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

347 #9

Merged
merged 2,104 commits into from
Nov 28, 2020
Merged

347 #9

merged 2,104 commits into from
Nov 28, 2020

Conversation

tooptoop4
Copy link
Owner

No description provided.

wendigo and others added 30 commits November 10, 2020 13:25
Make both 128x64 and 128x128 multiplications use one generic method
Remove conditional statements from 64x64 -> 128 bit multiplication
Also rename KerberosKms to HadoopKerberosKms.
Attempting to invoke a test run with no time limit remaining would have
no chance to succeed, and more importantly would lead to a immediate
failure like

```
io.prestosql.tests.product.launcher.cli.TestRun	Failure: java.lang.IllegalArgumentException: timeout must be > 0
	at net.jodah.failsafe.internal.util.Assert.isTrue(Assert.java:27)
	at net.jodah.failsafe.Timeout.of(Timeout.java:109)
	at io.prestosql.tests.product.launcher.cli.TestRun$Execution.call(TestRun.java:176)
	at io.prestosql.tests.product.launcher.cli.SuiteRun$Execution.runTest(SuiteRun.java:229)
	at io.prestosql.tests.product.launcher.cli.SuiteRun$Execution.executeSuiteTestRun(SuiteRun.java:221)
	at io.prestosql.tests.product.launcher.cli.SuiteRun$Execution.call(SuiteRun.java:168)
	at io.prestosql.tests.product.launcher.cli.SuiteRun$Execution.call(SuiteRun.java:118)
```
Signed-off-by: Ke Zhu <kzhu@us.ibm.com>
electrum and others added 29 commits November 25, 2020 18:25
We should not pass column metadata to filterColumns like we do not pass
TableMetadata to filterTables method.

Requiring ColumnMetadata makes it cumbersome to call filterColumns
method as it requires to have column metadata in hand which not always
is trivial.
Ensure the operator infos can be merged, rather than ignoring the
inconsistencies and returning a "random" one.
TableHandle should be cached if there is a value,
or tableHandle is empty but missing value cashing is expected.
Loader in cache.get(key, loader) is able to refer
to state other than that in the key.
This is a preparation step for enhancing some loaders
with necessary runtime information (e.g. query id)
which cannot be stored within cache key.
Some assertion were using explicit matching while others were using
`anyTree` for no apparent reason. Make the assertions consistent across
the class.
Factor out the repetitive stanzas so that tested aspects are easier to
see. Since upon failure the actual failing query is reported, it does
not limit ability to hand-test selected cases.
Update operator id independently of whether
DynamicFilterSourceOperatorFactory is created
(for broadcast joins it might be created in single task only)
This way operator summaries can be correctly merged at pipeline level.
Additional information from ConnectorSession is required
to support audit in not public connectors
to correlate query in Presto with query w RDBMS
Gather auxiliary statistics from all operators
created for a given plan node. Previously, auxiliary
stats were only collected for input and output operators
for a given plan node. However, because creation of DynamicFilterSourceOperator
(input operator for join plan node) is conditional for some task
HashCollisionPlanNodeStats might be lost.
Collision stats were not correctly reported previously
for join operators. Now they are reported like:

 Fragment 1 [HASH]
    CPU: 154.12ms, Scheduled: 339.28ms, Input: 100 rows (9.33kB); per task: avg.: 33.33 std.dev.: 0.47, Output: 125 rows (13.67kB)
    Output layout: [nationkey, regionkey, nationkey_0, name_1, comment_3]
    Output partitioning: SINGLE []
    Stage Execution Strategy: UNGROUPED_EXECUTION
    InnerJoin[("regionkey" = "regionkey_2")][$hashvalue_7, $hashvalue_8]
    │   Layout: [nationkey:bigint, regionkey:bigint, nationkey_0:bigint, name_1:varchar(25), comment_3:varchar(152)]
    │   Estimates: {rows: ? (?), cpu: ?, memory: ?, network: ?}
    │   CPU: 90.00ms (40.36%), Scheduled: 194.00ms (35.79%), Output: 125 rows (13.67kB)
    │   Left (probe) Input avg.: 2.08 rows, Input std.dev.: 86.44%
    │   Right (build) Input avg.: 6.25 rows, Input std.dev.: 87.18%
    │   Collisions avg.: 5.64 (100.00% est.), Collisions std.dev.: 0.00%
    │   Distribution: REPLICATED
    │   dynamicFilterAssignments = {regionkey_2 -> #df_385}

Aggregation collision stats are reported like:

    │  └─ Aggregate(FINAL)[nationkey, regionkey][$hashvalue]
    │     │   Layout: [nationkey:bigint, regionkey:bigint, $hashvalue:bigint]
    │     │   Estimates: {rows: ? (?), cpu: ?, memory: ?, network: ?}
    │     │   CPU: 26.00ms (11.66%), Scheduled: 57.00ms (10.52%), Output: 25 rows (675B)
    │     │   Input avg.: 2.08 rows, Input std.dev.: 86.44%
    │     │   Collisions avg.: 0.00 (100.00% est.), Collisions std.dev.: 0.00%
@tooptoop4 tooptoop4 merged commit 41d2fd9 into tooptoop4:master Nov 28, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.