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.

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

Multiple optimizer passes #3880

Merged
merged 10 commits into from
Oct 19, 2022
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
6 changes: 3 additions & 3 deletions benchmarks/expected-plans/q19.txt
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@ Projection: SUM(lineitem.l_extendedprice * Int64(1) - lineitem.l_discount) AS re
Projection: lineitem.l_extendedprice, lineitem.l_discount
Filter: part.p_brand = Utf8("Brand#12") AND part.p_container IN ([Utf8("SM CASE"), Utf8("SM BOX"), Utf8("SM PACK"), Utf8("SM PKG")]) AND lineitem.l_quantity >= Decimal128(Some(100),15,2) AND lineitem.l_quantity <= Decimal128(Some(1100),15,2) AND part.p_size <= Int32(5) OR part.p_brand = Utf8("Brand#23") AND part.p_container IN ([Utf8("MED BAG"), Utf8("MED BOX"), Utf8("MED PKG"), Utf8("MED PACK")]) AND lineitem.l_quantity >= Decimal128(Some(1000),15,2) AND lineitem.l_quantity <= Decimal128(Some(2000),15,2) AND part.p_size <= Int32(10) OR part.p_brand = Utf8("Brand#34") AND part.p_container IN ([Utf8("LG CASE"), Utf8("LG BOX"), Utf8("LG PACK"), Utf8("LG PKG")]) AND lineitem.l_quantity >= Decimal128(Some(2000),15,2) AND lineitem.l_quantity <= Decimal128(Some(3000),15,2) AND part.p_size <= Int32(15)
Inner Join: lineitem.l_partkey = part.p_partkey
Filter: lineitem.l_shipmode IN ([Utf8("AIR"), Utf8("AIR REG")]) AND lineitem.l_shipinstruct = Utf8("DELIVER IN PERSON") AND lineitem.l_quantity >= Decimal128(Some(100),15,2) AND lineitem.l_quantity <= Decimal128(Some(1100),15,2) OR lineitem.l_quantity >= Decimal128(Some(1000),15,2) AND lineitem.l_quantity <= Decimal128(Some(2000),15,2) OR lineitem.l_quantity >= Decimal128(Some(2000),15,2) AND lineitem.l_quantity <= Decimal128(Some(3000),15,2)
Filter: lineitem.l_quantity >= Decimal128(Some(100),15,2) AND lineitem.l_quantity <= Decimal128(Some(1100),15,2) OR lineitem.l_quantity >= Decimal128(Some(1000),15,2) AND lineitem.l_quantity <= Decimal128(Some(2000),15,2) OR lineitem.l_quantity >= Decimal128(Some(2000),15,2) AND lineitem.l_quantity <= Decimal128(Some(3000),15,2) AND lineitem.l_shipmode IN ([Utf8("AIR"), Utf8("AIR REG")]) AND lineitem.l_shipinstruct = Utf8("DELIVER IN PERSON")
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The order changed here

TableScan: lineitem projection=[l_partkey, l_quantity, l_extendedprice, l_discount, l_shipinstruct, l_shipmode]
Filter: part.p_size >= Int32(1) AND part.p_brand = Utf8("Brand#12") AND part.p_container IN ([Utf8("SM CASE"), Utf8("SM BOX"), Utf8("SM PACK"), Utf8("SM PKG")]) AND part.p_size <= Int32(5) OR part.p_brand = Utf8("Brand#23") AND part.p_container IN ([Utf8("MED BAG"), Utf8("MED BOX"), Utf8("MED PKG"), Utf8("MED PACK")]) AND part.p_size <= Int32(10) OR part.p_brand = Utf8("Brand#34") AND part.p_container IN ([Utf8("LG CASE"), Utf8("LG BOX"), Utf8("LG PACK"), Utf8("LG PKG")]) AND part.p_size <= Int32(15)
TableScan: part projection=[p_partkey, p_brand, p_size, p_container]
Filter: part.p_brand = Utf8("Brand#12") AND part.p_container IN ([Utf8("SM CASE"), Utf8("SM BOX"), Utf8("SM PACK"), Utf8("SM PKG")]) AND part.p_size <= Int32(5) OR part.p_brand = Utf8("Brand#23") AND part.p_container IN ([Utf8("MED BAG"), Utf8("MED BOX"), Utf8("MED PKG"), Utf8("MED PACK")]) AND part.p_size <= Int32(10) OR part.p_brand = Utf8("Brand#34") AND part.p_container IN ([Utf8("LG CASE"), Utf8("LG BOX"), Utf8("LG PACK"), Utf8("LG PKG")]) AND part.p_size <= Int32(15) AND part.p_size >= Int32(1)
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The order changed here

TableScan: part projection=[p_partkey, p_brand, p_size, p_container]
9 changes: 8 additions & 1 deletion datafusion/core/src/config.rs
Original file line number Diff line number Diff line change
Expand Up @@ -64,6 +64,9 @@ pub const OPT_PARQUET_ENABLE_PAGE_INDEX: &str =
pub const OPT_OPTIMIZER_SKIP_FAILED_RULES: &str =
"datafusion.optimizer.skip_failed_rules";

/// Configuration option "datafusion.optimizer.max_passes"
pub const OPT_OPTIMIZER_MAX_PASSES: &str = "datafusion.optimizer.max_passes";

/// Definition of a configuration option
pub struct ConfigDefinition {
/// key used to identifier this configuration option
Expand Down Expand Up @@ -241,7 +244,11 @@ impl BuiltInConfigs {
rule. When set to false, any rules that produce errors will cause the query to fail.",
true
),
]
ConfigDefinition::new_u64(
OPT_OPTIMIZER_MAX_PASSES,
"Number of times that the optimizer will attempt to optimize the plan",
3
)]
}
}

Expand Down
9 changes: 8 additions & 1 deletion datafusion/core/src/execution/context.rs
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,7 @@ use crate::physical_optimizer::repartition::Repartition;

use crate::config::{
ConfigOptions, OPT_BATCH_SIZE, OPT_COALESCE_BATCHES, OPT_COALESCE_TARGET_BATCH_SIZE,
OPT_FILTER_NULL_JOIN_KEYS, OPT_OPTIMIZER_SKIP_FAILED_RULES,
OPT_FILTER_NULL_JOIN_KEYS, OPT_OPTIMIZER_MAX_PASSES, OPT_OPTIMIZER_SKIP_FAILED_RULES,
};
use crate::datasource::file_format::file_type::{FileCompressionType, FileType};
use crate::execution::{runtime_env::RuntimeEnv, FunctionRegistry};
Expand Down Expand Up @@ -1599,6 +1599,13 @@ impl SessionState {
.get_bool(OPT_OPTIMIZER_SKIP_FAILED_RULES)
.unwrap_or_default(),
)
.with_max_passes(
self.config
.config_options
.read()
.get_u64(OPT_OPTIMIZER_MAX_PASSES)
.unwrap_or_default() as u8,
)
.with_query_execution_start_time(
self.execution_props.query_execution_start_time,
);
Expand Down
1 change: 1 addition & 0 deletions datafusion/core/tests/sql/information_schema.rs
Original file line number Diff line number Diff line change
Expand Up @@ -708,6 +708,7 @@ async fn show_all() {
"| datafusion.explain.logical_plan_only | false |",
"| datafusion.explain.physical_plan_only | false |",
"| datafusion.optimizer.filter_null_join_keys | false |",
"| datafusion.optimizer.max_passes | 3 |",
"| datafusion.optimizer.skip_failed_rules | true |",
"+-------------------------------------------------+---------+",
];
Expand Down
4 changes: 2 additions & 2 deletions datafusion/core/tests/sql/subqueries.rs
Original file line number Diff line number Diff line change
Expand Up @@ -394,9 +394,9 @@ order by cntrycode;"#;
Projection: AVG(customer.c_acctbal) AS __value, alias=__sq_1
Aggregate: groupBy=[[]], aggr=[[AVG(customer.c_acctbal)]]
Filter: customer.c_acctbal > Decimal128(Some(0),15,2) AND substr(customer.c_phone, Int64(1), Int64(2)) IN ([Utf8("13"), Utf8("31"), Utf8("23"), Utf8("29"), Utf8("30"), Utf8("18"), Utf8("17")])
TableScan: customer projection=[c_phone, c_acctbal], partial_filters=[CAST(customer.c_acctbal AS Decimal128(30, 15)) > Decimal128(Some(0),30,15), substr(customer.c_phone, Int64(1), Int64(2)) IN ([Utf8("13"), Utf8("31"), Utf8("23"), Utf8("29"), Utf8("30"), Utf8("18"), Utf8("17")])]"#
TableScan: customer projection=[c_phone, c_acctbal], partial_filters=[CAST(customer.c_acctbal AS Decimal128(30, 15)) > Decimal128(Some(0),30,15), substr(customer.c_phone, Int64(1), Int64(2)) IN ([Utf8("13"), Utf8("31"), Utf8("23"), Utf8("29"), Utf8("30"), Utf8("18"), Utf8("17")]), customer.c_acctbal > Decimal128(Some(0),15,2)]"#
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Example of an optimization that we previously missed

.to_string();
assert_eq!(actual, expected);
assert_eq!(expected, actual);

// assert data
let results = execute_to_batches(&ctx, sql).await;
Expand Down
73 changes: 51 additions & 22 deletions datafusion/optimizer/src/optimizer.rs
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,7 @@ use datafusion_common::{DataFusionError, Result};
use datafusion_expr::logical_plan::LogicalPlan;
use log::{debug, trace, warn};
use std::sync::Arc;
use std::time::Instant;

/// `OptimizerRule` transforms one ['LogicalPlan'] into another which
/// computes the same results, but in a potentially more efficient
Expand Down Expand Up @@ -71,6 +72,8 @@ pub struct OptimizerConfig {
skip_failing_rules: bool,
/// Specify whether to enable the filter_null_keys rule
filter_null_keys: bool,
/// Maximum number of times to run optimizer against a plan
max_passes: u8,
}

impl OptimizerConfig {
Expand All @@ -81,6 +84,7 @@ impl OptimizerConfig {
next_id: 0, // useful for generating things like unique subquery aliases
skip_failing_rules: true,
filter_null_keys: true,
max_passes: 3,
}
}

Expand All @@ -107,6 +111,12 @@ impl OptimizerConfig {
self
}

/// Specify how many times to attempt to optimize the plan
pub fn with_max_passes(mut self, v: u8) -> Self {
self.max_passes = v;
self
}

/// Generate the next ID needed
pub fn next_id(&mut self) -> usize {
self.next_id += 1;
Expand Down Expand Up @@ -189,38 +199,57 @@ impl Optimizer {
where
F: FnMut(&LogicalPlan, &dyn OptimizerRule),
{
let start_time = Instant::now();
let mut plan_str = format!("{}", plan.display_indent());
let mut new_plan = plan.clone();
log_plan("Optimizer input", plan);
let mut i = 0;
while i < optimizer_config.max_passes {
log_plan(&format!("Optimizer input (pass {})", i), &new_plan);

for rule in &self.rules {
let result = rule.optimize(&new_plan, optimizer_config);
match result {
Ok(plan) => {
new_plan = plan;
observer(&new_plan, rule.as_ref());
log_plan(rule.name(), &new_plan);
}
Err(ref e) => {
if optimizer_config.skip_failing_rules {
// Note to future readers: if you see this warning it signals a
// bug in the DataFusion optimizer. Please consider filing a ticket
// https://github.com/apache/arrow-datafusion
warn!(
for rule in &self.rules {
let result = rule.optimize(&new_plan, optimizer_config);
match result {
Ok(plan) => {
new_plan = plan;
observer(&new_plan, rule.as_ref());
log_plan(rule.name(), &new_plan);
}
Err(ref e) => {
if optimizer_config.skip_failing_rules {
// Note to future readers: if you see this warning it signals a
// bug in the DataFusion optimizer. Please consider filing a ticket
// https://github.com/apache/arrow-datafusion
warn!(
"Skipping optimizer rule '{}' due to unexpected error: {}",
rule.name(),
e
);
} else {
return Err(DataFusionError::Internal(format!(
"Optimizer rule '{}' failed due to unexpected error: {}",
rule.name(),
e
)));
} else {
return Err(DataFusionError::Internal(format!(
"Optimizer rule '{}' failed due to unexpected error: {}",
rule.name(),
e
)));
}
}
}
}
log_plan(&format!("Optimized plan (pass {})", i), &new_plan);

// TODO this is an expensive way to see if the optimizer did anything and
// it would be better to change the OptimizerRule trait to return an Option
// instead
let new_plan_str = format!("{}", new_plan.display_indent());
if plan_str == new_plan_str {
// plan did not change, so no need to continue trying to optimize
debug!("optimizer pass {} did not make changes", i);
break;
}
Comment on lines +239 to +247
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1! This is probably a follow-up item; but instead of having the optimizer decide on this (by returning an option), it might also make sense to compute a unique plan id (bottom up) so that we can also use this to detect optimization cycles.

A very basic example is (assuming each letter is a unique plan id) A -> B -> C -> A -> B -> [max passes times more], where even though the previous plan is different from the current one we would still need to exit the loop. Having a unique id would mean we can just store a set somewhere and check against if known_plans.contains(new_plan.id) and it would break the loop.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's a great idea. Thanks @isidentical. I will write up an issue.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I filed #3892 to track this

plan_str = new_plan_str;
i += 1;
}
log_plan("Optimized plan", &new_plan);
log_plan("Final optimized plan", &new_plan);
debug!("Optimizer took {} ms", start_time.elapsed().as_millis());
Ok(new_plan)
}
}
Expand Down
22 changes: 20 additions & 2 deletions datafusion/optimizer/src/projection_push_down.rs
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@ fn optimize_plan(
_optimizer_config: &OptimizerConfig,
) -> Result<LogicalPlan> {
let mut new_required_columns = required_columns.clone();
match plan {
let new_plan = match plan {
LogicalPlan::Projection(Projection {
input,
expr,
Expand Down Expand Up @@ -509,7 +509,25 @@ fn optimize_plan(

from_plan(plan, &expr, &new_inputs)
}
}
};

// when this rule is applied multiple times it will insert duplicate nested projections,
// so we catch this here
let with_dupe_projection_removed = match new_plan? {
LogicalPlan::Projection(p) => match p.input.as_ref() {
LogicalPlan::Projection(p2) if projection_equal(&p, p2) => {
LogicalPlan::Projection(p2.clone())
}
_ => LogicalPlan::Projection(p),
},
other => other,
};
Comment on lines +516 to +524
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is the fix for #3881


Ok(with_dupe_projection_removed)
}

fn projection_equal(p: &Projection, p2: &Projection) -> bool {
p.expr.len() == p2.expr.len() && p.expr.iter().zip(&p2.expr).all(|(l, r)| l == r)
}

#[cfg(test)]
Expand Down
6 changes: 6 additions & 0 deletions docs/source/user-guide/configs.md
Original file line number Diff line number Diff line change
Expand Up @@ -40,9 +40,15 @@ Environment variables are read during `SessionConfig` initialisation so they mus
| datafusion.execution.batch_size | UInt64 | 8192 | Default batch size while creating new batches, it's especially useful for buffer-in-memory batches since creating tiny batches would results in too much metadata memory consumption. |
| datafusion.execution.coalesce_batches | Boolean | true | When set to true, record batches will be examined between each operator and small batches will be coalesced into larger batches. This is helpful when there are highly selective filters or joins that could produce tiny output batches. The target batch size is determined by the configuration setting 'datafusion.execution.coalesce_target_batch_size'. |
| datafusion.execution.coalesce_target_batch_size | UInt64 | 4096 | Target batch size when coalescing batches. Uses in conjunction with the configuration setting 'datafusion.execution.coalesce_batches'. |
| datafusion.execution.parquet.enable_page_index | Boolean | false | If true, uses parquet data page level metadata (Page Index) statistics to reduce the number of rows decoded. |
| datafusion.execution.parquet.pushdown_filters | Boolean | false | If true, filter expressions are be applied during the parquet decoding operation to reduce the number of rows decoded. |
| datafusion.execution.parquet.reorder_filters | Boolean | false | If true, filter expressions evaluated during the parquet decoding opearation will be reordered heuristically to minimize the cost of evaluation. If false, the filters are applied in the same order as written in the query. |
| datafusion.execution.time_zone | Utf8 | UTC | The session time zone which some function require e.g. EXTRACT(HOUR from SOME_TIME) shift the underline datetime according to the time zone, |
| then extract the hour. |
| datafusion.execution.time_zone | Utf8 | UTC | The session time zone which some function require e.g. EXTRACT(HOUR from SOME_TIME) shift the underline datetime according to the time zone, |
| then extract the hour |
| datafusion.explain.logical_plan_only | Boolean | false | When set to true, the explain statement will only print logical plans. |
| datafusion.explain.physical_plan_only | Boolean | false | When set to true, the explain statement will only print physical plans. |
| datafusion.optimizer.filter_null_join_keys | Boolean | false | When set to true, the optimizer will insert filters before a join between a nullable and non-nullable column to filter out nulls on the nullable side. This filter can add additional overhead when the file format does not fully support predicate push down. |
| datafusion.optimizer.max_passes | UInt64 | 3 | Number of times that the optimizer will attempt to optimize the plan |
| datafusion.optimizer.skip_failed_rules | Boolean | true | When set to true, the logical plan optimizer will produce warning messages if any optimization rules produce errors and then proceed to the next rule. When set to false, any rules that produce errors will cause the query to fail. |