You are misusing the binding.
There are three different ways of binding variables with cx_Oracle as one can see here :
1) by passing a tuple to a SQL statement with numbered variables :
sql = "select * from sometable where somefield = :1 and otherfield = :2"
cur.execute(sql, (aValue, anotherValue))
2) By passing keyword arguments to a SQL statement with named variables :
sql = "select * from sometable where somefield = :myField and otherfield = :anotherOne"
cur.execute(sql, myField=aValue, anotherOne=anotherValue)
3) By passing a dictionary to a SQL statement with named variables :
sql = "select * from sometable where somefield = :myField and otherfield = :anotherOne"
cur.execute(sql, {"myField":aValue, "anotherOne":anotherValue})
Remarks
Why does your code work then ?
Let’s try to understand what happens here :
bind= {"var" : "ciao"}
sql = "select * from sometable where somefield = :bind and otherfield = :bind"
cur.execute(sql,(bind["var"], bind["var"]))
Oracle will understand that it expects one variable. This is a named variable, linked by name bind. You should then give a parameter as a named parameter like this :
cur.execute(sql, bind="ciao")
Or using a dictionary, like that :
cur.execute(sql, {bind:"ciao"})
However, as cx_Oracle receives a tuple instead, it fallbacks in a binding by number, as if your SQL statement was :
sql = "select * from sometable where somefield = :1 and otherfield = :2"
And as you are passing bind['var'] twice, which is just the string "ciao". It is mapping the two tuple items to the numbered variables :
cur.execute(sql, ("ciao", "ciao"))
That runs by chance but the code is very misleading.
Tuple with a single value to bind
Also note that the first option requires a tuple. But if you have a single value to bind, you can use this notation to create a tuple of a single value :
sql = "select * from sometable where somefield = :1"
cur.execute(sql, (aValue,))
[EDIT] : Thanks to @tyler-christian for mentioning that passing a dict was supported by cx_Oracle.