Join issue in BigQuery with two million row tables

Question

I posted this in the BigQuery issue tracker: (please star the issue if it affects you) https://code.google.com/p/google-bigquery/issues/detail?id=89&q=join%20each

What steps will reproduce the problem?

1. See job personal-real-estate:job_up2I9A31Bo8NSvwD0XTWG2tBoVA
2. I run

SELECT *        FROM  
      (select *,integer(AD_STREET_NO_PROP) as str_no_prop, integer(CD_ADDR_ZIP_PROP) as CD_ADDR_ZIP_PROP1 from [acris_nyc.nyc_dof_SOA] 
      where NM_RECIPIENT_1 like '%THE MICHAEL R. BLOOMBERG REVOCABLE%') AS s 
        JOIN   each  
      (select *,integer(hnum_lo) as str_num,integer(zip) as zip1 from [acris_nyc.nyc_dof_tc_Tentative_Assessment_Roll] where owner like '%BLOOM%' and txcl = '1') AS a  
      on s.str_no_prop = a.str_num and s.ad_street_1_prop = a.str_name order by NEW_FV_T desc limit 100

What is the expected output? What do you see instead?

I expect one record to be returned.

containing

17 as the str_num and "EAST 79 STREET" as the str_name

What version of the product are you using? On what operating system?

BigQuery on 4/22/2014 from chrome browser

Please provide any additional information below.

I try a very similar query on a much smaller set of tables and it works as expected.

SELECT *  FROM (select *, integer(number) as inumber from [test_1.table1] where owner like '%BLOOM%') as a join each (select *, integer(number) as inumber from [test_1.table2] where owner like '%BLOOM%') as b on a.inumber=b.inumber and a.street = b.street

returns 

Row a_number    a_street    a_owner a_inumber   b_number    b_street    b_owner b_inumber    
1   00000017    EAST 79 STREET  BLOOMBERG, MICHAEL R    17  17  EAST 79 STREET  THE MICHAEL R. BLOOMBERG REVOCABLE  17   

If I query the individual tables in the 1 million row case they contain the data that should match when the join completes.

Is there any way to debug the actual join operation?

Thanks.

Answer 1

Just to close the loop on this question; after investigation it turned out to be an error in the data that was masked by automatic-whitespace-deletion that was being done in the browser. See https://code.google.com/p/google-bigquery/issues/detail?id=89&q=join%20each for more information.