Trash - 2018-01-10

« first day (439 days earlier) ← previous day next day → last day (2254 days later) »

3:38 PM

you might appreciate this quote: "Of course I expected using a persisted computed column to be a clever solution without any overhead other than a compute scalar, not expose how stupidly these columns are implemented…"

Paul White

That sort of statement just annoys me. Who said it?

Joe Obbish

@PaulWhite A developer that I'm working with here. Why does it annoy you?

I understand his perspective very well. We're getting multiple scans of a table when filtering against a computed column even with TF 176

I'll work on a repro that I can share

Paul White

Because it assumes the people that wrote the implementation are stupid, which they're not. By and large, they're much cleverer than any of us, and unable to defend themselves.

Joe Obbish

@PaulWhite I understand your point of view. Personally I look at it a different way

It's not so personal

I like to think that I'm not stupid, but I personally write stupid code all the time

Paul White

This isn't about you.

Joe Obbish

3:43 PM

And it's software. There are lots of reasons why the ideal implementation couldn't be released

maybe the developers working on it didn't have enough time

The implementation does not meet the needs of the end users. Isn't that the most important thing?

Maybe I'll end up telling the developer to just construct the plan that he needs with a USE PLAN hint. That's not annoying?

Paul White

I have no problem with people asking for better solutions to real problems. I have a problem with people assuming others are stupid.

Joe Obbish

@PaulWhite Ok. Well if it bothers you feel free to delete the message

Paul White

@JoeObbish That won't unbother me :)

Joe Obbish

That wasn't intended though

for all we know it's a PM's fault

maybe some developer at MSFT is just as frustrated by it

anyway, apologies for causing offense

Paul White

I'm not offended, I'm irked. And it's a general irk, not directed at your coworker - I hear far too much of this sort of thing.

Now of course I agree that it would be nice to have persisted computed columns be "properly persisted" when required. I wrote a whole thing about that.

And the choice to expand could be more informed with better costing support.

And match-back could be better with an implementation tweak.

But the general idea of having computed columns like a lightweight indexed view is a good one.

My point, if I have one, is that it is perfectly possible to suggest improvements without defaming good people one has never met.

Joe Obbish

3:53 PM

That's not what was said

As someone who works in a large software company there is an important distinction here

Paul White

Nevertheless, there are real people involved, who read things on the Internet and elsewhere. A little more consideration wouldn't cost anything.

Joe Obbish

Well, as I said you should remove my comment then.