Girish Sharma wrote:I don't have the cited reference available, but I suspect you must be overlooking some caveats. It can't be that simplistic.
Undo generation quantity (DESC order) :
Page 323 Expert Oracle Database Architecture 9I& 10G Prog. By Thomas Kyte
I don't have the cited reference available,But, I have [url http://books.google.co.in/books?id=_JVb4S-KcCcC&pg=PA323&lpg=PA323&dq=which+dml+generates+more+undo&source=bl&ots=95or3g5elD&sig=_QpMvybc2GFOh8xpUV7eT4AQqhA&hl=en&sa=X&ei=7FFvUcK7KcjmrAfMq4DIBQ&ved=0CE0Q6AEwBA#v=onepage&q=which%20dml%20generates%20more%20undo&f=false] this one.
but I suspect you must be overlooking some caveats. It can't be that simplistic.agreed, but at the same time it should not be that simplistic when it is asked. If I asks you what you like most to eat from Biscuits, Chocolate, Burger, then probably you will choose one of them, but obviously it is not really that simplistic. May be you like to eat more Biscuits in summer than winter or vice - versa. So my question should be what you like to most to eat in summer/winter from <options>. Life and IT both are same i.e. each and everything, every time, every where "IT DEPENDS".
does the DELETE still generate the most undo?With due respect, obviously, when we are dealing with more rows, more undo will be generated, but again in the proportionate only , it is well known and well documented too; more over I am really surprised to see above example in this context from your side. How and why you and Mark are advocating for number of rows, while it seems OP is asking for a single row (question style) and if still OP seems to ask something different and my reply is not correct or wrong, I am really sorry 100 times.
rp0428 wrote:The ridiculous is intentional, to make the point painfully obvious. Once the point is made, we can negotiate the actual boundary conditions.
RP, I agree with Ed. The response should have included some qualifying statement like "when dealing with the same row of data" to be more specific to ensure the OP understands the limitations of the answer.
Girish's response did have 'some qualifying statement. He provide the reference to the expert resource that he used.
Everything is subject to interpretation. Sometimes you just need to use common sense. I don't consider it common sense to interpret OPs question to be asking about possibilities where one of those ops is for ONE row and one is for 100 million rows. That is just ridiculous on the face of it, IMHO. :D
Let's assume this is a test question:If it were a written, multi-choice test, I would concur. If it were an interview (the OP really hasn't said) or just a discussion among colleagues, I wouldn't allow myself to be constrained by multiple choice. I don't know what actually prompted the OP's question, but we'd all have to agree we've seen plenty of people come here either preparing for or doing a post-mortem on an interview and it is painfully obvious they treat an interview like it were an exam, seeming to be oblivious to the idea they might actually discuss a question and answer with the interviewer.
Which DML generates the most UNDO ?
Remember - this is a TEST. On tests you need to select 'the best, or most correct' answer.
I would choose 'B. DELETE' since I consider that to be the 'most correct' given the information available.
Why don't you and Ed tell us 'which of those three possible answers' you would choose?
RegardsNo need to apologize. The back-and-forth is educational for all. I certainly don't take any of it personally and I trust that neither do the other participants. The OP probably got much more than he expected, but if he has read through it all, he also has a far better understanding than if the only response he had gotten was simply "answer 'B'". He definitely got the "why" part of his question.