C++ and Java TPCH Implementations Generate Different Random Data

[tdcmeehan]

The C++ (Velox) and Java TPCH implementations generate different random strings for varchar columns, making exact data comparison impossible.

Your Environment

• Presto version used: Latest
• Storage (HDFS/S3/GCS..): N/A
• Data source and connector used: TPCH
• Deployment (Cloud or On-prem): N/A
• Pastebin link to the complete debug logs: N/A

Expected Behavior

The data should match between both connectors.

Current Behavior

The data is different.

Possible Solution

Fix the TPCH Velox connector to match the Java implementation.

Steps to Reproduce

-- Java
  SELECT custkey, address FROM tpch.tiny.customer WHERE custkey < 10 ORDER BY custkey;
   custkey |                address
  ---------+---------------------------------------
         1 | IVhzIApeRb ot,c,E
         2 | XSTf4,NCwDVaWNe6tEgvwfmRchLXak
         3 | MG9kdTD2WBHm
         4 | XxVSJsLAGtn
         5 | KvpyuHCplrB84WgAiGV6sYpZq7Tj
         6 | sKZz0CsnMD7mp4Xd0YrBvx,LREYKUWAh yVn
         7 | TcGe5gaZNgVePxU5kRrvXBfkasDTea
         8 | I0B10bB0AymmC, 0PrRYBCP1yGJ8xcBPmWhl5
         9 | xKiAFTjUsCuxfeleNqefumTrjS

  -- C++
  SELECT c_custkey, c_address FROM tpchstandard.tiny.customer WHERE c_custkey < 10 ORDER BY c_custkey;
   c_custkey |               c_address
  -----------+---------------------------------------
           1 | j5JsirBM9PsCy0O1m
           2 | 487LW1dovn6Q4dMVymKwwLE9OKf3QG
           3 | fkRGN8nY4pkE
           4 | 4u58h fqkyE
           5 | hwBtxkoBF qSW4KrIk5U 2B1AU7H
           6 |  g1s,pzDenUEBW3O,2 pxu0f9n2g64rJrt5E
           7 | 8OkMVLQ1dK6Mbu6WG9 w4pLGQ n7MQ
           8 | j,pZ,Qp,qtFEo0r0c 92qobZtlhSuOqbE4JGV
           9 | vgIql8H6zoyuLMFNdAMLyE7 H9

Screenshots (if appropriate)

Context

I tried to write some tests that compare the results between the two connectors and encountered this issue.

[tdcmeehan]

Corresponding Velox issue: facebookincubator/velox#14875