ARROW-11973 [Rust][DataFusion] Boolean kleene kernels#9772
ARROW-11973 [Rust][DataFusion] Boolean kleene kernels#9772ch-sc wants to merge 6 commits intoapache:masterfrom
Conversation
|
@ch-sc thanks for your PR In the JIRA ticket it is mentioned that these semantics are default for SQL - however - these return I think it is a useful addition to Arrow side though - for DataFusion we should just think about what behavior we want to support. |
|
Hi @Dandandan, I actaully checked on Postgres (11) beforehand and it did return exactly the values I provided in the table. Which Postgres version are you using? |
|
I tried on mysql, postgres and big query and was happy with the table you provided on JIRA. I am also curious how I got a different result than @Dandandan xD |
|
Haha @ch-sc @jorgecarleitao I tried it on this website which I used before for some simple checks https://rextester.com/l/postgresql_online_compiler So forget what I said, it is very useful to do this both for Arrow & DataFusion 👍 |
Codecov Report
@@ Coverage Diff @@
## master #9772 +/- ##
==========================================
- Coverage 82.59% 82.36% -0.24%
==========================================
Files 248 252 +4
Lines 58294 58893 +599
==========================================
+ Hits 48149 48507 +358
- Misses 10145 10386 +241
Continue to review full report at Codecov.
|
|
@Dandandan @jorgecarleitao the integration tests fail because of scarce disk space: Any ideas what we can do about this? |
|
For the record the issue of the integration test running out of space is known and not related to this PR |
[PR 9772](#9772) introduced a bug. The boolean kleene kernel would not iterate over the bit chunks of the batch, if there is no validity bitmap on the left or the right input. It will only process the bits of the remainder word. The inital unit test didn't pick a large enough batch size to test this scenario, which is why this was not detected by the tests earlier. Closes #9965 from ch-sc/ARROW-12294-boolean-kleene-kernels-no-remainder Authored-by: Christoph Schulze <christoph.schulze@signavio.com> Signed-off-by: Andrew Lamb <andrew@nerdnetworks.org>
[PR 9772](apache/arrow#9772) introduced a bug. The boolean kleene kernel would not iterate over the bit chunks of the batch, if there is no validity bitmap on the left or the right input. It will only process the bits of the remainder word. The inital unit test didn't pick a large enough batch size to test this scenario, which is why this was not detected by the tests earlier. Closes #9965 from ch-sc/ARROW-12294-boolean-kleene-kernels-no-remainder Authored-by: Christoph Schulze <christoph.schulze@signavio.com> Signed-off-by: Andrew Lamb <andrew@nerdnetworks.org>
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #9772 +/- ##
==========================================
- Coverage 82.59% 82.36% -0.24%
==========================================
Files 248 252 +4
Lines 58294 58893 +599
==========================================
+ Hits 48149 48507 +358
- Misses 10145 10386 +241 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
This PR adds two boolean kernels
kleene_orandkleene_and.As described in the corresponding JIRA ticket, the kleene operator handels null values differently compared to plain OR and AND operators.