On 2015-07-09 09:54, Chetan Mehrotra wrote:
Hi,
Looking at RDBDocumentStore it appears that it is using same table
schema for all collections. For e.g. columns like deletedOnce,
hasBinary are only required for NodeDocument. However they are present
in the tables
Yes.
Any specific reason for doing this and not going for schema per collection?
Simplicity and the complete lack of contract. How would the DS
implementation *know* what needs to be indexed?
This is fine for small collection like settings and clusterNodes. But
for bigger collection like journal the overhead of such empty columns
might be large. It would be better if each Document provides a set of
"might be large" is guesswork, no? The additional columns are all
numbers/flags.
column names along with types to be indexed and then RDBDocumentStore
create the correct schema.
Each *document*? Did you mean collection?
Best regards, Julian
