« first day (4525 days earlier)      last day (344 days later) » 

12:38 AM
Adobe makes a lot of doodoo software.
 
Adoodoo
 
1:28 AM
The word that my wife and I coincidentally used as our first Wordle guess yesterday turned out to be the right answer today
 
 
4 hours later…
5:20 AM
Morning
Wordle 706 6/6*

⬛🟨⬛⬛⬛
⬛⬛🟩🟨⬛
⬛🟨🟩⬛🟨
🟨⬛🟩🟨🟩
🟩⬛🟩🟩🟩
🟩🟩🟩🟩🟩
Not my best attempt
 
5:40 AM
Morning
Wordle 706 4/6*

🟩⬛⬛⬛🟩
🟩⬛⬛⬛🟩
🟩⬛🟩🟩🟩
🟩🟩🟩🟩🟩
@PaulWhite Congrats; and what a great starting word!
 
 
2 hours later…
7:16 AM
Wordle 706 5/6*

⬛⬛⬛⬛⬛
⬛⬛🟩⬛⬛
⬛⬛🟩🟨⬛
🟩🟨🟩⬛🟩
🟩🟩🟩🟩🟩
 
Oh, I thought you reused the starting word and got it in one guess.
 
@ErikDarling So you're adding an after insert trigger to generate a history row with the modification user based on a computed column, because otherwise you can't reliably know who performed the insert?
 
 
1 hour later…
8:22 AM
@HannahVernon thanks for the pull request. Hadn't been on github in a while.
 
8:35 AM
2 days ago, by Paul White
No one knows how to use git.
 
8:48 AM
@Zikato No, we both happened to choose the answer word today for our first guesses yesterday.
Independently choosing the same non-obvious starting word yesterday was coiincidence enough
But then the same word turned out to be the solution the next day
 
 
2 hours later…
10:38 AM
user image
5
 
So true
 
 
1 hour later…
12:06 PM
@PaulWhite I think the computed column goes away (I'm assuming) in this solution. He just has a regular ModifiedBy column that gets updated in the trigger, which causes that update to get logged in the history table.
 
@J.D. Yes, that's my reading too but I wanted to be sure
 
w0rd
 
Wordle 706 5/6

⬜⬜🟩⬜🟩
🟩⬜🟩⬜🟩
🟩⬜🟩⬜🟩
🟩⬜🟩🟩🟩
🟩🟩🟩🟩🟩
forked
 
@ErikDarling Due to?
 
12:14 PM
@PaulWhite it being a bad idea
 
@ErikDarling OK but I'm kinda interested in the specifics, if you have the time & energy to explain. Did a temporal table not meet the requirements, or was a trigger considered undesirble, or...?
> Queries containing columns of the new JSON type will return the data as varchar(max) type.
🤨
 
@PaulWhite it was hard to figure out a way to track inserts reliably without either losing the computed column or doing some kind of no-op update to newly inserted rows via a trigger to get them into the history table
 
@ErikDarling Yes, a no-op update on insert is p much required
You can only capture the correct value of the computed column by copying it to the history table
 
12:35 PM
Yeah so it felt a bit easier to use CDC for a few different reasons
 
12:51 PM
@ErikDarling I can't shake my head enough at this
 
Temporal tables also make it difficult to distinguish different operations made very close together in time. If the ValidFrom and ValidTo values are the same, there's nothing to determine which came first. That's important if you want to distinguish an insert from an update, for example.
 
@SeanGallardy i'm sure it was quite low effort
@PaulWhite i've never run into that, but the no-op update thing was like one step too hacky for even me to push forward on
the real task is coming up with triggers that update the cdc table reliably. it feels easy at first glance, which means i'm going to have to get those weird cdc functions involved.
 
@ErikDarling It's difficult to demo fully on db<>fiddle because one can't create users or logins, but it's easy to see the order of operations would be non-deterministic with equal values for ValidFrom dbfiddle.uk/zFBHwaez
One can work around it by including the transaction id, but still.
 
oh that is weird
> (e.g. JOIN, WHERE, HAVING clauses)
@J.D. an index typically won't help a having clause since the result has to be computed at runtime and filtered later
it may help you get a result to filter faster, but you can't directly seek to anything
 
1:14 PM
if someone has sysadmin, can't they get access to all the necessary information to decrypt data?
 
@ErikDarling Heh was literally about to ask this.
Or at least to manage the permissions one would need to be able to?
@ErikDarling word that's fair. Never thought that through but makes sense.
@PaulWhite That's because of the precision limits of datetime2?...does CDC have any similar potential issues? (If memory serves me correctly, CDC isn't sequenced by time, rather by the LSN or some sort of GUID or something?)
 
@J.D. No, the fundamental limit is the accuracy (resolution) of the clock feeding the value. You can have 7 digits of precision and only 3 of accuracy.
19 mins ago, by Paul White
One can work around it by including the transaction id, but still.
@ErikDarling Not with Always Encrypted, no
Only the client has the necessary encryption key
 
interesting
the one time i ran into it, it played a real devil with performance
 
Yes, you can get a bit of extra perf in limited cases by using deterministic encryption, but that's less secure
Seeks for equality tests etc.
mfw people leave a comment asking for a doc link instead of editing the link in
😩
 
i wouldn't edit the doc link in either
 
1:27 PM
Wordle 706 3/6*

⬛⬛⬛🟩⬛
🟩⬛🟩🟩🟩
🟩🟩🟩🟩🟩
 
imagine if someone confused your edit with you being the person who left that shit of an answer
2
embarrassing
 
Ha ha ha
Well, it is the answer to the question
 
Clicking the click bait
 
i wonder who told him about that
 
1:31 PM
Which problem? The index range one or PSPO?
 
@ErikDarling that's kindof terrible
 
Really bad yeah
 
@PaulWhite the index one discussed in the post
do you know what the tf does?
 
@ErikDarling Yes
@ErikDarling Doesn't repro for me on first try
Changing it to a DESC index key reproduces the error, as expected
 
what does the tf help with in that scenario?
 
1:46 PM
Disables an optimization for descending indexes
 
lousy optimizations
 
So a multi-seek happens in some order, Ozar then Skeet or vice versa. Depending on the index key spec and order by clause, you either need to seek Skeet or Ozar first to avoid a sort. You can also scan (range scan) an index forward or backward
It's easy to get the logic wrong
I forget how many descending index bugs there have been over the years
CREATE INDEX DisplayName ON dbo.Users(DisplayName DESC);

-- Not ordered by DisplayName ASC
SELECT *
FROM dbo.Users
WHERE DisplayName IN (N'Brent Ozar', N'Jon Skeet')
ORDER BY DisplayName ASC;
 
would have been easier to just build parallel backwards scans eh
 
It's funny because if you specify ORDER BY ASC you get DESC and vice versa
Off by one I guess
 
@PaulWhite so I tried scanning these docs and a few related ones this morning, but didn't readily find anything. But if sysadmin doesn't have access to manage the permissions for managing the encryption, how does one acquire said permissions?
@PaulWhite Also, you really hate comments eh? I'm with Erik. Rather teach a person to fish then do the work for them.
 
2:00 PM
Wordle 706 3/6

⬜⬜⬜🟩🟩
⬜⬜⬜🟩🟩
🟩🟩🟩🟩🟩
 
I know I said "or" but that was more of a hint that both the docs link and quoted information from the docs would make the answer better.
 
@J.D. Yes, but you're always wrong. They'll be notified of the edit.
@J.D. One doesn't, inside the database engine. Only the client can decrypt.
 
@PaulWhite Thought I briefly saw the docs mention, the initial implementation can be done either in T-SQL or PowerShell. But perhaps I misread. I am always wrong ya'know.
 
> To ensure Always Encrypted keys and protected sensitive data are never revealed in plaintext to the database environment, the Database Engine can't be involved in key provisioning and data encryption, or decryption operations. Therefore, Transact-SQL (T-SQL) doesn't support key provisioning or cryptographic operations.
> For the same reason, encrypting existing data or re-encrypting it (with a different encryption type or a column encryption key) needs to be performed outside of the database (SQL tools can automate that).
 
what's the last version of SSMS that had database diagrams? 17?
 
2:08 PM
I thought they were back?
 
They're definitely in 18.something.
 
so did i but i'm not seeing them in 18 or 19
 
Using v18.12.1 on my end with them.
 
i thought they used to be right under the main database node in object explorer
 
It should be under each user database node itself.
 
2:10 PM
ooooooh, user DBs only
was that always the case?
 
Not sure, but makes sense to me I guess.
 
i feel like i made a DB diagram of msdb one time
 
Brave of you...
 
couldn't find it this morning so i went back in to make a new one and... now i'm here
 
2:12 PM
but why dot gif
 
No idea. Maybe they don't support the objects needed for db diagrams in system dbs
 
@PaulWhite Interesting. But you can view the encryption key via T-SQL?...for what purpose?
 
> You need to store column master keys in a trusted key store outside of the database system, such as Azure Key Vault, Windows certificate store, or a hardware security module.
@J.D. The column encryption key might be involved in key rotation (I forget the details) but you can't do anything with it without the column master key.
 
Nvm, I understand what the docs are saying now. Needed more time for my brain to wake up. Thanks.
 
@J.D. Feel free to leave a better comment or make a better edit
 
2:17 PM
@PaulWhite I think the original comment was better than the answer :)
 
You were out-voted on that, I'm afraid
@ErikDarling Specifically, the trace flag disables an optimization which skips a step of the logic when it is safe to do (ha ha).
I find it hard to believe they ran any tests.
Any that passed, anyway
DECLARE @T table (i integer NOT NULL INDEX ii (i DESC));
INSERT @T (i) VALUES (1), (2), (3);

-- Ordered descending!
SELECT Ascending = T.i
FROM @T AS T
WHERE T.i IN (1, 2, 3)
ORDER BY T.i ASC; -- We asked for ascending

-- Ordered ascending!
SELECT Descending = T.i
FROM @T AS T
WHERE T.i IN (1, 2, 3)
ORDER BY T.i DESC; -- We asked for descending
🤣
 
2:32 PM
@PaulWhite Interesting. I couldn't repro on my end.
 
@J.D. Using Brent's example? Not surprised. He's fixed it now.
You also need to be running SQL Server 2022 CU4
 
@PaulWhite ah, that may be the difference then. Nvm.
P funny bug though.
 
Apr 9 at 18:21, by Paul White
Reading comprehension being what it is these days
I would have provided a db<>fiddle but that's still on RTM
 
True. I missed the part where this related to a BrentO article, so haven't read said article yet.
My brain's on holiday mode today anyway. Food trucks coming to the office for lunch, and then off to the extended weekend.
 
Computers aren't really worth thinking about anyway
Ha ha Hell Pizza have a promo going called After Life Pay. You write the cost of your pizza order into your Will, and pay when you die. Awesome. 666 lucky applicants will be accepted into the offer
 
2:50 PM
@JohnK.N. my pleasure
 
@PeterVandivier nice domain name
@PaulWhite is it available to u.s. residents?
 
@ErikDarling Sadly not
 
3:13 PM
i suppose it would be a difficult delivery anyway
though i wouldn't have to pay for the plane tickets until after i die
 
> 6) Only available to Applicants in New Zealand.
Of course they'll just forgive the debt so it's basically a free double pizza promo, but I like the idea
 
@ErikDarling really need to use enclaves
 
3:32 PM
@SeanGallardy i'd rather put it in an autoclave
 
3:45 PM
@SeanGallardy The restrictions make enclaves all but unworkable
 
@PaulWhite 2019 and 2022 are less restrictive
 
@SeanGallardy If you're happy with VBS
But I mean the list of caveats and restrictions is almost endless
 
Is it really the S?
 
Only within the VM, it says
Imagine going to all that trouble
> VBS enclaves help protect your data from attacks inside the VM. However, they don't provide any protection from attacks using privileged system accounts originating from the host.
 
4:04 PM
@HannahVernon I think it's no visible on the NYT version but the original version, it was all in JavaScript behind and it was two word lists - valid hidden words and valid guesses. In my analysis implementation, I just have the one dictionary table, with a flag marking whether it's allowed to be an answer or not.
 
ahhh
cheers
 
It doesn't work 100% since the NYT takeover. I had 3 or 4 words from the non-answer list as the answer
 
Probably ones I did by hand and never ran through my analyzer which would not have had that as a possible answer remaining. I play Squaredle which is kind of like Boggle and a lot of words in that each day don't count as regular words but only bonus words, and it's really annoying.
Today, canard was a bonus word, as was napper - I some of them seem common enough they should be the regular words
 
4:22 PM
ok, it was only 2 words

2023-03-27 - guano
2023-04-09 - snafu
 
4:37 PM
Lol, I remember those, and yeah, I didn't run the analyzer, but I remember being surprised by both
But would have been clearly obvious to me
 
Has BINGO already been the word of the day?
 
Same with SNAFU - 4 letters in both cases with 2 direct hits - hard not to get those manually.
ANd I know all of you are thinking...
Is C*NTY a valid probe word?
2
Yes, yes it is
 
5:14 PM
@PaulWhite Can I do the same with Erik's invoices?
 
@Zikato I can't see why you'd need to - his rates are so very reasonable
 
5:34 PM
Unlike the Hell Pizza?
 
Erik obviously doesn't need to promote himself as desperately as those pizza from hell guys
 
@Zikato Where you at?
FYI this is the new puzzles interface I was telling you and @PaulWhite about
 
Looks like a deep fake
 
deep dish fake
 
@SeanGallardy wth? How did you get there?
 
5:49 PM
@Zikato That's what puzzles is now, this is why I asked if you all noticed it was WAY different
 
Is that on worldwide or just Florida?
 
ystery?
So you all don't see this view?
 
no, is that the main menu?
 
@SeanGallardy what is this?
@Zikato isn't all of america gonna be florida?
 
@mustaccio I’ll give you a coupon for that
 
5:53 PM
@Zikato When I click puzzles it takes me to a new updated puzzle screen, an when I press on the number at the top it shows me this "view" and anytime I got to the next "number" it does this view with an animation of me moving and showing my "friends" and where they are. You're the only one that shows up right now.
@Lamak We can only hope
@Lamak This is the chess.com app
 
@SeanGallardy oooh
 
I only get this screen
 
@Zikato So that's what mine USED to be and then the app updated one day and it wasn't
May 12 at 13:04, by Sean Gallardy
Chess.com changed their puzzles :/
Maybe do to EU regulations and oversight it won't get to you for 5 more years
 
seems weird that the new one compares streaks instead of rating
 
yeah I don't know what my rating is now
but the puzzles "feel" easier
 
5:56 PM
it's not the 837?
 
no let me see if I can get a diff screenshot
 
you have 1770
 
yeah
it's so hard to find on this app, ugh
 
6:17 PM
Looks like Sean's accidentally playing the puzzles vs another person variant?
With the other person (@Zikato) not taking part
That's my guess anyway
 
I don't get long streaks a lot at 2400+
 
I had two losses today out of three attempts
Down to 2066
5 points for a win, 10+ for a loss really makes it hard to progress
Without spending forever on each puzzle, perhaps
@Zikato Oh I see it's streak length
 
7:02 PM
@Zikato is very streaky
 
7:37 PM
Like bacon
 
7:50 PM
Making some Friday Friends: twitter.com/erikdarlingdata/status/…
 
Curious, I noticed this once before. Sometimes my single Close Vote is enough to close a question, sometimes not. Why is that? Example: dba.stackexchange.com/q/327588/150011
 
It’s usually tied to your score with a tag the question is posted in
 
8:07 PM
var walksWithADrink is translated by the EF into SELECT * FROM [runsWithScissors]
2
@ErikDarling to the golden badge in the corresponding tag
 
8:26 PM
@mustaccio thank you, trusted user
 
 
1 hour later…
9:27 PM
@ErikDarling oOo interesting. So if you have accumulated enough of a score for a certain tag, you can close questions tagged with that tag?...seems like people could easily abuse by adding the wrong tags to a question prior to closing it?
 
@ErikDarling Was this satirical or was this a legit statement. Hard to tell from people who drink the koolaid like a greek frat on a long weekend.
> EF Core 365 AI Enterprise
I'm fairly certain it was, but hard to tell
Because whomever names products at MS should have been fired years ago
 
 
2 hours later…
11:04 PM
@J.D. They anticipated and closed that particular loophole.
If you edit the tags on a question, your close vote just counts normally.
 
11:57 PM
@JoshDarnell ah good to know. Cheers.
 

« first day (4525 days earlier)      last day (344 days later) »