TIL that MySQL and MariaDB allow you to specify MATCH PARTIAL or MATCH FULL on your FKs, but completely ignore them, and also causes them to ignore the cascade clauses.

Exciting news! Shrink is now available in Hyperscale! Cloud first!!11!

Now you can avoid paying for storage you're no longer using by paying for CPU and I/O to shrink

A chairde -Morning all!

@HannahVernon It's a good one here for us shiftless Irish - long, thanks to St. Brigid! of Kildare!

Mirror image (sort of)

@PaulWhite It's called hyperscale, not hyper-cost-savings :P

@PaulWhite Have you been messing with my chat post? I specifically linked the repost of the repost of the repost, was there a problem with that?

So aggressive. Do u have a problem with that m8? No, I thought you did it by accident.

Oh sorry. I was trying to show how many different times I've found weird stuff in MySQL...

It is fixed now

thanks

@SeanGallardy Indeed. I particularly liked the advice that the process might run for several days, so you should issue it from a connection that's not likely to get disconnected (from the cloud!)

@PaulWhite I thought when I read it "OK MySQL allows you to use syntax for features which aren't supported and just ignores it, that's their strange philosophy" but then I saw "and also causes ON DELETE and ON UPDATE clauses to be ignored. For these reasons, specifying MATCH should be avoided." and I just fell about laughing.

Another gem:

> If ON UPDATE CASCADE or ON UPDATE SET NULL recurses to update the same table it has previously updated during the same cascade, it acts like RESTRICT.

So it just ignores your cascade with no warning. SQL Server famously throws an error when you try this.

@ErikReasonableRatesDarling Do you have a demo anywhere of sp_prepexec not sniffing parameters? Counterexample: dbfiddle.uk/s816PmrB

@PaulWhite Yep, and then the complaints about the log throughput and all the other things while it's running. Hyperscale still needs a ton of work IMO but apparently people are using it.

I think it has a reputation for being the least expensive option

Just sign here for the next 3-5 years and you're locked in at a low rate

> The LINQ expression 'giant lambda string here' could not be translated. Either rewrite the query in a form that can be translated, or switch to client evaluation explicitly by inserting a call to 'AsEnumerable', 'AsAsyncEnumerable', 'ToList', or 'ToListAsync'.

Well how about you tell me why you're refusing to translate my query?

Do I really need to pull up an EF Core debugger, and go down into 25 nested classes to find out which operator I used in a slightly unexpected way that made you go beserk?

sigh...

I'm sure lambdas made programming easier in some fashion, but they sure make debugging a chore

Especially without source

Well the issue isn't lambdas per se, it's that LINQ/EF uses expressions of lambdas. So you type what looks like actual code, it then goes and turns that into an expression tree (a bit like a query plan) and then transmogrifies it so it's almost unrecognizable.

And that it seemingly can't throw proper error messages when it hits an expression it doesn't know what to do with. It just pushes it back up the stack and says the whole query can't be translated, without saying what exactly it was.

Yes there's heaps of scaffolding. The layers and generated names make debugging frustrating

@PaulWhite I do wish lambdas would cease to exist

@SeanGallardy what's the prob with lambdas?

EF Core being such a leaky abstraction makes me think of a bucket with not just a pinhole, but the whole bottom of the bucket gone.

@Charlieface Much as Paul mentioned it's syntactical sugar which makes debugging 10000% harder

I need clippy to pop up and say, "I see you're too lazy to make another method or function, would you like me to make your code messy with a lambda?"

Only because Visual Studio gets itself in knots when you do Hot Reload on a lambda, it doesn't have to be that way.

As someone who looks at a lot of memory dumps, it has nothing to do with VS

@Charlieface I wonder if there's an EF Core feature request for being more specific about the failed translation. I can't find one. I thought I had seen it be more specific in certain scenarios, but maybe I'm crazy.

@SeanGallardy Meaning you can't correlate the mangled function names to the lambdas? Does it not work if you load up the PDB? can't say I've tried myself.

@JoshDarnell yes, does that already in some situations, but having gone through the code it seems to me that's making it worse. Because it tries to log the error without throwing immediately, so that it can try multiple ways to translate, then it just bails out and says "no can do". Would have been better to just immediately throw an exception in a lot of cases.

The code itself is gnarly in the extreme: think FizzBuzzEnterprise on stilts.

You can get the location from the symbols and most modern debuggers do a good job at getting you around the right spot. Tell me if you saw a callstack like this if you'd think this was at all helpful, and you would know what the code was doing and what was being attempted.

SomeRes!CompleteAction::__2::<lambda>
SomeRes!ProcessItem::__22::<lambda>
SomeRes!InvokeStubWithExceptionPolicyAndTracing::__l6::<lambda>::operator()

Dunno, in VS you can just double click the stack frame and it'll take you to the exact line. What are using, is it windbg?

Yes, VS is woefully terrible at memory dumps from random systems with varying degrees of symbol availability

You're assuming I'm the one with code access and writing the code, whereas I'm getting the dumps from a system I don't have access with to code I also may not have access

yeah that might be an issue...

currently managed to get an ExecutionEngineException in VS and had to shut it down, after trying to Hot Reload in a function with a complex LINQ expression. Twice.

VS couldn't even stop the debuggee process, had to kill IIS.

@SeanGallardy what do you do with a dump without code? It's not like you can write the fix, how does that work?

See, have you tried turning it off and on worked again!

@Charlieface Tell Crowdstrike their user-mode level detours are crashing the process and it's not a product bug

3rd party drivers are also a good mix - mostly Oracle

bleuggh

At least you're not the poor sysadmin who has to physically climb two stories up the side of a building to plug a USB stick into an Intel NUC so that it can boot up and wipe the offending driver.

not me either, but heard stories.

My first line of defence is turn off, turn on. Second line is have you checked the cables. That fixes 95% of issues IME

@PaulWhite i think the closest i have is something with sp_prepare

the most surprising thing in that post is that pspo kicked in for it

@Charlieface Yeah I read that Stack Exchange blog on moving to the cloud as well :P

huh?

Joking about how in their blog it was beyond difficult for someone to get in a car/train/tax/whatever and go to their rented datacenter space to work on the servers.

friend told me about this one, I think it was some kind of display billboard thing. There were loads of these sorts of devices buried in funny places.

it is outrageous that with 17 comments (including one from our dear leader) no one has asked to see the restore command being run

shame on all involved

lots of people getting bogged down in permissions for the .bak not having read the error message.

> The operating system returned the error '5(Access is denied.)' while attempting 'RestoreContainer::ValidateTargetForCreation' on 'C:\Program Files\Microsoft SQL Server\MSSQL15.SQLEXPRESS\MSSQL\DATA\<Prod>_<Customer>.mdf'

Quite clearly it's the target not the source that's the issue.

computers hard

not as hard as reading comprehension, clearly

clearly

@Charlieface that was my point! not a very long walk to see that one!

Maybe someone should comment on that answer :P

better dead than commented

@ErikReasonableRatesDarling Thanks. Yeah, there's also the old one on the other site with the JDBC settings workaround but I was specifically interested in prepexec because I could have sworn that had the same issue but now I can't reproduce it even as far back as 2008 R2

@ErikReasonableRatesDarling There you go.

@ErikReasonableRatesDarling I almost deleted the answer but left a comment to remind myself to check on it later

@PaulWhite hm. For what it’s worth I had tests for that in my old site demo script (not part of the post) noting that it didn’t have that problem.

I wonder if it changed around the recompile thing in 2008 sp1(?)

ysterious

To help AIs understand the world, researchers put them in a robot

@PaulWhite v thick plot on this one

Isn't it

I deleted that answer. Our resident Comment Crew will need to find a new place to fail to use the site correctly

I upvoted John's answer

Larry Ellison is the Trade Offer guy isn't he

I get a Hawaiian island, you get an invoice

Pretty much