Pandas deserialization fails whenever kdb sends it a single row table where one of the columns has a blank char.
To reproduce:
from qpython import qconnection
q = qconnection.QConnection(pandas = True, host = 'myhost', port = myport)
print 'initiating connection(s)'
q.open()
while True:
msg = q.receive(data_only = True, raw = False)
print 'message received'
Now, on the kdb side:
/send a table (which just so happens to have blank/null chars)
neg[4](`upd;`trade;([] col1:`a`b`c;col2:" a"))
/no problem
/send last row of that table
neg[4](`upd;`trade;-1#([] col1:`a`b`c;col2:" a"))
/no problem
/send two rows (2 blank chars)
neg[4](`upd;`trade;2#([] col1:`a`b`c;col2:" a"))
/no problem
/send first row of that table (one blank char)
neg[4](`upd;`trade;1#([] col1:`a`b`c;col2:" a"))
/crashes with error "AttributeError: 'float' object has no attribute 'meta'"
As described in: http://stackoverflow.com/questions/39536229/issue-with-null-character-received-by-qpython-pandas-from-kdb
Pandas deserialization fails whenever kdb sends it a single row table where one of the columns has a blank char.
Now, on the kdb side: