Alexander Drogin

Another edge case of an overflowing record in a SQL table As a follow-up to my previous post, I'd like to linger a little longer on the topic of wide rows and the effects of row overflow in SQL Server, and specifically on the storage impact that we can obser...

SQL Server allows table rows to exceed the 8KB data page limit. But... some curious side effects can happen.

www.keytogoodcode.com/post/another...

Performance benefits of SetLoadFields in a slow network Introduction of the SetLoadFields function in AL language that allowed to limit the number of fields returned to the Business Central server by a SQL query was a big step forward towards achieving bet...

SetLoadFields function helps developers to avoid unnecessary table joins and reduce the amount of data sent over the network. But how significant is the benefit of a smaller dataset? When does it really make a difference? Some measurements in a blog post.

www.keytogoodcode.com/post/perform...

It looks like I need to practice my skills in using the Ctrl+F function. I once asked how I could generate a hash with the Cryptography Management codeunit based on a binary key. Easy, it turns out. There is a function for this that accepts a base64 key. I just did not find it.

JIT loads vs Full record load Since the time when I came across this post by Waldo (or actually, he mentioned his test results in one online discussion) I was curious to try it myself and dig into its unexpected results. The quest...

How can just in time loads be faster than a full record load? After all, this means two queries reading the same table. And two are slower than one, right? Right, but only if we allow these two queries to run until completion.

www.keytogoodcode.com/post/jit-loa...

To sort, or not to sort, that is the question Back in the old Navision days, when clients' data were stored in the Navision native database instead of Azure SQL Server, and functions like FindSet and FindFirst did not exist in the AL language, ev...

Which side are you on when it comes to the question whether SetCurrentKey is beneficial for data access performance in AL? Should we use it or drop it?

I give a definite and unequivocal answer: it depends.

www.keytogoodcode.com/post/to-sort...

If the other side was using a text key, this would make things easier for us. The key is actually a hex number shared as text, but they parse it into its actual hexadecimal value before running the hash function.

This could be another option for a contribution to the system app. The function that generates the hash must decode the Base64 string in this case. But as it is, the function would take the encoded string as the key.

Yes, that's what I was considering too. Just thinking that the developer experience would be a bit different from classic arrays, as we cannot allocate a specific amount of memory, but have to add one element at a time.

It could be on overload. Although AL is generally not too good in manipulating byte arrays. Would it even be a good idea to to have a function accepting a list of bytes, considering that AL arrays can only have fixed size?

I can't see any way around it, but maybe someone has an idea how to generate a hash in AL with a byte sequence as a key instead of a text string?

I have a good question for #bcalhelp. I need to generate a hash with a function from Cryptography Management CU, which requires a text argument as the key. But I need to send a byte array. Text in AL is UTF-8, so bytes above 7F turn into 2-byte sequences, and my key never matches the counterpart.

Garbage collection of event subscribers I once wrote a post on event subscribers and the presumed performance benefits of single instance subscriber codeunits, but the topic keeps reappearing every now and then, and I feel it could be a good idea to add a few more arguments to the same old debate. Because why not run a few useless, but highly entertaining tests on a Saturday evening.One statement in defense of the single instance event subscribers is that the subscriber codeunit does not have to be disposed after the execution and rem

@nataliekarolak.bsky.social has been concerned about the lack of activity on BSky. I have a suggestion. How about garbage collection? Seems to be a good topic to talk about.

www.keytogoodcode.com/post/garbage...

GitHub - adrogin/BCVisuals: Visual control add-ins for Business Central Visual control add-ins for Business Central. Contribute to adrogin/BCVisuals development by creating an account on GitHub.

I worked a little more on the BC cost tracer and added a couple new features: node grouping and a new layout with the fCoSE algorithm, which is very efficient in compound nodes clustering. Thanks to this algorithm, item ledger entries can be grouped by the document no.
github.com/adrogin/BCVi...

Looking forward to be on stage at @bctechdays.com with @stefanmaron.com

And the blank message in this lig entry does not help either. It would be worth adding something meaningful.

Search in the repository points to Master Data Management in ALAppExtensions, just a feature usage log.

No other method I'm aware of. And this way, it looks good. It's a clear intention: "I'm only going to use these fields". And then the runtime will adjust the list anyway, if any other fields must be included.

InStream/OutStream are essentially memory buffers with read/write pointers. In this snippet, CreateInStream instantiates the stream, but does not read the data, whilst ReadAs actually fills the buffer. I don't see as a hack, but it surely can be confusing.

Buffered inserts and deadlocks Recently I was debugging a very interesting deadlock case which I want to share in a blog post now. It was especially curious because, as it turned out, it was caused by the Business Central's bulk in...

I almost never use the return value of the Insert function in AL, but recently I found it useful in avoiding a specific deadlock situation. And the deadlock itself was a very curious one.

www.keytogoodcode.com/post/buffere...

The holidays are almost over, next week I am back to work and other Business Central activities. But these two weeks made a great breakaway tour.

Filtering on FlowFields It has been said so many times that filtering on FlowFields is bad for performance that we are used to avoiding this kind of filters in AL at all costs. So whenever a developer needs to select a subse...

Are you afraid of filters on FlowFields in AL? Well... There is definitely a reason to be cautious. But it may not be as bad as it seems, and certainly not as bad as some other methods which are used to avoid these filters.

www.keytogoodcode.com/post/filteri...

Table indexes and inserts Following my series of posts where I have been addressing various aspects of table indexing, this post delves into the same old argument from a different point of view. Previously, I measured timings ...

Following my series of posts where I've been addressing various aspects of table indexing, this post delves into the same argument from a different point of view. Now I want to ask a question: how indexing impacts query performance from the insert/update side.

www.keytogoodcode.com/post/table-i...

Temporary tables in procedure parameters When we declare multiple instances of a record variable referencing the same table in AL code, we essentially receive different instances of the active record, but the same underlying database table b...

Did you ever get confused with Business Central's temporary tables and the way temp tables are passed to procedure parameters? I did too, so I wrote a post as a memo.

www.keytogoodcode.com/post/tempora...