On May 18, 2010, at 10:06 PM, Dan Kuebrich wrote:
> My apologies; I expressed my question rather incoherently.
>
> if the question is, I want to multiple insert like [{'a':1, 'b':2, 'c':3},
> {'a':2}, {'a':3, 'b':4}], etc. only some dictionaries are missing different
> keys, that is not allowed. The structure of the SQL statement as parsed by
> MySQL determines for which columns the server-side default is emitted and
> there is only one version of that with an executemany. So every param
> dictionary must have at least all the keys which the first one does.
>
> This was the basic scenario of my question. However, unless I misunderstand,
> it does seem to be "allowed": the missing values are replaced with NULL in
> the generated query.
it should not be in SQLA 0.6. Specific logic was added to disallow this.
> But what I actually meant to ask was this: sqlalchemy handles missing
> dictionary entries as NULL for the purpose of inserts, is there an easy way
> to have it generate a SQL statement with DEFAULT for missing dictionary
> entries?
>
> Eg. execute(testtable.insert(), [{'a':1, 'b':2, 'c':3}, {'a':2}, {'a':3,
> 'b':4}]) => INSERT INTO testtable (a,b,c) VALUES
> (1,2,3),(2,DEFAULT,DEFAULT),(3,4,DEFAULT)
ive never seen that syntax before (i.e. "DEFAULT" is actually present as a
value). Assuming its valid, you could achieve it using a "client side"
default as I mentioned earlier. i.e. Column('foo', Integer,
default=text('DEFAULT')). But again, not with an executemany() and
heterogeneous dictinoaries as you have above. The SQL statement is rendered
only once, and either has a bind parameter for a particular position or not.
--
You received this message because you are subscribed to the Google Groups
"sqlalchemy" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en.
