fix/ORA-00907 when comparing ref cursors with BINARY_... columns

[opendba]

fixing "ORA-00907: missing right parenthesis" when comparing ref cursors with BINARY_DOUBLE/BINARY_FLOAT columns, because these column types do not support type length

Stack trace:

1. ut_get_data
   ORA-00907: missing right parenthesis
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 318
   ORA-06512: at "UT3.UT_COMPOUND_DATA_HELPER", line 596
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 309
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 366
   ORA-06512: at "UT3.UT_EQUAL", line 225
   ORA-06512: at "UT3.UT_EXPECTATION", line 26
   ORA-06512: at "UT3.UT_EXPECTATION", line 138
   ORA-06512: at "TDS2.UT_TDSPATANSINTPKG", line 117
   ORA-06512: at "TDS2.UT_TDSPATANSINTPKG", line 117
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 318
   ORA-06512: at "UT3.UT_COMPOUND_DATA_HELPER", line 596
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 309
   ORA-06512: at "UT3.UT_DATA_VALUE_REFCURSOR", line 366
   ORA-06512: at "UT3.UT_EQUAL", line 225
   ORA-06512: at "UT3.UT_EXPECTATION", line 26
   ORA-06512: at "UT3.UT_EXPECTATION", line 138
   ORA-06512: at "TDS2.UT_TDSPATANSINTPKG", line 117
   ORA-06512: at "TDS2.UT_TDSPATANSINTPKG", line 117

[lwasylow]

Hi thanks for looking into this and catching that error. Do you know what's coming out of cursor desc? The else condition should not put a len when column_len is not null. Would probably be best to look why it didn't catch that and fix that line to be more generic for other types. Would be also very useful to provide a test with failing scenario, as a regression stop and showcase.

else 
      l_col_type := a_data_info.column_type
        ||case when a_data_info.column_len is not null
          then '('||a_data_info.column_len||')'
          else null
          end;
    end if;

[opendba]

Hi,

Thanks for review.

a_data_info.column_len is 8 for BINARY_DOUBLE column and this is the generated query part that fails, since oracle does not want type len to be specified:

 ,"PI_DOUBLE" BINARY_DOUBLE(8) PATH 'PI_DOUBLE' , "PI_FLOAT" BINARY_FLOAT(4) PATH 'PI_FLOAT'

I've added simple test case based on success_on_same_data that tries to compare binary_% data types, but fails with ORA-907 without this fix.

[jgebal]

Can you also create issue for this?
It makes it easier to track releases.

[opendba]

Added issue for this: #899

[jgebal]

@opendba This looks great.
Thank you.
It doesn't often happen that users fix tge bugs the find, usually they only report them.

Great contribution!

[jgebal]

Resloves #899

[jgebal]

@lwasylow
Are you ok to merge this PR?

[opendba]

it was a very simple fix.

[lwasylow]

Yes no problem