David G. Johnston wrote: > > > It may be worth updating the docs here... > > > So, my first pass at this. I'm probably going a bit outside what would normally be covered in the SQL > Command section but it does feel right at first blush. > > Also attached; not compiled. > > As a bug fix I've updated the description of "COMMIT" here since it can cause a ROLLBACK to be issued and that isn't documented. > > "(pseudo) sub-transaction" seemed like a reasonable choice of terminology to introduce rather than just "mark". > Having it mentioned in context in the BEGIN docs, instead of just a "see also", should aid in understanding > how the whole transaction system fits together. The advanced features of the tutorial cover this to some degree > (I didn't re-read it prior to writing this up) but I'm inclined to believe people wanting to understand transactions, > lacking a top-level chapter on the topic, will know of BEGIN and start their discovery there. I think that it is a good idea to explain the behavior of aborted transactions. Shouldn't that go to the hackers list though? I don't like the term "pseudo sub-transaction". What's pseudo about it? > + > + <para> > + <application>psql</application> makes use of savepoints to implment its > + <literal>ON_ERROR_ROLLBACK</literal> behavior. > + </para> s/implment/implement/ Yours, Laurenz Albe
