« first day (5132 days earlier)      last day (23 days later) » 

00:01
I think we all know they were made up on demand
Beyond the first twenty or so
I suppose the Ferengi had a sort of extreme form of capitalism
The Federation arrangement seems closest to Marxism in many ways
> From each according to his ability, to each according to his needs
Replicators are the key technology we need to focus on next
I suppose everything would have to cost nothing for that all to work out
00:17
That's probably the natural consequence of a post-scarcity society
Once you have a replicator, there's not much need to 'buy' anything
Which is just as well because AI and robots took all the jobs
Just no damn sythohol
Yeah, that was clearly a plot error
I wonder if I could replicate Paris
Just the best bits
Or visit the holodeck
Gonna need a bigger holodeck
00:29
I guess heaps of people would move to space and other worlds so France would be quite empty
They’d be the last to leave
Picard was quite keen on Paris as I recall
It seems your needs could be met
Well I’d naturally just go live in the 80s
Just remember the Temporal Prime Directive
Don’t bang your mom?
00:40
I think that captures the heart of it yeah
Maybe not in the most sophisticated way
What I lack in sophistication I make up for in not banging my mom
Keep up the good work then, I guess
Wouldn’t be fair to raise the bar now
01:00
Words Sean lives by
Especially if there are weights on the bar
01:11
🌶️
I hope this answered @hannah’s question anyway
Unfortunately, service broker will not be part of our brave new world
01:41
Did we get nerd sniped
01:54
@ErikReasonableRatesDarling I don't have Peyronie
I'd hereby like to call to point of order and take votes on burning down nuget and the .net GAC
 
3 hours later…
05:16
@HannahVernon it would be nice if they bothered to build an intelligent one first. Changed phone provider last week - asked the simplest questions of their so-called "AI Assistant" - no joy - nearly driven to heavy drinking! Thank God for that girl from Bengaluru - I was even going to propose... <sniffle...>
Ooops, - a chairde - Morning all!
Wordle 1,313 2/6*

⬛⬛🟩🟨🟨
🟩🟩🟩🟩🟩
Yaay... back on form after a few bad days!
@ErikReasonableRatesDarling Most French people would be quite happy if Paris were relocated to Mars!
06:06
@HannahVernon There are some hilarious names for genes - mostly in fruitflies (Drosophila melanogaster) - my favourite is Cheap Date - a fly that's very susceptible to alcohol (do I have that one???) or the Ken and Barbie gene - afflicted flies have no external genitalia! Those in the world of human genetics have been renaming them because of potential issues with patients - it wouldn't be very nice to be told that your child has a Swiss Cheese mutation... (brain full of holes...)
Some further reading - here and here.
 
2 hours later…
08:05
@PaulWhite When parts of the UI get corrupted then you are unable to start any of the new fancy tools as they seem to be located in some funny location. The old tools just work regardless of whether the UI is corrupted or not.
Wordle 1,313 4/6*

⬛⬛🟨🟨⬛
🟨⬛⬛🟨🟨
🟩🟩🟩🟩⬛
🟩🟩🟩🟩🟩
08:57
Wordle 1,313 2/6*

🟨🟨🟩⬛🟨
🟩🟩🟩🟩🟩
 
1 hour later…
10:22
Wordle 1,313 2/6*

🟨🟨🟩⬛🟨
🟩🟩🟩🟩🟩
@PaulWhite jinx
Seems like we chose the same starting word today
Maybe. I have two variations with just the 4th letter changed
I guess we'll find out tomorrow
10:48
Ah, no, our words are different
 
2 hours later…
13:09
well that was exciting
13:46
@SeanGallardy Nice! Well done 😄
14:00
Wordle 1,313 4/6*

⬛⬛⬛🟨⬛
🟨🟨⬛⬛⬛
⬛🟨🟩🟩🟨
🟩🟩🟩🟩🟩
@ErikReasonableRatesDarling I can only imagine if AI were actually intelligent how much wider the wealth gap would become. Also I didn't get notified when you @-mentioned me.
Presumably SE AI hasn't been added to chat yet
Congrats everyone on getting Wordle in two today. Apparently it was easy for all but me.
@Vérace perhaps it would be better named NAAI as in Not Actually Artificial Intelligence.
14:35
@HannahVernon The way I was feeling, more like NAAFIFFS! ;-)
Wordle 1,313 4/6*

⬜🟨⬜🟨⬜
⬜🟩⬜🟨⬜
⬜🟩🟩🟩🟩
🟩🟩🟩🟩🟩
14:59
frog waving hello
Frogs aren't that nice
15:57
I see where they get it from...
what did you do to the frogs to make them mean
16:55
@Lamak Squinting frog?
I aint do nothin'
two correct apostrophes in one sentence
the world really is changing
a' changin'?
Apologies to Bob!
why would you apologize to bob
17:13
I think it's "A-Changin''" - at least according to Mr. Google!
@HannahVernon not quite
But I got a Donald, sort of, frog.
17:58
TIL that adaptive join spills force the scan branch. That was interesting.
18:21
@Forrest you had a spill from the seek/nested loop branch?
of course i've seen adaptive joins with two seek branches so i don't want to get cornered totally by anyone (ahem)
ashktually
certain elements in chat are far more particular than others
18:41
I enter the room to see people talking about nested loop spills
What happened here
that's my current inquiry
I imagine a silly person might think the correlated input was independent of the non-correlated input
So, the join could choose either
That would be a silly person, though
Or perhaps their spilling NLJ upgrade came with a correlated hash join
"Closing 1 app and restarting ... Task Manager" long delay
What's the difficulty there Windows
do you have a lot of tasks to manage
18:48
Not just at the moment
Context: Wife has decided she wants to use her ancient Surface Book after all
There are pending Windows Updates
She had a browser open with about a dozen tabs. That was it.
install linux
i hear it revives old hardware
Sir, I am the support staff
of course she might become a hacker
it's a risk
linux doesn't need support. it's self-supporting. can run itself without interference for dozens of years.
Oh well, that sounds ideal
I suppose technically the nested loop branch of an adaptive join might have a scan
with maybe an eager spool or something
but that's not currently possible IIRC
is not
but the hash can have a seek with some supplied predicate
18:54
Maybe SQL Server 2025 will save the bush's blushes
@ErikReasonableRatesDarling really, how interesting, do you have a newsletter
@PaulWhite here's how to order
order is much sought after
Anyway, this Surface runs W10. I can confirm Task Manager is better
Notepad is objectively worse
i do sort of miss w10
it would be nice if we have a w10.5
Only natural to feel a fondness for the last ever version of Windows
it's not a great situation where you notice an OS so much
19:00
Exactly right
hmpf
i should go to the gym before morale improves too much
19:14
@ErikReasonableRatesDarling No, it started building the hash table, spilled, and then continued as a hash join despite being under the row threshold
That makes more sense
Also frustrating is how an adaptive row threshold of 1.5 is morally equivalent to a threshold of 100k. But that's a blog post for later.
20:08
@ErikReasonableRatesDarling Becoming a hacker seems like a pretty safe move. You can get a quick presidential pardon if you run into any trouble.
@Forrest nested loops probably don’t have the faculties to deal with spills, but maybe I’m thinking like an e-core.
@JoshDarnell I could use a few of those. Have a few ideas in the works.
@ErikReasonableRatesDarling Probably a limitation of the AdaptiveBufReader
The thing that looks like a table spool source if you expand the adaptive plan with TF 9415
20:26
So nested loops can read hash spills?
Just not adaptive ones?
Well, the only thing that can transition a hash join to an apply is adaptive join, so that's the only place the issue arises
where else might you encounter an apply running off a hash table?
I sort of figured that the spilling operator would be responsible for dealing with that and that’s why the limitation existed
My guess is the AdaptiveBufReader is just pretty basic
Intern left or whatever
I suppose you could argue a hash that spills is a sign that apply wouldn't be a good option
MemFree
That seems a bit unfair since hash bailout turns into nested loops
Maybe not apply nested loops
20:32
sort of yeah
Anyway, big inputs get HJ, small get apply, so you could argue anything big enough to spill means HJ is the go, regardless of what the AJ threshold says
4u
I don’t fully understand why memory grants under the max can’t adjust at runtime
(Up obviously)
In general, you mean? Because some of them do, like index building sorts
No for queries
I know some of them do, I think there’s an exception for columnstore inserts as well
Maybe behind a trace flag
20:38
Yeah there are a few batch mode things
It sounds like a good idea until you try it
Hindsight is a powerful tool in query optimization
Hence memory grant feedback
Just giving things extra memory grant as they go along doesn't work very well in practice
Kinda defeats the purpose of setting a limit in the first place
I’m not suggesting going beyond the limit
I mean the limit set by the initial memory grant, not the instance-wide limit
Limits based on estimates!
Yes
But you have these limits because you want to guarantee a decent outcome for all concurrently running queries. Semaphores and whatnot
Obviously the same semaphore rules would apply
150% of the increase being available
I wonder how highly concurrent environments are where really painful spills occur
20:48
That doesn't work though. The semaphore can split 100% of what's available among queries. Any one of those going beyond its initial limit could have catastrophic effects on all the others.
The grant is a guarantee that memory will be available if you need it.
If you're going to allow something 150% of its grant, you're really just granting it 150% in first place
But I know what you're saying about it being dynamic
Like the index building case
It all sounds very sensible until you implement it
Devil being in the details and all
I meant 150% of the increase would have to be available so it wouldn’t tank anything currently running or anything seeking to run
Not that the query should get a 150% increase
Normal resource semaphore rules
You can't see the future though
A generous and flexible semaphore could easily find itself painted into a corner
"Well, this is awkward"
It's not like they didn't explore this sort of thing before deciding on MGF
21:16
They’re not exactly forthcoming with the details of that exploration
Maybe you have some background on it
MGF can’t see the future for what the state of the server would be the next time the query runs with an increased grant
I’ve seen some whopping increases there
21:39
Maybe MGF needs more GF-1
21:57
It should dump its girlfriend?
22:13
Dump_Preemptive
22:39
@ErikReasonableRatesDarling Right, but at least you're making changes on the basis of past information, not just increasing the current grant in the hope it will be sufficient
doesn't do much for the poor query that's going to take an hour to run in all its spilling glory
sounds like you have a query tuning problem
have you updated statistics and defragmented your indexes
it's not like MIN_GRANT_PERCENT and MAX_GRANT_PERCENT don't exist
22:56
resource governor also exists
but you know
in the age of iqp qo and qe should snuggle up more

« first day (5132 days earlier)      last day (23 days later) »