* MSQ: Support for querying lookup and inline data directly.
Main changes:
1) Add of LookupInputSpec and DataSourcePlan.forLookup.
2) Add InlineInputSpec, and modify of DataSourcePlan.forInline to use
this instead of an ExternalInputSpec with JSON. This allows the inline
data to act as the right-hand side of a join, if needed.
Supporting changes:
1) Modify JoinDataSource's leftFilter validation to be a little less
strict: it's now OK with leftFilter being attached to any concrete
leaf (no children) datasource, rather than requiring it be a table.
This allows MSQ to create JoinDataSource with InputNumberDataSource
as the base.
2) Add SegmentWranglerModule to CliIndexer, CliPeon. This allows them to
query lookups and inline data directly.
* Updates based on CI.
* Additional tests.
* Style fix.
* Remove unused import.
* MSQ: Support multiple result columns with the same name.
This is allowed in SQL, and is supported by the regular SQL endpoint.
We retain a validation that INSERT ... SELECT does not allow multiple
columns with the same name, because column names in segments must be
unique.
changes:
* adds support for boolean inputs to the classic long dimension indexer, which plays nice with LONG being the semi official boolean type in Druid, and even nicer when druid.expressions.useStrictBooleans is set to true, since the sampler when using the new 'auto' schema when 'useSchemaDiscovery' is specified on the dimensions spec will call the type out as LONG
* fix bugs with sampler response and new schema discovery stuff incorrectly using classic 'json' type for the logical schema instead of the new 'auto' type
* Frames: Ensure nulls are read as default values when appropriate.
Fixes a bug where LongFieldWriter didn't write a properly transformed
zero when writing out a null. This had no meaningful effect in SQL-compatible
null handling mode, because the field would get treated as a null anyway.
But it does have an effect in default-value mode: it would cause Long.MIN_VALUE
to get read out instead of zero.
Also adds NullHandling checks to the various frame-based column selectors,
allowing reading of nullable frames by servers in default-value mode.
Fixes#13837.
### Description
This change allows for input source type security in the native task layer.
To enable this feature, the user must set the following property to true:
`druid.auth.enableInputSourceSecurity=true`
The default value for this property is false, which will continue the existing functionality of needing authorization to write to the respective datasource.
When this config is enabled, the users will be required to be authorized for the following resource action, in addition to write permission on the respective datasource.
`new ResourceAction(new Resource(ResourceType.EXTERNAL, {INPUT_SOURCE_TYPE}, Action.READ`
where `{INPUT_SOURCE_TYPE}` is the type of the input source being used;, http, inline, s3, etc..
Only tasks that provide a non-default implementation of the `getInputSourceResources` method can be submitted when config `druid.auth.enableInputSourceSecurity=true` is set. Otherwise, a 400 error will be thrown.
* smarter nested column index utilization
changes:
* adds skipValueRangeIndexScale and skipValuePredicateIndexScale to ColumnConfig (e.g. DruidProcessingConfig) available as system config via druid.processing.indexes.skipValueRangeIndexScale and druid.processing.indexes.skipValuePredicateIndexScale
* NestedColumnIndexSupplier uses skipValueRangeIndexScale and skipValuePredicateIndexScale to multiply by the total number of rows to be processed to determine the threshold at which we should no longer consider using bitmap indexes because it will be too many operations
* Default values for skipValueRangeIndexScale and skipValuePredicateIndexScale have been initially set to 0.08, but are separate to allow independent tuning
* these are not documented on purpose yet because they are kind of hard to explain, the mainly exist to help conduct larger scale experiments than the jmh benchmarks used to derive the initial set of values
* these changes provide a pretty sweet performance boost for filter processing on nested columns
* Always use file sizes when determining batch ingest splits.
Main changes:
1) Update CloudObjectInputSource and its subclasses (S3, GCS,
Azure, Aliyun OSS) to use SplitHintSpecs in all cases. Previously, they
were only used for prefixes, not uris or objects.
2) Update ExternalInputSpecSlicer (MSQ) to consider file size. Previously,
file size was ignored; all files were treated as equal weight when
determining splits.
A side effect of these changes is that we'll make additional network
calls to find the sizes of objects when users specify URIs or objects
as opposed to prefixes. IMO, this is worth it because it's the only way
to respect the user's split hint and task assignment settings.
Secondary changes:
1) S3, Aliyun OSS: Use getObjectMetadata instead of listObjects to get
metadata for a single object. This is a simpler call that is also
expected to be less expensive.
2) Azure: Fix a bug where getBlobLength did not populate blob
reference attributes, and therefore would not actually retrieve the
blob length.
3) MSQ: Align dynamic slicing logic between ExternalInputSpecSlicer and
TableInputSpecSlicer.
4) MSQ: Adjust WorkerInputs to ensure there is always at least one
worker, even if it has a nil slice.
* Add msqCompatible to testGroupByWithImpossibleTimeFilter.
* Fix tests.
* Add additional tests.
* Remove unused stuff.
* Remove more unused stuff.
* Adjust thresholds.
* Remove irrelevant test.
* Fix comments.
* Fix bug.
* Updates.
changes:
* introduce ColumnFormat to separate physical storage format from logical type. ColumnFormat is now used instead of ColumnCapabilities to get column handlers for segment creation
* introduce new 'auto' type indexer and merger which produces a new common nested format of columns, which is the next logical iteration of the nested column stuff. Essentially this is an automatic type column indexer that produces the most appropriate column for the given inputs, making either STRING, ARRAY<STRING>, LONG, ARRAY<LONG>, DOUBLE, ARRAY<DOUBLE>, or COMPLEX<json>.
* revert NestedDataColumnIndexer, NestedDataColumnMerger, NestedDataColumnSerializer to their version pre #13803 behavior (v4) for backwards compatibility
* fix a bug in RoaringBitmapSerdeFactory if anything actually ever wrote out an empty bitmap using toBytes and then later tried to read it (the nerve!)
* select sum(c) on an unnested column now does not return 'Type mismatch' error and works properly
* Making sure an inner join query works properly
* Having on unnested column with a group by now works correctly
* count(*) on an unnested query now works correctly
While using intermediateSuperSorterStorageMaxLocalBytes the super sorter was retaining references of the memory allocator.
The fix clears the current outputChannel when close() is called on the ComposingWritableFrameChannel.java
* Reworking s3 connector with
1. Adding retries
2. Adding max fetch size
3. Using s3Utils for most of the api's
4. Fixing bugs in DurableStorageCleaner
5. Moving to Iterator for listDir call
array columns!
changes:
* add support for storing nested arrays of string, long, and double values as specialized nested columns instead of breaking them into separate element columns
* nested column type mimic behavior means that columns ingested with only root arrays of primitive values will be ARRAY typed columns
* neat test refactor stuff
* add v4 segment test
* add array element indexes
* add tests for unnest and array columns
* fix unnest column value selector cursor handling of null and empty arrays
* Refactoring and bug fixes on top of unnest. The filter now is passed inside the unnest cursors. Added tests for scenarios such as
1. filter on unnested column which involves a left filter rewrite
2. filter on unnested virtual column which pushes the filter to the right only and involves no rewrite
3. not filters
4. SQL functions applied on top of unnested column
5. null present in first row of the column to be unnested
changes:
* fixes inconsistent handling of byte[] values between ExprEval.bestEffortOf and ExprEval.ofType, which could cause byte[] values to end up as java toString values instead of base64 encoded strings in ingest time transforms
* improved ExpressionTransform binding to re-use ExprEval.bestEffortOf when evaluating a binding instead of throwing it away
* improved ExpressionTransform array handling, added RowFunction.evalDimension that returns List<String> to back Row.getDimension and remove the automatic coercing of array types that would typically happen to expression transforms unless using Row.getDimension
* added some tests for ExpressionTransform with array inputs
* improved ExpressionPostAggregator to use partial type information from decoration
* migrate some test uses of InputBindings.forMap to use other methods
* Adds new implementation of 'frontCoded' string encoding strategy, which writes out a v1 FrontCodedIndexed which stores buckets on a prefix of the previous value instead of the first value in the bucket
* Refactoring and bug fixes on top of unnest. The filter now is passed inside the unnest cursors. Added tests for scenarios such as
1. filter on unnested column which involves a left filter rewrite
2. filter on unnested virtual column which pushes the filter to the right only and involves no rewrite
3. not filters
4. SQL functions applied on top of unnested column
5. null present in first row of the column to be unnested
* Various changes and fixes to UNNEST.
Native changes:
1) UnnestDataSource: Replace "column" and "outputName" with "virtualColumn".
This enables pushing expressions into the datasource. This in turn
allows us to do the next thing...
2) UnnestStorageAdapter: Logically apply query-level filters and virtual
columns after the unnest operation. (Physically, filters are pulled up,
when possible.) This is beneficial because it allows filters and
virtual columns to reference the unnested column, and because it is
consistent with how the join datasource works.
3) Various documentation updates, including declaring "unnest" as an
experimental feature for now.
SQL changes:
1) Rename DruidUnnestRel (& Rule) to DruidUnnestRel (& Rule). The rel
is simplified: it only handles the UNNEST part of a correlated join.
Constant UNNESTs are handled with regular inline rels.
2) Rework DruidCorrelateUnnestRule to focus on pulling Projects from
the left side up above the Correlate. New test testUnnestTwice verifies
that this works even when two UNNESTs are stacked on the same table.
3) Include ProjectCorrelateTransposeRule from Calcite to encourage
pushing mappings down below the left-hand side of the Correlate.
4) Add a new CorrelateFilterLTransposeRule and CorrelateFilterRTransposeRule
to handle pulling Filters up above the Correlate. New tests
testUnnestWithFiltersOutside and testUnnestTwiceWithFilters verify
this behavior.
5) Require a context feature flag for SQL UNNEST, since it's undocumented.
As part of this, also cleaned up how we handle feature flags in SQL.
They're now hooked into EngineFeatures, which is useful because not
all engines support all features.
With SuperSorter using the PartitionedOutputChannels for sorting, it might OOM on inputs of reasonable size because the channel consists of both the writable frame channel and the frame allocator, both of which are not required once the output channel has been written to.
This change adds a readOnly to the output channel which contains only the readable channel, due to which unnecessary memory references to the writable channel and the memory allocator are lost once the output channel has been written to, preventing the OOM.
* Window planning: use collation traits, improve subquery logic.
SQL changes:
1) Attach RelCollation (sorting) trait to any PartialDruidQuery
that ends in AGGREGATE or AGGREGATE_PROJECT. This allows planning to
take advantage of the fact that Druid sorts by dimensions when
doing aggregations.
2) Windowing: inspect RelCollation trait from input, and insert naiveSort
if, and only if, necessary.
3) Windowing: add support for Project after Window, when the Project
is a simple mapping. Helps eliminate subqueries.
4) DruidRules: update logic for considering subqueries to reflect that
subqueries are not required to be GroupBys, and that we have a bunch
of new Stages now. With all of this evolution that has happened, the
old logic didn't quite make sense.
Native changes:
1) Use merge sort (stable) rather than quicksort when sorting
RowsAndColumns. Makes it easier to write test cases for plans that
involve re-sorting the data.
* Changes from review.
* Mark the bad test as failing.
* Additional update.
* Fix failingTest.
* Fix tests.
* Mark a var final.
* Improve memory efficiency of WrappedRoaringBitmap.
Two changes:
1) Use an int[] for sizes 4 or below.
2) Remove the boolean compressRunOnSerialization. Doesn't save much
space, but it does save a little, and it isn't adding a ton of value
to have it be configurable. It was originally configurable in case
anything broke when enabling it, but it's been a while and nothing
has broken.
* Slight adjustment.
* Adjust for inspection.
* Updates.
* Update snaps.
* Update test.
* Adjust test.
* Fix snaps.
* use custom case operator conversion instead of direct operator conversion, to produce native nvl expression for SQL NVL and 2 argument COALESCE, and add optimization for certain case filters from coalesce and nvl statements
* Sort-merge join and hash shuffles for MSQ.
The main changes are in the processing, multi-stage-query, and sql modules.
processing module:
1) Rename SortColumn to KeyColumn, replace boolean descending with KeyOrder.
This makes it nicer to model hash keys, which use KeyOrder.NONE.
2) Add nullability checkers to the FieldReader interface, and an
"isPartiallyNullKey" method to FrameComparisonWidget. The join
processor uses this to detect null keys.
3) Add WritableFrameChannel.isClosed and OutputChannel.isReadableChannelReady
so callers can tell which OutputChannels are ready for reading and which
aren't.
4) Specialize FrameProcessors.makeCursor to return FrameCursor, a random-access
implementation. The join processor uses this to rewind when it needs to
replay a set of rows with a particular key.
5) Add MemoryAllocatorFactory, which is embedded inside FrameWriterFactory
instead of a particular MemoryAllocator. This allows FrameWriterFactory
to be shared in more scenarios.
multi-stage-query module:
1) ShuffleSpec: Add hash-based shuffles. New enum ShuffleKind helps callers
figure out what kind of shuffle is happening. The change from SortColumn
to KeyColumn allows ClusterBy to be used for both hash-based and sort-based
shuffling.
2) WorkerImpl: Add ability to handle hash-based shuffles. Refactor the logic
to be more readable by moving the work-order-running code to the inner
class RunWorkOrder, and the shuffle-pipeline-building code to the inner
class ShufflePipelineBuilder.
3) Add SortMergeJoinFrameProcessor and factory.
4) WorkerMemoryParameters: Adjust logic to reserve space for output frames
for hash partitioning. (We need one frame per partition.)
sql module:
1) Add sqlJoinAlgorithm context parameter; can be "broadcast" or
"sortMerge". With native, it must always be "broadcast", or it's a
validation error. MSQ supports both. Default is "broadcast" in
both engines.
2) Validate that MSQs do not use broadcast join with RIGHT or FULL join,
as results are not correct for broadcast join with those types. Allow
this in native for two reasons: legacy (the docs caution against it,
but it's always been allowed), and the fact that it actually *does*
generate correct results in native when the join is processed on the
Broker. It is much less likely that MSQ will plan in such a way that
generates correct results.
3) Remove subquery penalty in DruidJoinQueryRel when using sort-merge
join, because subqueries are always required, so there's no reason
to penalize them.
4) Move previously-disabled join reordering and manipulation rules to
FANCY_JOIN_RULES, and enable them when using sort-merge join. Helps
get to better plans where projections and filters are pushed down.
* Work around compiler problem.
* Updates from static analysis.
* Fix @param tag.
* Fix declared exception.
* Fix spelling.
* Minor adjustments.
* wip
* Merge fixups
* fixes
* Fix CalciteSelectQueryMSQTest
* Empty keys are sortable.
* Address comments from code review. Rename mux -> mix.
* Restore inspection config.
* Restore original doc.
* Reorder imports.
* Adjustments
* Fix.
* Fix imports.
* Adjustments from review.
* Update header.
* Adjust docs.
This function is notorious for causing memory exhaustion and excessive
CPU usage; so much so that it was valuable to work around it in the
SQL planner in #13206. Hopefully, a warning comment will encourage
developers to stay away and come up with solutions that do not involve
computing all possible buckets.
You can now do the following operations with TupleSketches in Post Aggregation Step
Get the Sketch Output as Base64 String
Provide a constant Tuple Sketch in post-aggregation step that can be used in Set Operations
Get the Estimated Value(Sum) of Summary/Metrics Objects associated with Tuple Sketch
The FiniteFirehoseFactory and InputRowParser classes were deprecated in 0.17.0 (#8823) in favor of InputSource & InputFormat. This PR removes the FiniteFirehoseFactory and all its implementations along with classes solely used by them like Fetcher (Used by PrefetchableTextFilesFirehoseFactory). Refactors classes including tests using FiniteFirehoseFactory to use InputSource instead.
Removing InputRowParser may not be as trivial as many classes that aren't deprecated depends on it (with no alternatives), like EventReceiverFirehoseFactory. Hence FirehoseFactory, EventReceiverFirehoseFactory, and Firehose are marked deprecated.
* move numeric null value coercion out of expression processing engine
* add ExprEval.valueOrDefault() to allow consumers to automatically coerce to default values
* rename Expr.buildVectorized as Expr.asVectorProcessor more consistent naming with Function and ApplyFunction; javadocs for some stuff
* Speed up composite key joins on IndexedTable.
Prior to this patch, IndexedTable indexes are sorted IntList. This works
great when we have a single-column join key: we simply retrieve the list
and we know what rows match. However, when we have a composite key, we
need to merge the sorted lists. This is inefficient when one is very dense
and others are very sparse.
This patch switches from sorted IntList to IntSortedSet, and changes
to the following intersection algorithm:
1) Initialize the intersection set to the smallest matching set from the
various parts of the composite key.
2) For each element in that smallest set, check other sets for that element.
If any do *not* include it, then remove the element from the intersection
set.
This way, complexity scales with the size of the smallest set, not the
largest one.
* RangeIntSet stuff.
* merge druid-core, extendedset, and druid-hll into druid-processing to simplify everything
* fix poms and license stuff
* mockito is evil
* allow reset of JvmUtils RuntimeInfo if tests used static injection to override
Clint Wylie
Gian Merlino
zachjsh
Abhishek Agarwal
Karan Kumar
Soumyava
Paul Rogers
abhagraw
imply-cheddar
Laksh Singla
Anshu Makkar
Tejaswini Bandlamudi
hqx871
Kashif Faraz
Suneet Saldanha