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

Upgrade to 149 #51

Merged
merged 790 commits into from
Nov 17, 2016
Merged

Upgrade to 149 #51

merged 790 commits into from
Nov 17, 2016

Conversation

dabaitu
Copy link

@dabaitu dabaitu commented Nov 8, 2016

Upgrade to OSS 0.149 - previous attempts to upgrade to 0.151, 0.152, 0.153, 0.154 all failed due to certain queries return no data without any error. The diff between those versions to 0.143(twitter-master) is massive so upgrading with smaller steps.
This upgrade does not involve refactoring the event listener.

haozhun and others added 30 commits May 18, 2016 14:01
We should make dates work, but this fixes the tests for now.
Syntax:
DEALLOCATE PREPARE <name>

Confirms existence of prepared statement with name <name> and adds the
statement to a set of deallocated prepared statements. Sends the set of
deallocated statements back to the client in an HTTP header. The client
is responsible for maintaining the prepared statements for a session.
Avoid calling Page.getBlocks when each invocation does not correspond to other
operations that are at least on the order of Page.positionCount.

Perf profiling showed significant CPU time spent inside JVM_Clone. Perf
backtrace points to Page.getBlocks and MultiChannelGroupByHash. Perf also
showed significant CPU time spent inside TypeArrayKlass::allocate_common, which
is likely related.
- Modify the GroupIdNode and GroupIdOperator to pass aggregation
arguments through.
- Change PruneUnreferencedOutputs optimizer to only
pass through columns that are also aggregation arguments.

The expected behavior is that arguments to the aggregations reference
values in the source relation rather than the outputs of the GroupId
operation (which can produce nulls for these columns). Prior to this
change, aggregation arguments and grouping columns referenced the same
symbol, producing incorrect results.
Summarizing TaskInfo loses the pipeline stats which are needed by
explain analyze. This commit changes TaskInfoFetcher to requst complete
TaskInfo for the last request that it sends.
When creating a prepared statement with a name that already exists,
overwrite the existing statement with the new one.  This follows the sql
spec under 20.6 <prepare statement>.
Always use dot as the decimal separator (some locales use commas).
A window frame like "ROWS BETWEEN UNBOUNDED PRECEDING AND 2 PRECEDING"
was not treated as an empty frame when appropriate.
This file only needs 100 lines to create the Hive test tables.
Previously, joins ignored coercions that had been introduced earlier
in the planning stage. This caused non-existent function signatures to
present in the plan that would fail to be resolved during execution.
Raghav Sethi and others added 28 commits June 21, 2016 13:44
New dependencies: jQuery, jQuery.sparkline, Reactable
Upgrades/old dependencies: React, Bootstrap, d3, Highlightjs
Features:
- Hip new React-based design
- Live cluster statistics
- Live query performance statistics
- Improved sorting and filtering of cluster queries
Default behavior for GroupId was to pass down parent partitioning
preferences unchanged and also pass up child actual partitioning
unchanged. However:

- Children of the GroupId plan node do not have access to new symbols
(i.e. passthrough symbols) that are generated by the GroupId. These
must be translated.
- We cannot pass through preferences for symbols that are modified by
GroupId (i.e. the set difference of the distinct and common grouping
columns). These must be removed.
The Java 7 syntax is equally concise.
* The new implementation does not support non-aligned getRange and copyRange.
* Special handling for zero-column InterleavedBlock is removed because they
  are no longer constructed.
* This fixes bugs related handling of zero-column InterleavedBlock
  and zero-position InterleavedBlock. Specifically,
  * Before this commit, calling copyRegion on a zero-column InterleavedBlock
    causes div-by-zero exception
  * Before this commit, calling getRegion and semiCompact subsequently on a
    zero-length block can produce a zero-column InterleavedBlock
@billonahill
Copy link
Collaborator

👍

@dabaitu dabaitu merged commit e5084e2 into twitter-forks:twitter-master Nov 17, 2016
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.