functions
This commit fixes an issue where a [NOT] EXISTS subquery that contains
an aggregate function will sometimes be incorrectly rewritten into a
join, thereby returning incorrect results.
Change-Id: I18b211d76ee3de77d8061603ff5bb1fbceae2e60
Reviewed-on: http://gerrit.cloudera.org:8080/266
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
The templates for starting the services of the cluster had a bad
declaration of the shebang that made it impossible to start kms when
using a non-bash default shell.
Change-Id: I6b105b328dc61e71095c2d5e5d6859f65ca56a18
Reviewed-on: http://gerrit.cloudera.org:8080/293
Reviewed-by: Martin Grund <mgrund@cloudera.com>
Tested-by: Internal Jenkins
This commit fixes the issue where no lineage events are generated for
create and alter view statements.
Change-Id: Ib8c4513219569f62eb26a0eb09a8c2a762054b70
Reviewed-on: http://gerrit.cloudera.org:8080/265
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
While cherry-picking commit c289974 from cdh5-2.2.0_5.4.x, a line that causes a dataset
change was somehow missed by git's cherry-pick. This patch adds that line back in.
Change-Id: Id163643c55d44fe6a9503b54397f1f370b198554
I did a local benchmark and there's minimal performance impact(<1%)
Change-Id: I8d84a145acad886c52587258b27d33cff96ea399
(cherry picked from commit 7e750ad5d90007cc85ebe493af4dce7a537ad7c0)
Reviewed-on: http://gerrit.cloudera.org:8080/189
Reviewed-by: Juan Yu <jyu@cloudera.com>
Tested-by: Internal Jenkins
We used mkdir without a -p to load a hive13 parquet table. This failed when trying to load
the metadata, which is necessary for changes that mutate the warehouse snapshot. With this
change, it now passes.
Change-Id: Id163643c55d44fe6a9503b54397f1f370b198554
Fixes a bug where FIRST_VALUE may produce incorrect results (or a DCHECK
failure in debug) when there is a window like "ROWS X PRECEDING Y PRECEDING",
such that X < Y and X > the size of a partition.
For windows with an end boundary that is PRECEDING (i.e.
the entire window is before a row), there is some special handling between
partitions, and the logic was not correct in some corner cases for FIRST_VALUE.
Change-Id: Ied5d440684e99dcaf60b47489c90300891f09b91
Reviewed-on: http://gerrit.cloudera.org:8080/236
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
The IN predicate wasn't using the decimal type when comparing decimal
values. I benchmarked this on a modified version of TPCDS-Q8 (i.e. a
query with a huge decimal IN predicate) and there is a ~5% performance
degradation with codegen enabled (surprisingly, there appears to be a
slight performance gain with codegen disabled). We should be able to
remove this penalty when we add constant injection via codegen.
Change-Id: Ie1296fd50c68d06a343701442da49fe8d3cd16dd
Reviewed-on: http://gerrit.cloudera.org:8080/230
Reviewed-by: Skye Wanderman-Milne <skye@cloudera.com>
Tested-by: Internal Jenkins
This patch removes the logic from the python test file, it should really live in the
code that sets up the test-warehouse.
Change-Id: Id04dc90c7ab813af2f347ec79e9e43d76de794a2
Reviewed-on: http://gerrit.cloudera.org:8080/224
Reviewed-by: Ippokratis Pandis <ipandis@cloudera.com>
Tested-by: Internal Jenkins
Implicit casting to decimals allows truncating digits from the left of the
decimal point (see TypesUtil). A literal that is implicitly cast to a decimal
with truncation is wrapped into a CastExpr so the BE can evaluate it and report
a warning. This behavior is consistent with casting/overflow of non-constant
exprs that return decimal.
IMPALA-1837: Without the CastExpr wrapping, such literals can exceed the max
expected byte size sent to the BE in toThrift().
Change-Id: Icd7b8751b39b8031832eec04bd8eac7d7000ddf8
Reviewed-on: http://gerrit.cloudera.org:8080/195
Reviewed-by: Marcel Kornacker <marcel@cloudera.com>
Tested-by: Internal Jenkins
Re-enables data error tests which were not being included in
run-tests.py. Broken tests were updated, with one exception which
is tracked by IMPALA-1862. Depends on a related change to
Impala-lzo.
Change-Id: I4c42498bdebf9155a8722695a3305b63ecc6e5f3
Reviewed-on: http://gerrit.cloudera.org:8080/194
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
Those tests were added as part of the new hash table implementation, as we didn't have
tests with right joins and duplicates (and other conjuncts) as well as aggregation
distinct queries with group bys on multiple columns. Adding them as a separate
patch, To improve testing coverage in the 2.2 release branch.
Change-Id: Id1b4f27fa6e587b2031635974ac9d2d39a1b015a
Reviewed-on: http://gerrit.cloudera.org:8080/193
Reviewed-by: Ippokratis Pandis <ipandis@cloudera.com>
Tested-by: Internal Jenkins
This patch fixes a regression introduced in:
c6907e4c2eabf5d73f83cc8e16b7f35a13c3b59f
IMPALA-1376: Split up Planner into multiple classes.
The problem was that in single-node planning the root analyzer
was passed to generate the plan for an INSERT/CTAS' query stmt.
The fix is to instead pass the stmt's analyzer that contains
information about evaluated constant predicates.
Change-Id: I551f471c978bc1f6bdff0d98e4826856d1e4860f
Reviewed-on: http://gerrit.cloudera.org:8080/191
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
The filesizes changed slightly, causing the S3 CI build to fail.
Let's regex the file sizes in the compute stats expected results.
Change-Id: Ie95bdf3a253a28aa2b6f3deb281948780ca2cc6a
Reviewed-on: http://gerrit.cloudera.org:8080/200
Reviewed-by: Ishaan Joshi <ishaan@cloudera.com>
Tested-by: Dan Hecht <dhecht@cloudera.com>
As needed, fix up file paths and other misc things to get
more test cases running against S3.
Change-Id: If4eaf9200f2abd17074080a37cd0225d977200ad
Reviewed-on: http://gerrit.cloudera.org:8080/167
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
The following commits disabled tests to unblock the full data load:
a00a9a5e53f7a8e7a1e3c931ea0e4b7db21c6f00
bf29d06f2e53bb924d250275d51f5ccd1213531d
This patch re-enables those tests and adds new tests to guard against
regressions to HIVE-6308.
Unfortunately, we cannot completely remove the analysis check for HIVE-6308
in our code, because there is still one case where COMPUTE STATS will fail on
a Hive-created Avro table: If there is a mismatch in column names between
the Avro schema and the column defs given to a CREATE TABLE in Hive.
Change-Id: I81ae6b526db02fdfc634e09eeb9d12036e2adfdd
Reviewed-on: http://gerrit.cloudera.org:8080/180
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
Verify DDL and queries when a table spans multiple filesystems
and across tables that live on different filesystems.
Change-Id: I4258bebae4a5a2758666f5c2e283eb2d205c995e
Reviewed-on: http://gerrit.cloudera.org:8080/166
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
Analytic function evaluation was broken when partition or order by exprs
evaluated to NaN (i.e. 0/0). We were relying on the comparison of the
current row with the previous row to be equal (i.e. x == x), but x != x
if x is NaN, and in the case of the very first row in the stream, some
logic breaks if x != x. The fix is to handle the very first row specially.
Change-Id: I1c33445d55a70c7f107f05eeadef272b7973ee11
Reviewed-on: http://gerrit.cloudera.org:8080/179
Reviewed-by: Matthew Jacobs <mj@cloudera.com>
Tested-by: Internal Jenkins
clause
This commit fixes the issue where if a query contains a count(distinct)
expression in conjunction with a limit, the limit is incorrectly applied
to the wrong place in the generated plan, thereby producing incorrect
results.
Change-Id: I776e1b78461323e7ab72d491dcec7a9acd9e75f9
Reviewed-on: http://gerrit.cloudera.org:8080/196
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
expressions
This commit enables fast partition pruning for cases where constant
expressions appear in binary or IN predicates. During partition pruning,
the constant expressions are evaluated in the BE and are replaced by the
computed results as LiteralExprs.
Change-Id: Ie8a2accf260391117559dc6c0a565f907c516478
Reviewed-on: http://gerrit.cloudera.org:8080/144
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
1) Fix up locations to take FILESYSTEM_PREFIX into account
so we can run the test against non-default FS.
2) Fix up results and catch sections.
3) Since S3 doesn't support INSERT, split the test into
another version that expects different results for the
INSERT part. The rest of the test is identical, and
we can remove this new .test file once INSERT is supported.
Change-Id: I50d21048b846aa985d1eefc50fc33bda05ebe509
Reviewed-on: http://gerrit.cloudera.org:8080/146
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
After the hive/hdfs rebase, the indexed lzo file names changed. This patch uses a
wildcard rather than a specific file name to protect against such changes. It's safe
because the test simply expects a partition that does not have index files.
Change-Id: I6d32609b62df83fe2a8ef935d7ca6506ecff5e0d
Reviewed-on: http://gerrit.cloudera.org:8080/150
Reviewed-by: Ishaan Joshi <ishaan@cloudera.com>
Tested-by: Internal Jenkins
In some cases, the lookup for the first_value_rewrite function could
return a different fn with the wrong (yet technically compatible) type.
first_value_rewrite takes 2 parameters, the first is the parameter to
first_value, and the second is an integer explicitly added by the FE during
analysis. first_value_rewrite has signatures where the first parameter can
be of any type and the second is always a BIGINT. When the FE inserts the
second parameter, it may not actually be a BIGINT (e.g. it could be a
TINYINT, SMALLINT, etc.). In this case, the function arguments will not
match one of the registered signatures exactly and some compatable function
will be returned. For example, FIRST_VALUE(1.1) has a DECIMAL parameter
and if a TINYINT/SMALLINT/INT parameter is added for the rewrite, then
the first_value_rewrite fn lookup happened to match the fn taking a
FLOAT and BIGINT. (Ideally DECIMAL would not be implicitly castable to
a FLOAT/DOUBLE, but NumericLiterals do allow this casting.)
As a result, the agg fn actually returned a FLOAT while the AnalyticExpr
was of the type DECIMAL, and the analytic tuple contained a DECIMAL slot
which would throw a DCHECK in the BE (or perhaps crash in retail builds).
This fixes the issue by setting the NumericLiterals to be explicitlyCast,
i.e. the type will not change if re-analyzed. Then the correct fn signature
is found.
Change-Id: I1cefa3e29734ae647bd690263bb63f08f10ea8b9
Reviewed-on: http://gerrit.cloudera.org:8080/136
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
Add skip markers for S3 that can be used to categorize the tests that
are skipped against S3 to help see what coverage is missing. Soon
we'll be reworking some tests and/or adding new tests to get back the
important gaps.
Also, add a mechanism to parameterize paths in the .test files, and
start using these new variables. This is a step toward enabling some
more tests against S3.
Finally, a fix for buildall.sh to stop the minicluster before applying
the metastore snapshot. Otherwise, this fails since the ms db is in
use.
Change-Id: I142434ed67bed407e61d7b2c90f825734fc0dce0
Reviewed-on: http://gerrit.cloudera.org:8080/127
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
Allow values larger than 64KB to be written to Parquet files. This was
previously limited by a fixed data page size. This commit removes that
limitation by allowing the page size to grow when necessary. This occurs
when there are enough unique values to switch from dictionary encoding
to plain encoding, and then there are huge values larger than the
default 64KB page size. In this case, it may be possible to write files
larger than one HDFS block, but this is an edge case and not worth
introducing additional complexity to handle.
Change-Id: I165ef44ba48ff0c3c3203860157a61c45f77df8b
Reviewed-on: http://gerrit.cloudera.org:8080/120
Reviewed-by: Skye Wanderman-Milne <skye@cloudera.com>
Tested-by: Internal Jenkins
When an analytic fn does not contain partition or order by exprs (i.e. empty
OVER() clause), we should not be comparing the previous and current rows.
It is not necessary because the analytic fn is applied to the entire input,
and attempting to access the child tuples could reference invalid memory
because there might be nullable tuples. When there are either partition or
orderby exprs, then there is a sort node preceding the analytic node and the
sort node always produces non-null tuples (though tuples may have all null
slots).
Change-Id: I5788295682b4c9a1dd8a3078e11da5767f12214c
Reviewed-on: http://gerrit.cloudera.org:8080/129
Reviewed-by: Ippokratis Pandis <ipandis@cloudera.com>
Tested-by: Internal Jenkins
The change of the PARTITION_FANOUT from 32 to 16 exposed a pathological case due to
the lack of coordination across concurrently executing spilling nodes of the same query.
In particular, when we repartition a partition we try to initialize hash tables for the
new partitions. But each hash table needs a block (for the nodes). In case there were not
any IO-sized blocks available, because they had been consumed by other nodes, we would get
into a loop trying to repartition those smaller partitions that couldn't initialize their
hash table. Additional repartitions that, among others, would need additional blocks for
the new streams. These partitions would end up being very small, still we would fail the
query when we were reaching the MAX_PARTITION_DEPTH limit, which was fixed to 4.
This patch fixes the problem by initializing the hash tables during repartitions with
small pages. That is, the hash tables always first use a 64KB and a 512KB block for their
nodes before switching to IO-sized blocks. This helps the partitioning algorithm to
finish when we end up with partitions that can fit in those small pages. The performance
may not be optimal, still the memory consumption is lower and the algorithm finishes. For
example, without this patch and with PARTITION_FANOUT == 16 in order to run TPC-H Q18 and
Q20 we needed 3.4GB and 3.1GB respectively. With this patch TPC-H Q18 needs ~1GB and Q20
975MB.
This patch also removes the restriction of stopping repartitioning when we are reaching
4 levels of repartitioning. Instead, whenever we repartition we compare the size of
the input partition to the size of the largest new partition. If there is no reduction
on the size we stop the algorithm. Otherwise, we keep on repartitioning. That should
help in cases of skew (e.g. due to bad hashing). There is a new MAX_PARTITION_DEPTH limit
of 16. It is very unlikely we will ever hit this limit.
Change-Id: Ib33fece10585448bc2d07bb39d0535d78b168ccc
Reviewed-on: http://gerrit.cloudera.org:8080/119
Reviewed-by: Ippokratis Pandis <ipandis@cloudera.com>
Tested-by: Internal Jenkins
The S3 work really enabled any Hadoop FileSystem to work with Impala,
but a small tweak is needed for LocalFileSystem due to how the Hadoop
Path code deals with URIs that don't have an authority component.
While we aren't claiming support for arbitrary FileSystem's at this
tiem, it is useful to test this. Since the S3 testing is done as a
nightly test rather than pre-checkin, we can use the LocalFileSystem to
regression test that:
1) Impala can access table data living on a secondary filesystem,
i.e. not the filesystem specified by fs.defaultFS.
2) Impala does not make assumptions that the filesystem has type
DistributedFileSystem.
Change-Id: Ie9b858ea440c9b3b332602e034c8052b168c57da
Reviewed-on: http://gerrit.cloudera.org:8080/121
Reviewed-by: Martin Grund <mgrund@cloudera.com>
Tested-by: Internal Jenkins
This change adds support for column lineage logging in Impala to be
consumed by Navigator. This feature is disabled by default and is enabled
by setting the -lineage_event_log_dir flag. When lineage logging is
enabled, the serialized column lineage graph is computed for each query
and stored in a specialized log file in JSON format.
Change-Id: Ib8d69cdbcc435be1e9c9694998c1d33ec1245b10
Reviewed-on: http://gerrit.cloudera.org:8080/70
Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Tested-by: Internal Jenkins
These are the backend changes necessary for reading structs in Parquet
files. I wrote this against Alex's preliminary frontend work, and
ad-hoc tables containing structs work. We won't be able to add
automated tested until the FE changes are in as well, but I'd like to
get these changes in so we can at least get converage of our existing
workloads.
The bulk of the changes are in the Parquet scanner. The rest is around
changing the column index of a slot descriptor to a column path, in
order to support nested columns.
Change-Id: Ifbd865b52c2b4679d81643184b1f36bf539ffcfd
Reviewed-on: http://gerrit.cloudera.org:8080/62
Reviewed-by: Marcel Kornacker <marcel@cloudera.com>
Tested-by: Internal Jenkins
Currently, we only consider files hidden if they have the special
prefixes "." or "_". However, some tools use special suffixes
to indicate a file is being operated on, and should be considered
invisible.
This patch adds the following hidden suffixes:
'.tmp' - Flume's default for temp files
'.copying' - hdfs put may produce these
Change-Id: I151eafd0286fa91e062407e12dd71cfddd442430
Reviewed-on: http://gerrit.cloudera.org:8080/80
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
Specifically:
- Hive needs some jars from hadoop/tools/lib
- Hive has an dependency on apache.snapshots ( added in fe/pom.xml )
- Beeline has to explicitly told not to use jline.
Change-Id: Id38956b748f8f667a39505c92355f0298f308718
Conflicts:
testdata/bin/load-hive-builtins.sh
The HMS does not properly migrate column stats when moving a table across
databases (HIVE-9720). Apart from losing the stats, this issue also prevents
the newly renamed table from being dropped.
To work around the issue, this patch manually drops+adds the column stats
when renaming a table across databases.
Change-Id: If901c5d1e9a6b2cedc35034a537f18c361c8ffa1
Reviewed-on: http://gerrit.cloudera.org:8080/72
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
Additionally, this patch also disabled the hbase/none test dimension if the
TARGET_FILESYSTEM environment variable is set to either s3 of isilon.
Change-Id: I63aecaa478d2ba9eb68de729e9640071359a2eeb
Reviewed-on: http://gerrit.cloudera.org:8080/74
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
Some Hadoop filesystems, like the S3-based ones, are not block based.
Since Impala derives scan ranges from file blocks, synthesize file
blocks for these filesystems. Otherwise, files are always assigned a
single scan range, limiting parallelism.
An alternate approach would be to modify the planner's compute scan range
code. However, there would be some downsides to that approach: (a) we'd
need to plumb through more information from the catalog to the frontend,
increasing the catalog size, and (b) we'd be doing more work on each
query rather than once at metadata load time.
Change-Id: If53cba6e25506545eae78190601fbee0147547b3
Reviewed-on: http://gerrit.cloudera.org:8080/54
Reviewed-by: Dan Hecht <dhecht@cloudera.com>
Tested-by: Internal Jenkins
The problem was that VARCHAR was not present in a few switch statements
for updating/populating column stats in various places.
Change-Id: I0b2b316b734d27a7ff08701b0986014be2473443
Reviewed-on: http://gerrit.cloudera.org:8080/65
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: Internal Jenkins
This should fix the last y2k38 problem. Previously calling
unix_timestamp() with a input of '2038-01-19 03:14:08' or later would
return a negative value due to a 32 bit int overflow. This patch
switches from 32 to 64 bit ints.
Change-Id: Ic9180887d6c828f6ecd25435be86fd0bd52d3f0d
Reviewed-on: http://gerrit.cloudera.org:8080/61
Reviewed-by: Casey Ching <casey@cloudera.com>
Tested-by: Internal Jenkins
In the case, that a cached table was created in a version of Impala that
did not have the property for the cache replication factor, the loading
of the table will fail until the table is un-cached and cached again.
This patch fixes this behavior and ignores this missing parameter.
Change-Id: I118020dd5bd7fb203d91853d5ef946f2c4c8a695
Reviewed-on: http://gerrit.cloudera.org:8080/48
Reviewed-by: Martin Grund <mgrund@cloudera.com>
Tested-by: Internal Jenkins
First change for IMPALA-1209 to address Impala limitations when
using HDFS encryption. This adds a KMS process to the testdata
cluster. This was tested manually by creating a key and an
encryption zone.
Change-Id: I499154506386f04e71c5371b128c10868b1e1318
Reviewed-on: http://gerrit.cloudera.org:8080/41
Reviewed-by: Matthew Jacobs <mj@cloudera.com>
Tested-by: Internal Jenkins
No changes to writing were made. No changes to reading Impala written
files were made.
Hive writes TIMESTAMP values to parquet files differently than Impala
does. Hive converts the value from local time to UTC before writing;
Impala does not. This change adds a startup flag that will convert UTC
to local when reading files written by Hive.
The Hive-file detection actually checks for "parquet-mr" (which is the
library Hive uses) in the file metadata. A slight possibility exists
that TIMESTAMP values written by something other than Hive but also
using parquet-mr may become incorrect. The possibility should be very
small because TIMESTAMP values are stored and encoded in a non-standard
way other applications are unlikely to be aware of.
Flags from be/src/exec/hdfs-parquet-scanner.cc:
-convert_legacy_hive_parquet_utc_timestamps (When true, TIMESTAMPs
read from files written by Parquet-MR (used by Hive) will be
converted from UTC to local time. Writes are unaffected.) type: bool
default: false
Change-Id: I79a499fe24049b7025ee2dd76c9c3e07010d346a
Reviewed-on: http://gerrit.cloudera.org:8080/35
Reviewed-by: Casey Ching <casey@cloudera.com>
Tested-by: Internal Jenkins
When a table is loaded in the catalog, we will now perform a check to
verify that the cache directive ID and cache replication factor is still
valid and the data is current.
If the cache directive does no longer exist, we issue a error message
and mark the table / partition as uncached. Furthermore, the replication
factor is updated with the information from the actual cache directive.
In the case of insert statement there is a special situation as the
catalog update is happening synchronously and will try to access the
cache directive information that might be stale. Thus in this insert
path, we catch the possible not found exception and reset the caching
information.
Change-Id: I882041ce5395b8a3d17e9fc2750053393340df65
Reviewed-on: http://gerrit.cloudera.org:8080/40
Reviewed-by: Martin Grund <mgrund@cloudera.com>
Tested-by: Internal Jenkins
We build some of the jars that hive needs fe/target/. A recent change resulted in these
jars not being loaded, causing a bad hive environment. This patch restores proper
behaviour.
Change-Id: Icb27ab04f7f77cb4ddab51326eedfd11a6cdf960
Reviewed-on: http://gerrit.sjc.cloudera.com:8080/5930
Reviewed-by: Alex Behm <alex.behm@cloudera.com>
Tested-by: jenkins
