Add support for reading orc column statistics

[wypb]

Currently, the dwrf module only supports reading the column statistics of dwrf format, and does not support reading the column statistics of orc format. When I use velox to read the orc table, the following exception occurs:

presto:tpch_velox> create table region_orc with(format = 'ORC') as select * from tpch.sf10.region;
CREATE TABLE: 5 rows

Query 20230804_032253_00003_tubug, FINISHED, 1 node
Splits: 22 total, 22 done (100.00%)
0:35 [5 rows, 0B] [0 rows/s, 0B/s]

presto:tpch_velox> select * from region_orc;

Query 20230804_043524_00023_m8qw7, FAILED, 1 node
Splits: 2 total, 0 done (0.00%)
0:40 [0 rows, 0B] [0 rows/s, 0B/s]

Query 20230804_043524_00023_m8qw7 failed:  Operator::getOutput failed for [operator: TableScan, plan node ID: 0]: vector

with this pr, we support read orc column statistics, than we can read orc data through velox:

presto:tpch_velox> select * from region_orc;
 regionkey |    name     |                                                       comment
-----------+-------------+---------------------------------------------------------------------------------------------------------------------
         0 | AFRICA      | lar deposits. blithely final packages cajole. regular waters are final requests. regular accounts are according to
         1 | AMERICA     | hs use ironic, even requests. s
         2 | ASIA        | ges. thinly even pinto beans ca
         3 | EUROPE      | ly final courts cajole furiously final excuse
         4 | MIDDLE EAST | uickly special accounts cajole carefully blithely close requests. carefully final asymptotes haggle furiousl
(5 rows)

Query 20230804_044016_00026_m8qw7, FINISHED, 1 node
Splits: 2 total, 2 done (100.00%)
0:14 [5 rows, 603B] [0 rows/s, 44B/s]

presto:tpch_velox>

CC: @Yuhta

[netlify]

✅ Deploy Preview for meta-velox canceled.

Name	Link
🔨 Latest commit	48f1ee0
🔍 Latest deploy log	https://app.netlify.com/sites/meta-velox/deploys/666a560f236a4b00092a640b

[Yuhta]

We have some problem here, the wrapper is just a reference but here we are returning actual values. If ORC does not need it, can we keep the signature and make it DWRF only (with a check)?

[wypb]

@Yuhta Thank you for your review, It should be that temporary local variables are used here, I will modify it.

[wypb]

Hi @Yuhta, Thank you for your review, I have modified the problematic code and the code has been rebased. You can check this diff for the relevant modified code.

[Yuhta]

We still have the same problem, I think for this function you should keep it as it is, and create another one returning proto::orc::ColumnStatistics if it is required by ORC as well. Wrappers can only be created on call sites.

[wypb]

@Yuhta I looked at the code again, ORC does not use this method. I'll keep the previous code logic. Thank you for testing this.

[Yuhta]

A better name is dwrfStatistics

[wypb]

Fixed, thanks.

[wypb]

Hi @Yuhta Do you have any other comment for this PR? thank you.

[wypb]

Hi @Yuhta PTAL, Thank you.

[wypb]

HI @kevinwilfong I moved the implementation of options() and context() to ColumnStatisticsBase.h file to reduce duplication of code. I also synchronized the latest code of the main branch to make CI green. please help me review it again when you have a chance. Thanks!

[facebook-github-bot]

@kevinwilfong has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[facebook-github-bot]

@kevinwilfong merged this pull request in 0c0a973.

[conbench-facebook]

Conbench analyzed the 1 benchmark run on commit 0c0a9737.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details.