Allow per-column output converters

[mkleehammer]

Global converters can already be configured by SQL type, but sometimes it might be useful to have a per-column setting. For example, a single column might be encrypted or compressed.

These would need to be configured with the execute or immediately after and would only apply for the single SQL statement.

Prompted by a comment in #160 .

[Lexcon]

Mike, just to share this: Py3.6 supports an ANSI encoding with is basically 'bytes'. Using this encoding it's able to compress/encrypt a string, store and retrieve it binary. Not per-column yet as the intended change but quite handy.

Config: windows10, py36, pyodbc4.0.0, mssql server 2014 express

Example:
import pyodbc, zlib

conn = pyodbc.connect('connection_string_here')

teststr = """Binary distributions for OS X and Windows are now uploaded to PyPI and can now be installed using pip. Source distributions are also uploaded and can be used by other platforms to compile. (PyPI only allows binary uploads for OS X and Windows.)"""
zipped = zlib.compress(teststr.encode('ansi'))

conn.setdecoding(pyodbc.SQL_CHAR, encoding='ansi')

value = zipped.decode('ansi')
cursor = conn.cursor()
cursor.execute("create table zippedcontent (zipped varchar(8000))")
cursor.execute("insert into zippedcontent (zipped) values (?)", value)
v = cursor.execute("select zipped from zippedcontent").fetchone()[0]
print(zlib.decompress(v.encode('ansi')).decode('ansi')) # yeah!

[Lexcon]

By the way: this is easily accomplished by adding the column name to GetDataUser and from there from pass the column name to the converter function, then you could do this:

def myconverter(content, fieldname):
__# converts ip adress b'127.0.0.1' to its numerical value, leaves other values unmodified
__if fieldname == 'ipaddress'
____return struct.unpack("!I", socket.inet_aton(content.decode('ascii'))[0]
__return content

connection.add_output_converter(pyodbc.SQL_VARCHAR, myconverter)