« first day (482 days earlier)      last day (207 days later) » 
02:00 - 17:0017:00 - 00:00

2:05 AM
I've seen, like, one index intersection plan ever.
 
2:27 AM
did you master it?
 
2:46 AM
You guessed it.
 
if anyone wants easy rep:
0
Q: Window Functions: What is the purpose of Rows Unbounded Preceeding?

J.D.What is the purpose of using the Rows Unbounded Preceeding clause in a window function? I think I understand that it's basically saying not to limit how far back to look when rolling up for the aggregate function, but how is that different than not using the clause at all? Could you please provi...

 
That's my sock, so it would be unethical.
 
 
3 hours later…
6:11 AM
You should totally post this as an answer so I can accept it, because that literally explained everything I needed to know in 60 seconds. :) Also, thanks for the info on the window spool operator! — J.D. 3 hours ago
New user confuses comment box with answer box
Somewhat ironically, Josh is the best and most helpful main-site user.
 
6:50 AM
@JoshDarnell My favourite reference for ramp-up reads is blogs.msdn.microsoft.com/ialonso/2011/12/09/…
 
 
4 hours later…
10:22 AM
@PaulWhite explain this rating!
 
11:06 AM
@ErikDarling Least noise generated, most helpful flags and post edits.
 
11:28 AM
Sounds like moderator material
 
12:01 PM
@PaulWhite do you have time to review a short blog post?
 
12:17 PM
@ErikDarling Sure. I just finished a draft of my personal income tax so I could do with something non-numeric to read.
@ErikDarling Everyone's a moderator on SE
 
12:42 PM
i thought you took a vow of poverty
 
Well I didn't say the numbers were very big
 
maybe if you included wine gifts ;)
 
sh!
 
heh heh heh
 
if the tax man comes knocking will Paul delete the transcript?
 
12:55 PM
i hear jacinda herself shows up looking cross at you
 
1:09 PM
50% of my income last year was interest and dividends - I guess I am retired after all!
 
truly living the life
 
1:22 PM
in this answer, i assume part of the window spool magic is the presence of a sequence project in the second plan
3
Q: Window Functions: What is the purpose of Rows Unbounded Preceeding?

J.D.What is the purpose of using the Rows Unbounded Preceeding clause in a window function? I think I understand that it's basically saying not to limit how far back to look when rolling up for the aggregate function, but how is that different than not using the clause at all? Could you please provi...

i assume there's something smart to say about it
 
@ErikDarling also mentioned in this article by Aaron: sqlperformance.com/2012/07/t-sql-queries/running-totals
I suppose you'll add a link to your blog post as well?
 
i'm no longer able to edit those posts
:P
 
But you can edit your post here, so I don't follow ...
 
so can you
aaron doesn't say anything more insightful than i do
 
No, I suppose you both say the same thing
 
1:30 PM
If you liked it you should have run it in batch mode
 
i will add bitch mode
 
Window Agg FTW
The Sequence Project is for a row_number, which is only needed if the window frame is being measured in rows
 
so that is not the source of magic
is there any way to tell if you have a disk or memory spool by looking at the plan?
 
um logical reads for the worktable perhaps?
 
nothing in the xml?
 
1:34 PM
no
 
nuts
 
there was a connect thing for it, don't know if it made the transition
 
how weird, I got a spam email for this
I'm almost certain that this guy is taking posts from all of you
well...not all of you, just the famous ones ;)
 
lol
what was the email?
 
I deleted it, but it was something along the lines of "Get the latest news about everything SQL here"
and I wonder how they got my email
 
1:44 PM
same
 
I didn't get one
you guys need to surf safe
I bet Lamak signed up for free powerpoint templates
 
I didn't get one either
Seems like an aggregator of SQL Server Central content? I didn't click around much
 
there are posts from redgate an other sites
 
@PaulWhite So cool, thanks! I didn't pick up on this being an Enterprise Edition feature from the other article.
Also thanks for reading my blog. I recently realized that no one I work with reads it and it bummed me out.
Maybe they stopped being able to understand the content once I mastered index, query, and server tuning.
 
You are quite new to the blogging world so I wouldn't worry about it
 
1:53 PM
blog for yourself, not for other people
 
@Lamak Whole posts or just extracts?
 
readers = comments
you don't want that
 
@JoeObbish I don't understand this.
Isn't the whole point of blogging to share knowledge with other people?
 
@PaulWhite didn't fiddle enough...but there seems to be whole posts...
 
Well that's not really on is it
 
1:58 PM
@ErikDarling True. Scary stuff.
 
@PaulWhite was that comment for me?, if so, sorry, don't really understand it
 
@JoshDarnell That's part of it for sure, but writing about something is good to (a) make sure you understand it yourself; and (b) don't have to work it all out from scratch next time you need the info yourself
@Lamak Yes: stealing whole posts is naughty "not really on"
 
@PaulWhite ah, I see...wasn't familiar with that expression
ah, Brent already brought attention to this….does that mean I'm just reposting him?, damn
 
@PaulWhite That makes sense. Blogging has certainly helped me in both of those areas.
 
It is also nice to get to a million page views, no doubt about that
I imagine, anyway
 
2:02 PM
Right.
My gripe about my coworkers not reading was not about views, by the way. We just have worked together for 5 years, and are all somewhat close. And I know they follow several blogs.
So it was more of personal whining than anything.
 
Face it they just pretend to like you
I see all known problems with Query Store have been fixed in CTP 3 so Erin and Erik can be friends again
 
That's likely. I'm very intimidating, so they certainly wouldn't want to let me know they don't like me.
 
Right!
 
I think the new QDS changes are really cool. It will be interesting to see if it helps with adoption.
 
Seems to involve Postgres so what do you expect
Oh it's Oracle LOL
> In conclusion, not only does SQL Server perform much worse than Oracle, but also it doesn’t scale.
Seems fair
 
2:10 PM
Oh jeez, can I really not edit my own Tweets?
 
@JoshDarnell hint: it won't
 
I can continue to tell my developers that SQL Server performs poorly with superfluous WHILE loops. Cool.
 
while loop == superfluous
 
Read it. Said "meh". Closed tab.
@JoshDarnell No just delete then try again
> After I posted thus
Josh developing a kiwi accent
 
lol
I guess I'll just leave it like that.
No sense in double donking people.
@ErikDarling So negative!
 
2:14 PM
No one on Twitter can spell or read anyway
@ErikDarling It might if they went back in time and implemented it that way originally
A mistake MS seem to make a lot
Yeah yeah we know this has a reputation for being dumb, but look we fixed it up mostly in the fourth release!
I can only imagine how easy and convenient to use XE will be around 2025
 
i didn't see: did they give you the ability to put it in a different filegroup?
or is it still stuck in primary?
 
@JoshDarnell This was the first thing I read when I got in chat and I'm not going to attempt to find the context. It's funnier this way..
 
@Zane Glad your day is off to a good start =P
 
Last 2 days at xcel
 
2:23 PM
that's a long way to go to say no
 
Is xcel responsible for spreading sheet?
@ErikDarling Does QS work with readable secondaries?
 
ask someone with an ag
:D
i would imagine so
 
@PaulWhite It doesn't capture any queries on the secondary
but you can look at the QS reports there
 
@Forrest Thanks. Seems like a bit of a gap.
 
@PaulWhite it is
 
2:29 PM
Well they should fix that
Sean!!!
 
it's tough since we run so much off our secondaries and it's a bit :shrug:
 
imagine if all DMVs across AGs were reliable
 
Right exactly
 
plan cache, index use
makes index tuning really hard
 
just add another hack like temp stats
 
2:30 PM
stuff that looks unused on primary is heavily used on secondary
 
seems like it all ought to work with Log Shipping
or maybe not
idk
@ErikDarling no
hth
 
🙄
i'm gonna log ship you to tasmania
 
how rude
just the sort of thing I'd expect from a Java expert
 
Watch out for that furry tornado monster thing.
That's the only thing I know about Tasmania.
 
v helpful
 
2:42 PM
@PaulWhite that's a really weird post
 
yes
 
I'm going to try it
 
Another odd thing from the same person.
 
Can you believe they disabled one-boxing for tweets
 
I keep forgetting that's broken.
Oh I thought it was a bug.
 
2:45 PM
@PaulWhite good
 
Oh so that's how muting works on Twitter
 
Thanks, Taryn.
 
I can't believe they did dev to improve the chat functionality
 
@PaulWhite lol
 
@JoshDarnell don't blame me
 
2:47 PM
Manual one-boxing for Joe
@JoshDarnell Yeah probably
 
another stunning #sqlhelp success
 
#sqlhelpthathelps
 
@Taryn You are our local proxy for SE platform issues =)
Just like Sean is at fault for all SQL Server problems.
 
lol
 
huh
he's right about the scalability of that code
it does horribly at 100 threads
 
3:01 PM
sounds like an OLTP problem
 
redundant
 
also like a meh problem
 
OLTProblem
 
How does it do at 100 fib[er|re]s?
 
I got 4850799 work units done over one minute with 1 session
I got 4025837 work units done over one minute with 100 sessions
negative scaling confirmed
 
3:03 PM
49 mins ago, by Josh Darnell
@ErikDarling So negative!
 
watched captain marvel this weekend
 
oh yes?
 
i wonder at what point marvel writers stopped caring
 
ah
anyone seen x-men dark phoenix yet?
 
I couldn't type "inb4 Erik says he hated it" fast enough.
 
3:05 PM
@PaulWhite no but looking forward to it
 
I have high hopes for that one
 
@JoshDarnell i didn't hate it; it was just another very dull vehicle for pandering
 
Oh... Dark Phoenix is going to be baaaaaad.
 
WHY ZANE
 
@ErikDarling Was it woke?
 
3:06 PM
@PaulWhite Do you have a theory for what's going on with the while loop?
 
@JoeObbish I stopped reading it and genuinely don't care
 
Because they don't know what they are doing and have not for a single moment understood what makes X-Men special. It's also been a production dumpster fire and Dark Phoenix requires avengers level set up.
 
Oh I thought you meant bad as in good
 
@PaulWhite It's based on the Kelly Sue Deconick run so I would imagine it's very woke.
 
Scared to Google that
 
3:08 PM
@ErikDarling Ah, my mistake then.
 
inb4 Josh apologises
 
@PaulWhite oh
 
damn
 
all of you owe me an apology and $11
 
or you'll do what?
 
3:09 PM
SUE
 
I thought it was a fun take on the Kree vs Skrull story.
 
you were supposed to say SUE
 
i thought we were done with that bit
my bad
 
@PaulWhite She's a good writer and I actually love her ultra feminist book(bitch planet) because of how ludicrously over the top it is.
 
it's not been flogged to death yet
@JoshDarnell Is your keyboard stuck?
 
3:10 PM
It's important to drive a bit all the way into the ground before giving up on it.
 
yes
 
does anyone care about captain marvel spoilers?
 
not much
unless they relate to End Game
 
@ErikDarling I don't
 
@PaulWhite I'm in a different office today, using someone's impossible ergonomic keyboard.
 
3:12 PM
might go to see End Game tomorrow afternoon
 
won't you run out of time soon?
 
@JoshDarnell condolences
@JoeObbish me? to see it? I guess. Seems to be showing at least into next week here
 
@PaulWhite they don't
 
go nuts then
 
i felt like the movie lost its girl power cred when the bad guy was A WOMAN
 
3:13 PM
had they transitioned tho?
 
not that i'm aware of
 
good lord
 
HELP
 
use the onscreen keyboard
 
3:14 PM
the bad person also looked a lot like a hillary/warren mashup
 
buffett?
 
elizabeth
 
I never understood what the phrase "wooden acting" meant until recently
that is a phrase, right?
 
yep yep
 
3:15 PM
sure is
 
@JoeObbish which film helped you with that?
 
@PaulWhite it was a person
a certain captain
 
Surely not Picard
 
definitely not
 
He has a new film in the works which has me worried
 
3:16 PM
 
rigid
hey perhaps this all means DC will start making good movies
 
I figure I can just stare at that image for 2 hours and get the full experience of watching captain marvel
 
jeepers
I have to say that is all v disappointing
but ultimately a time saver for me
 
I didn't watch the movie to be fair
 
Has anyone in here used a tool for reading XE files other than writing XQuery?
 
3:26 PM
I send them to Erin
 
Wait isn't that the wrong SQL Skills person?
 
Probably
On the odd occasion I run an XE, I use SSMS to read the file
 
Cool, thanks. The only thing I really don't like about SSMS is that it doesn't show resolved callstacks.
 
well that's sucky
 
It is. If I query the data with XQuery, it shows up as resolved. For some reason, viewing the exact same thing in the SSMS file target UI shows addresses only.
 
3:32 PM
1 hour ago, by Paul White
Sean!!!
 
@JoshDarnell I think there's some work around
ask @Forrest
also XE callstacks are a pain
 
Because of the formatting, or something else?
 
sometimes you don't get what you expect
Paul explained it once here
 
is there a hint or tf to get rid of the distinct sort in this plan?
nevermind, hash group
 
are we boycotting pastetheplan?
 
3:47 PM
i just don't want to cost brent any money
 
@ErikDarling it'll still spill
 
yep
hash spill is worse than sort spill, time wise
that might make a good post, who knows
 
no doubt
I would probably use the UNION ALL rewrite with FORCESEEK hints
Push the MAX down to the APPLY as well
 
yeah, for sure
 
gather streams now runs in negative time
 
3:55 PM
i've been seeing that too
 
what am I looking for there?
 
table insert is 10.2 s, gather streams is 8.125 s
time should be the sum of the operator and its children
 
time decreases at gather streams from the prior operators
 
@ErikDarling and the compute scalars
clearly using alternative mathematics
 
@PaulWhite churly
 
3:58 PM
@PaulWhite those are batch mode
 
that's nice
I have a special bowl of salt next to my keyboard for showplan metrics
 
nothing is reliable
showplan, set statistics, XE
what are we supposed to use? profiler?
 
clearly
 
seems safest
meanwhile, celebrate the rapid releases due to devops
 
good times
 
4:02 PM
the stats! they're alive!
 
sp_spacehelp
that's a new one
 
just use ultra lightweight profiling where all the numbers are zero
 
@PaulWhite will that finally give us an ACTUAL plan for all queries?
 
better
an ACKCHUAL plan
 
this is hysterical
 
4:04 PM
does it have batch mode spills?
 
so the times in query plans doen't include returning results
xe does
 
> returning results
 
Serves you right for ignoring hygiene
 
What if we add columns for tempdb spills to sys.dm_exec_query_stats that don't include batch mode spills and we don't tell anyone?
 
What sort of lame query returns results anyway
 
4:05 PM
returning results is slow, do it on the client tier
why are you using your $7k licensed cores to return results?
 
heh
is that you Brent
 
data is a liability
no one should have any
 
does it bring you joy
 
I know of some people who use $7k licensed cores to run C# code
can you believe that?
 
sounds like a scaling problem
 
4:08 PM
I hear sql server doesn't scale
have you checked out WHILE loops lately?
 
T-SQL anyway
T-SQL sucked when my grandad was young
btw @Joe did you compare safe and unsafe spook?
 
@PaulWhite Kind of, it depends on what you're looking for. All of the unsafe spooky code that I ran used the columns as inputs for the CLR
I never did a VARBINARY(8000) input that ran unsafe spooky hash
 
@JoeObbish sort of a broad-brush x is y% slower or faster than z in the same test
if you don't have that info, it's fine
I was just curious for a moment
 
~for a moment~
 
I feel like I'm letting Paul down
to use your code I think I'd have to convert the byte[] to a stackalloc or whatever it was called
 
4:12 PM
stackalloc requires unsafe
no it's fine really I can test it myself at some point
 
the best result I have in that area is 330 M hashes for 100% safe code vs 895 M hashes for unsafe code
 
prolly about right
 
the issue is that the safe code passes in BINARY(8000) and the unsafe code doesn't
 
hm yeah ok hey don't worry about it
 
the two paths that I took were "We could reasonably do this" vs "What's the best we can do if there's no rules or restrictions?"
 
4:18 PM
330M is still a big number I think
compared with, say, 4
 
or with
4 u
 
good point
 
4:44 PM
so uh
how is everyone
 
02:00 - 17:0017:00 - 00:00

« first day (482 days earlier)      last day (207 days later) »