on linux, unicode parameters cannot be longer than 128

I noticed the following when posting a query from linux using parameters:

If the parameter string is longer than 127 characters, pypyodbc will throw an error on row 1572 when trying this:

param_buffer.value = c_char_buf

It complains that c_char_buf is too large.

I checked: c_char_buf is created by UCS_buf() if parameter type is ‘u’ (unicode less than 255 characters):

c_char_buf = UCS_buf(param_val)
c_buf_len = len(c_char_buf)

UCS_buf(string) is defined to return encoding(utf-16-le) for linux, and return its argument for windows.

Thus on linux, due to 2 byte minimum char width for utf-16, the length of the byte array will be double the length of the underlying string.

As param_buffer.value size is defined as 255 no matter what platform, this leads to above error.

My fix was to double the buffer size for linux:

 if param_types[col_num][0] == 'u':
                sql_c_type = SQL_C_WCHAR
                sql_type = SQL_WVARCHAR 
                if sys.platform not in ('win32','cli'):
                          buf_size = 255 * 2 # double buffer size for linux                 
                else:
                         buf_size = 255
                ParameterBuffer = create_buffer_u(buf_size)   

Hope this helps someone.

[timjen3]

An alternative that works for me is to pass encoded bytes as parameters. I’m working with SQL Server and pypyodbc 1.3.4.3

For example:

cursor = pypyodbc.connect(connection_string).cursor()
city= "Louisville".encode("utf-8", "ignore")[:32]
state = "KY".encode("utf-8", "ignore")[:2]
uid = "UID123".encode("utf-8", "ignore")[:16]
cursor.execute("UPDATE MyTable SET City=?, State=? WHERE Id=?", (city, state, uid))

I purposefully left the byte-trunacting in (ie: [:32]) because that works properly.

[dmfreemon]

Related, see PR: https://github.com/jiangwen365/pypyodbc/pull/73/files