node-odbc causes error "CLI0002W Data truncated"

Question

I am using node-odbc to connect a Node.js-based application to a legacy IBM DB2 database.

• On Windows 7
• The DB2 ODBC drivers are 32bit, therefore node-odbc binaries have been created with node-gyp clean configure build --arch=ia32 --msvs_version=2015
• Node.js 4.4.5 in 32bit

This mostly works fine. However, I currently have Unicode disabled in the node-odbc binaries (flag UNICODE in binding.gyp) because of the following problem:

If I enable UNICODE, I can

1. Correctly retrieve data with simple SQL statements (like query = select * from sysibm.sysdummy1)
2. Run queries with non-string parameter binding (like query = select * from syscat.tables where tableid = ?, bindings = [0];

However, if I run a query with string parameters (like query = select * from syscat.tables where tabname = ?, bindings = ['SYSDUMMY1']), the ODBC driver throws the error message

{ [Error: [IBM][CLI Driver] CLI0002W Data truncated. SQLSTATE=01004] errors: [ { message: '[IBM][CLI Driver] CLI0002W Data truncated. SQLSTATE=01004', state: '01004' } ], error: '[node-odbc] SQL_ERROR', message: '[IBM][CLI Driver] CLI0002W Data truncated. SQLSTATE=01004', state: '01004' }

Its explanation at IBM documentation doesn't help me.

I looked at the UNICODE-enabled snippets in the relevant node-odbc code which write the string into the buffer which will be passed to the ODBC driver:

#ifdef UNICODE
      params[i].ParameterType     = SQL_WVARCHAR;
      params[i].BufferLength      = (length * sizeof(uint16_t)) + sizeof(uint16_t);
#else
      params[i].ParameterType     = SQL_VARCHAR;
      params[i].BufferLength      = string->Utf8Length() + 1;
#endif
      params[i].ParameterValuePtr = malloc(params[i].BufferLength);
      params[i].StrLen_or_IndPtr  = SQL_NTS;//params[i].BufferLength;

#ifdef UNICODE
      string->Write((uint16_t *) params[i].ParameterValuePtr);
#else
      string->WriteUtf8((char *) params[i].ParameterValuePtr);
#endif

Apparently, string->Write writes the string with a null-terminal into the buffer ParameterValuePtr. SQL_NTS means null-terminated string (see explanation of StrLen_or_IndPtr at IBM documentation).

Judging from the error message, I assume the buffer does not have the correct size, which is calculated as (length * sizeof(uint16_t)) + sizeof(uint16_t) (length of string in bytes for unicode characters + 1 additional character, the null terminator):

1. Either it's too small, so the ODBC driver cannot find the null character which terminates the string
2. Or too big, so maybe the drivers complains that the null character comes before the end of the buffer. This is really an assumption!

So: I wonder how the code should be corrected so that string-like parameters can be passed to the ODBC driver. Any ideas?