* Benchmarks: New SqlBenchmark, add caching & vectorization to some others.
- Introduce a new SqlBenchmark geared towards benchmarking a wide
variety of SQL queries. Rename the old SqlBenchmark to
SqlVsNativeBenchmark.
- Add (optional) caching to SegmentGenerator to enable easier
benchmarking of larger segments.
- Add vectorization to FilteredAggregatorBenchmark and GroupByBenchmark.
* Query vectorization.
This patch includes vectorized timeseries and groupBy engines, as well
as some analogs of your favorite Druid classes:
- VectorCursor is like Cursor. (It comes from StorageAdapter.makeVectorCursor.)
- VectorColumnSelectorFactory is like ColumnSelectorFactory, and it has
methods to create analogs of the column selectors you know and love.
- VectorOffset and ReadableVectorOffset are like Offset and ReadableOffset.
- VectorAggregator is like BufferAggregator.
- VectorValueMatcher is like ValueMatcher.
There are some noticeable differences between vectorized and regular
execution:
- Unlike regular cursors, vector cursors do not understand time
granularity. They expect query engines to handle this on their own,
which a new VectorCursorGranularizer class helps with. This is to
avoid too much batch-splitting and to respect the fact that vector
selectors are somewhat more heavyweight than regular selectors.
- Unlike FilteredOffset, FilteredVectorOffset does not leverage indexes
for filters that might partially support them (like an OR of one
filter that supports indexing and another that doesn't). I'm not sure
that this behavior is desirable anyway (it is potentially too eager)
but, at any rate, it'd be better to harmonize it between the two
classes. Potentially they should both do some different thing that
is smarter than what either of them is doing right now.
- When vector cursors are created by QueryableIndexCursorSequenceBuilder,
they use a morphing binary-then-linear search to find their start and
end rows, rather than linear search.
Limitations in this patch are:
- Only timeseries and groupBy have vectorized engines.
- GroupBy doesn't handle multi-value dimensions yet.
- Vector cursors cannot handle virtual columns or descending order.
- Only some filters have vectorized matchers: "selector", "bound", "in",
"like", "regex", "search", "and", "or", and "not".
- Only some aggregators have vectorized implementations: "count",
"doubleSum", "floatSum", "longSum", "hyperUnique", and "filtered".
- Dimension specs other than "default" don't work yet (no extraction
functions or filtered dimension specs).
Currently, the testing strategy includes adding vectorization-enabled
tests to TimeseriesQueryRunnerTest, GroupByQueryRunnerTest,
GroupByTimeseriesQueryRunnerTest, CalciteQueryTest, and all of the
filtering tests that extend BaseFilterTest. In all of those classes,
there are some test cases that don't support vectorization. They are
marked by special function calls like "cannotVectorize" or "skipVectorize"
that tell the test harness to either expect an exception or to skip the
test case.
Testing should be expanded in the future -- a project in and of itself.
Related to #3011.
* WIP
* Adjustments for unused things.
* Adjust javadocs.
* DimensionDictionarySelector adjustments.
* Add "clone" to BatchIteratorAdapter.
* ValueMatcher javadocs.
* Fix benchmark.
* Fixups post-merge.
* Expect exception on testGroupByWithStringVirtualColumn for IncrementalIndex.
* BloomDimFilterSqlTest: Tag two non-vectorizable tests.
* Minor adjustments.
* Update surefire, bump up Xmx in Travis.
* Some more adjustments.
* Javadoc adjustments
* AggregatorAdapters adjustments.
* Additional comments.
* Remove switching search.
* Only missiles.
* Add round support for DS-HLL
Since the Cardinality aggregator has a "round" option to round off estimated
values generated from the HyperLogLog algorithm, add the same "round" option to
the DataSketches HLL Sketch module aggregators to be consistent.
* Fix checkstyle errors
* Change HllSketchSqlAggregator to do rounding
* Fix test for standard-compliant null handling mode
* SQL: Allow select-sort-project query shapes.
Fixes#7768.
Design changes:
- In PartialDruidQuery, allow projection after select + sort by removing
the SELECT_SORT query stage and instead allowing the SORT and
SORT_PROJECT stages to apply either after aggregation or after a plain
non-aggregating select. This is different from prior behavior, where
SORT and SORT_PROJECT were only considered valid after aggregation
stages. This logic change is in the "canAccept" method.
- In DruidQuery, represent either kind of sorting with a single "Sorting"
class (instead of DefaultLimitSpec). The Sorting class is still
convertible into a DefaultLimitSpec, but is also convertible into the
sorting parameters accepted by a Scan query.
- In DruidQuery, represent post-select and post-sorting projections with
a single "Projection" class. This obsoletes the SortProject and
SelectProjection classes, and simplifies the DruidQuery by allowing us
to move virtual-column and post-aggregator-creation logic into the
new Projection class.
- Split "DruidQuerySignature" into RowSignature and VirtualColumnRegistry.
This effectively means that instead of having mutable and immutable
versions of DruidQuerySignature, we instead of RowSignature (always
immutable) and VirtualColumnRegistry (always mutable, but sometimes
null). This change wasn't required, but IMO it this makes the logic
involving them easier to follow, and makes it more clear when the
virtual column registry is active and when it's not.
Other changes:
- ConvertBoundsToSelectors now just accepts a RowSignature, but we
use the VirtualColumnRegistry.getFullRowSignature() method to get
a signature that includes all columns, and therefore allows us to
simplify the logic (no need to special-case virtual columns).
- Add `__time` to the Scan column list if the query is ordering by time.
* Remove unused import.
* Upgrade various build and doc links to https.
Where it wasn't possible to upgrade build-time dependencies to https,
I kept http in place but used hardcoded checksums or GPG keys to ensure
that artifacts fetched over http are verified properly.
* Switch to https://apache.org.
* Bump Checkstyle to 8.20
Moderate severity vulnerability that affects:
com.puppycrawl.tools:checkstyle
Checkstyle prior to 8.18 loads external DTDs by default,
which can potentially lead to denial of service attacks
or the leaking of confidential information.
Affected versions: < 8.18
* Oops, missed one
* Oops, missed a few
* refactor sql planning to re-use expression virtual columns when possible when constructing a DruidQuery, allowing virtual columns to be defined in filter expressions, and making resulting native druid queries more concise. also minor refactor of built-in sql aggregators to maximize code re-use
* fix it
* fix it in the right place
* fixup for base64 stuff
* fixup tests
* fix merge conflict on import order
* fixup
* fix imports
* fix tests
* review comments
* refactor
* re-arrange
* better javadoc
* fixup merge
* fixup tests
* fix accidental changes
* Added checkstyle for "Methods starting with Capital Letters" and changed the method names violating this.
* Un-abbreviate the method names in the calcite tests
* Fixed checkstyle errors
* Changed asserts position in the code
* * Add few methods about base64 into StringUtils
* Use `java.util.Base64` instead of others
* Add org.apache.commons.codec.binary.Base64 & com.google.common.io.BaseEncoding into druid-forbidden-apis
* Rename encodeBase64String & decodeBase64String
* Update druid-forbidden-apis
Not putting this to 0.13 milestone because the found bugs are not critical (one is a harmless DI config duplicate, and another is in a benchmark.
Change in `DumpSegment` is just an indentation change.
* Add checkstyle rules about imports and empty lines between members
* Add suppressions
* Update Eclipse import order
* Add empty line
* Fix StatsDEmitter
* Prohibit some guava collection APIs and use JDK APIs directly
* reset files that changed by accident
* sort codestyle/druid-forbidden-apis.txt alphabetically
This PR accumulates many refactorings and small improvements that I did while preparing the next change set of https://github.com/druid-io/druid/projects/2. I finally decided to make them a separate PR to minimize the volume of the main PR.
Some of the changes:
- Renamed confusing "Generic Column" term to "Numeric Column" (what it actually implies) in many class names.
- Generified `ComplexMetricExtractor`
* Rename io.druid to org.apache.druid.
* Fix META-INF files and remove some benchmark results.
* MonitorsConfig update for metrics package migration.
* Reorder some dimensions in inner queries for some reason.
* Fix protobuf tests.
Gian Merlino
Clint Wylie
Alexander Saydakov
Chi Cao Minh
Xue Yu
Jihoon Son
Gian Merlino
Merlin Lee
Fokko Driesprong
es1220
Jonathan Wei
Alexander Saydakov
Clint Wylie
Himanshu Pandey
Roman Leventov
Benedict Jin
Alexander Saydakov
David Lim
QiuMM
David Lim
Benedict Jin