExpr**)>&
+ in_predicate_checker,
+ const std::function<bool(const std::vector<VExpr*>&, const VSlotRef**,
VExpr**)>&
+ eq_predicate_checker) {
+ if (TExprNodeType::COMPOUND_PRED == expr->node_type()) {
+ DCHECK(expr->children().size() == 2);
+ auto compound_fn_name = expr->fn().name.function_name;
+ auto children_num = expr->children().size();
+ for (auto i = 0; i < children_num; ++i) {
+ VExpr* child_expr = expr->children()[i];
+ if (TExprNodeType::BINARY_PRED == child_expr->node_type()) {
+ SlotDescriptor* slot = nullptr;
+ ColumnValueRangeType* range_on_slot = nullptr;
+ if (_is_predicate_acting_on_slot(child_expr,
in_predicate_checker, &slot,
+ &range_on_slot) ||
+ _is_predicate_acting_on_slot(child_expr,
eq_predicate_checker, &slot,
+ &range_on_slot)) {
+ ColumnValueRangeType active_range =
+ *range_on_slot; // copy, in order not to affect
the range in the _colname_to_value_range
+ std::visit(
+ [&](auto& value_range) {
+ _normalize_binary_in_compound_predicate(
+ child_expr, expr_ctx, slot,
value_range, pdt,
+
_get_compound_type_by_fn_name(compound_fn_name));
+ },
+ active_range);
+
+ column_value_rangs->emplace_back(active_range);
+ }
+ } else if (TExprNodeType::COMPOUND_PRED ==
child_expr->node_type()) {
+ _normalize_compound_predicate(child_expr, expr_ctx, pdt,
column_value_rangs,
+ in_predicate_checker,
eq_predicate_checker);
+ }
+ }
+ }
+
+ return Status::OK();
+}
+
+template <PrimitiveType T>
+Status VScanNode::_normalize_binary_in_compound_predicate(
+ vectorized::VExpr* expr, VExprContext* expr_ctx, SlotDescriptor* slot,
+ ColumnValueRange<T>& range, PushDownType* pdt, const
TCompoundType::type& compound_type) {
+ DCHECK(expr->children().size() == 2);
+ if (TExprNodeType::BINARY_PRED == expr->node_type()) {
+ auto eq_checker = [](const std::string& fn_name) { return fn_name ==
"eq"; };
+ auto ne_checker = [](const std::string& fn_name) { return fn_name ==
"ne"; };
+ auto noneq_checker = [](const std::string& fn_name) {
+ return fn_name != "ne" && fn_name != "eq";
+ };
+
+ StringRef value;
+ int slot_ref_child = -1;
+ PushDownType eq_pdt;
+ PushDownType ne_pdt;
+ PushDownType noneq_pdt;
+ RETURN_IF_ERROR(_should_push_down_binary_predicate(
+ reinterpret_cast<VectorizedFnCall*>(expr), expr_ctx, &value,
&slot_ref_child,
+ eq_checker, eq_pdt));
+ RETURN_IF_ERROR(_should_push_down_binary_predicate(
+ reinterpret_cast<VectorizedFnCall*>(expr), expr_ctx, &value,
&slot_ref_child,
+ ne_checker, ne_pdt));
+ RETURN_IF_ERROR(_should_push_down_binary_predicate(
+ reinterpret_cast<VectorizedFnCall*>(expr), expr_ctx, &value,
&slot_ref_child,
+ noneq_checker, noneq_pdt));
+ if (eq_pdt == PushDownType::UNACCEPTABLE && ne_pdt ==
PushDownType::UNACCEPTABLE &&
+ noneq_pdt == PushDownType::UNACCEPTABLE) {
+ return Status::OK();
+ }
+ DCHECK(slot_ref_child >= 0);
+
+ if (eq_pdt == PushDownType::ACCEPTABLE) {
+ auto temp_range =
ColumnValueRange<T>::create_empty_column_value_range(
+ slot->type().precision, slot->type().scale);
+ auto fn_name = std::string("");
+ if (value.data != nullptr) {
+ if constexpr (T == TYPE_CHAR || T == TYPE_VARCHAR || T ==
TYPE_STRING ||
+ T == TYPE_HLL) {
+ auto val = StringValue(value.data, value.size);
+ RETURN_IF_ERROR(_change_value_range<true>(
+ temp_range, reinterpret_cast<void*>(&val),
+ ColumnValueRange<T>::add_fixed_value_range,
fn_name));
+ } else {
+ RETURN_IF_ERROR(_change_value_range<true>(
+ temp_range,
reinterpret_cast<void*>(const_cast<char*>(value.data)),
+ ColumnValueRange<T>::add_fixed_value_range,
fn_name));
+ }
+ range.intersection(temp_range);
+ }
+ *pdt = eq_pdt;
+
+ // exceed limit, no conditions will be pushed down to storage
engine.
+ if (range.get_fixed_value_size() >
_max_pushdown_conditions_per_column) {
+ range.set_whole_value_range();
+ *pdt = PushDownType::UNACCEPTABLE;
+ }
+ range.set_compound_type(compound_type);
+ }
+
+ if (ne_pdt == PushDownType::ACCEPTABLE) {
+ bool is_fixed_range = range.is_fixed_value_range();
+ auto not_in_range =
+
ColumnValueRange<T>::create_empty_column_value_range(range.column_name());
+ auto fn_name = std::string("");
+ if (value.data != nullptr) {
+ auto fn_name = std::string("");
+ if constexpr (T == TYPE_CHAR || T == TYPE_VARCHAR || T ==
TYPE_STRING ||
+ T == TYPE_HLL) {
+ auto val = StringValue(value.data, value.size);
+ if (is_fixed_range) {
+ RETURN_IF_ERROR(_change_value_range<true>(
+ range, reinterpret_cast<void*>(&val),
+ ColumnValueRange<T>::remove_fixed_value_range,
fn_name));
+ } else {
+ RETURN_IF_ERROR(_change_value_range<true>(
+ not_in_range, reinterpret_cast<void*>(&val),
+ ColumnValueRange<T>::add_fixed_value_range,
fn_name));
+ }
+ } else {
+ if (is_fixed_range) {
+ RETURN_IF_ERROR(_change_value_range<true>(
+ range,
reinterpret_cast<void*>(const_cast<char*>(value.data)),
+ ColumnValueRange<T>::remove_fixed_value_range,
fn_name));
+ } else {
+ RETURN_IF_ERROR(_change_value_range<true>(
+ not_in_range,
+
reinterpret_cast<void*>(const_cast<char*>(value.data)),
+ ColumnValueRange<T>::add_fixed_value_range,
fn_name));
+ }
+ }
+ }
+
+ if (is_fixed_range ||
+ not_in_range.get_fixed_value_size() <=
_max_pushdown_conditions_per_column) {
+ if (!is_fixed_range) {
+ _not_in_value_ranges.push_back(not_in_range);
+ }
+ *pdt = ne_pdt;
+ }
+ range.set_compound_type(compound_type);
+ }
+
+ if (noneq_pdt == PushDownType::ACCEPTABLE) {
+ const std::string& fn_name =
+
reinterpret_cast<VectorizedFnCall*>(expr)->fn().name.function_name;
+
+ // where A = nullptr should return empty result set
+ if (value.data != nullptr) {
+ if constexpr (T == TYPE_CHAR || T == TYPE_VARCHAR || T ==
TYPE_STRING ||
+ T == TYPE_HLL) {
+ auto val = StringValue(value.data, value.size);
+ RETURN_IF_ERROR(_change_value_range<false>(
+ range, reinterpret_cast<void*>(&val),
+ ColumnValueRange<T>::add_compound_value_range,
fn_name,
Review Comment:
The original intention is to reuse the fixed value code, but this may not be
necessary. In the latest code, the fixed value and range value in compound
conditions except leaf node of and node, have been uniformly stored in
`_compound_values`
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
