WeakRef usage

There was a good question from Markus in a recent comment on a different post of mine.

In children of a window we should keep track of the parent window by using WeakRef.
But what if I have a class cDepartment with a list of employees defined as Employees() as cEmployee – should that then better also be defined as WeakRef? And why or why not?

First I’ll say “it depends” as there’s certainly no single correct answer to such a question. In many cases it will depend on the expected usage.

WeakRefs are a curious thing. In Xojo most times you have a reference to an object its a strong reference. What happens is that the objects reference count is increased and only when that counter gets set to 0 will the object be destroyed.

For instance, if we have code like the following

Class Foo

  Sub Destructor()
  End Sub

End Class

dim firstRef as Foo
dim secondRef as Foo

firstRef = new Foo 
// the object firstRef refers to now has a reference count of 1

secondRef = firstRef 
// since secondRef refers to the SAME object as firstRef
// the object firstRef refers to now has a reference count of 2

firstRef = nil
// the object firstRef refers to now has a reference count of 1

secondRef = nil
// the object firstRef refers to now has a reference count of 0
// and it will be destroyed and the breakpoint in the destructor
// will be encountered

Now what happens if we use a WeakRef instead for secondRef ? If we turn that code into the following

Class Foo

  Sub Destructor()
  End Sub

End Class

Dim firstRef As Foo
dim secondRef as Foo
Dim weakSecondRef As WeakRef // <<<<<<< note the type !

firstRef = New Foo 
// the object firstRef refers to now has a reference count of 1

weakSecondRef = New WeakRef(firstRef)
// since weakSecondRef is a WeakRef the object 
// firstRef refers does NOT increment its reference count

firstRef = Nil
// the object firstRef refers to now has a reference count of 0
// and it will be destroyed and the breakpoint in the destructor
// will be encountered now

// but what about weakSecondRef's contents ???
// lets fetch that value and see whats in there now
secondRef = Foo(weakSecondRef.Value)

break // <<< here you will find that the returned value is NIL !

And if you try to use the object contained by the weakref secondRef you will find that it is nil. A weakref allows an object to be referenced in a way that it does not increment the original objects reference count which CAN let that object be nil’d in other code.

When you use a weakref you MUST check the VALUE of the weakref to see if what you get back is NIL because the object may have been set to nil elsewhere.

Now that we have the basics of weakrefs out of the way lets turn back to Markus’ original and my original reply of “it depends”.

n children of a window we should keep track of the parent window by using WeakRef.

Indeed you may want to have child windows retain a weakref to refer to a parent window. It would allow you to close the parent and not the child, as the weakref will not increment the reference count of the parent window. But, you may want the parent to not be set to nil when there are child windows.

This decision will depend entirely on how your UI is supposed to work. If those child windows should not exist without a parent then a hard reference to the parent may make perfect sense.

But what if I have a class cDepartment with a list of employees defined as Employees() as cEmployee – should that then better also be defined as WeakRef? And why or why not?

Again this will depend entirely on how the cDepartment class should behave. Can cEmployees be added & removed at runtime and should the cDepartment class be able to deal with this ? Or should cDepartment need to be reconstituted when such changes of its employees are made ? The correct behaviour isnt dictated by the underlying code and whether you should use weakrefs but by the needs of the application and how cDepartment should behave.

You could always use weakrefs but this could have unintended consequences for the cDepartment class as this would allows its cEmployees to be removed at runtime. If you had used a FOR EACH loop with an iterator to perform some function you would need to handle the exceptions that could be encountered. Iterators will be invalidated by removal of an employee while iterating over the list and will raise an exception of this occurs. Weakrefs would permit this. Hard references would not.

2019, The Lost Year, in review

2019 has been an interesting year to say the least.

Mine started off with the septic system needing to be replaced in the first week of January as the old system had been installed in a manner that it was not repairable and it had failed. So in the cold of January our yard was torn up and excavation work went on. One of our basement bedrooms had to have its floor cracked open and a new main line to the septic system installed. $25K later it was all back in working order and we just had landscaping to fix up.

Awesome start !

Work on Xojo 2019r1 was progressing when I was suddenly let go in March.

I panicked a bit. But things have worked out well for me so far and I’m happy doing what I’m doing for a really great group of clients. Since then I’ve created 300+ Feedback reports.

Xojo 2019r1 was released in April and thats the last release that I really contributed much to. 2019r1.1 went out in May.

Betas for Xojo 2019r2 started shortly after that and this brought API 2.0 and a fair number of bugs and complaints along with it. Especially aggravating were changes to many property names, method names and event names. Event names proved to be the most difficult for developers of all sorts, especially in the third party arena, to deal with. For my uses it was a beta that was going along fine until API 2.0 hit then it made it clear that my clients projects were NOT going to transition easily so I more or less stopped testing entirely.

Xojo 2019r2 was released in Oct. There was more angst expressed about the renames by users that were not part of the beta program.

About this time I started taking training to be a Ski Patroller. I’d thought about going and getting certified as a ski instructor again but Ski Patrol made more sense.

2019r2.1 with a number of bugs fixes followed in Nov. Of significant note 2019r2 was pulled from further distribution and 2.1 reverted the changes to event names. This was a welcome reversion. I think a number of people were surprised that Xojo took this course of action. I really didnt test much in this release.

Shortly after, in Dec, 2019r3 was shipped and it brought iOS dark mode along with it and a handful of other fixes that were not iOS related. Again since it did not contain a lot of fixes that were really relevant to my clients I mostly didnt test it and have only lightly used it.

And about this time I passed my written and diagnostic exams to progress on to the on snow training to be a Certified Canadian Ski Patroller. (YAY !!!!!!)

For some the year was a non-event as far as Xojo went. No new OOP features, classes, controls or other significant improvements were added. A lot of changes were made. But Xojo hasn’t become more stable, capable, or reliable as a result of those changes. Just different. It’s about the same as it always has been and several new bugs seemed to be a result of API 2.0 changes. While API 2.0 is out now it isn’t a game changer in most respects. At least not yet. It’s hard to know what else might come along in the future but the roadmap doesnt seem to have anything for API 2.0 further out beyond expanding to iOS. What remains is still things that have been on the to do list for several years like Web 2.0, Android, and plugins written in Xojo.

For me both personally and professionally it was a year with big changes in my personal life and professional life. For Xojo not so much.

How was your year ?

#if Targetxxxx

I’ve seen this a lot in posts on the forums and even in code submitted as part of bug reports

If TargetWindows Then
  // Windows only code not excluded from compilation.
End If

The code inside the if will only execute on Windows. However becuase this code is just a normal if it HAS to COMPILE on ALL targets even if the intention is that it only be used on Windows (which it obviously is)

If you intend some code to ONLY be used on Windows you’re better off to use this form

#If TargetWindows Then
  // Declares for example

This code will ONLY exist in a Windows builds and so can ONLY execute on Windows. And, because it will only be used on Windows the compiler will ignore the code inside the #if … #endif on all other platforms so the code doesn’t even have to compile on macOS or Linux (which it would in the first example shown)

Usually when you see the first form you really want the second using #if

Next time you find yourself writing

if Target

reconsider if you really should be using #if instead

Classes, Instances and Objects

There’s a fun “theoretical question” on the forums.

What is the relationship between a class, object, and instance?

Lets see if I can help any (or maybe just confuse the heck out of folks even more)

In Xojo in the IDE when you define a CLASS you are creating a “template” for how every item of this type will behave. You’re defining the properties, the methods, events etc etc etc. And EVERY item that IsA <this type I defined> returns true for IsA one of these.

AT runtime when you application runs and it uses the NEW operator to create a new item of this type you are creating new INSTANCES. Instances are the ACTUAL copies you create that you can manipulate at runtime.

And EVERY instance in Xojo IsA Object. Its the base object for EVERY dynamically created type where you use NEW to create one.

So CLASSES are a design time thing.

INSTANCES and OBJECTS are runtime things.

Silent readership

I’m curious about something. Actually I’m curious about a LOT of things; just ask my Canadian Skip Patrol instructors 😛
But I’m REALLY curious about one thing when it comes to this blog.

I can see there are a decent number of readers & views thanks to the stats that WordPress gives me. But, there’s very few comments.

I’m really curious why people don’t comment. Do the same things that make it so you don’t comment here also apply to other venues like the Xojo forums ? The reason I ask is that, as anyone who knows me well can vouch for, when I have something to say I say it. And if I have a strong opinion about something its nearly impossible to get me to shut up. Again you can ask anyone who knows me well – and if it’s Bob’s Keeney reading this I’m 100 sure he’s just shaking his head going “Yup!”

Even just a “Hey thanks I read that post and it was informative” is great feedback to get. Or “Hey I found a typo”. And yeah you can tell me I’m being dumb if I post something dumb 🙂

So what is it that keeps you from commenting ? And yes I get the irony of asking people who don’t comment on other posts to comment on this one 🙂

Performance tweaks

In Xojo there are a number of things you can do to improve the performance of your code.

There are various pragmas you can enable that can improve speed – I’d suggest only doing this once you code is fully debugged and known to be working properly as turning these checks of can mean IF you have a bug your application just crashes hard.

And there are other things you can do like lifting out code that repeatedly does lookups in arrays.

So instead of

dim something(1000) as String

// intervening code to load data into the array

for i as integer = 0 to something.ubound

   if something(i) = "abc" then
   elseif something(i) = "def" then
   elseif something(i) = "ghi" then
   end if

you might rewrite this as

dim something(1000) as String

// intervening code to load data into the array

// 1) eliminate repeatedly accessing the Ubound property
dim maximumIndex as integer = something.ubound 

for i as integer = 0 to maximumIndex

   // 2) eliminate repeatedly doing an array access
   dim thisItem as string = something(i)

   if thisItem = "abc" then
   elseif thisItem = "def" then
   elseif thisItem = "ghi" then
   end if


It would be nice if the compiler could/would do more optimization of code. There are a host of optimizations that you could apply manually like common subexpression elimination and loop invariant code motion.

But sometimes the biggest performance wins are not from tweaks like these. Often applying a different algorithm has a much bigger bang for the buck than anything.

When I worked at Xojo there were a couple areas where simply by inverting how some data was accessed in memory the entire process was sped up enormously.

Databases will do this and may use the MOST selective index to retrieve data first so the initial set of data is very tiny, instead of starting with the most broad set of data, and then winnowing it down further. By doing this they can decrease the amount of memory required and the amount of data to iterate though to satisfy whatever query you’ve executed.

When you have a performance issue I would START by reconsidering how the data is accessed and used and whether you can alter the algorithm to gain an initial big speed up by doing things in a different order.

And once you get the algorithm working the way you want then apply all the other tweaks to code to squeeze the most out of it.

And JUST to be clear make sure you do timings in a COMPILED and BUILT version NOT aa debug version using RUN as a debug version still has a lot of debug related code in it that can influence the results and MAY mislead you.

Structures vs Classes

While structures & classes may seem to be very similar in most respects there aren’t many cases when I would suggest you use a structure instead of a class.

A class with public member properties isnt that different from a structure in terms of how your code uses it.

The biggest difference is that for a class you must use new to get a new instance. With a structure you just declare it and use it.

Structures are handy for some things. If you have to read or write a specific binary file format you might use a structure. If you’re going to call some OS API via a declare then structures are useful. And if you’re reading or writing binary data to binary streams via sockets or other communication channels they are useful.

But there are downsides. Since structures are JUST data any methods to manipulate them have to go in a module. Unlike a class you cannot add methods to the structure itself. And so you end up using a very non-OOP style – maybe a module with the structure defined in there along with all the code that manipulates the structure.

Classes allow you to just bundle up the data and code into a single item and all the code that manipulates the class is part of the class – along with any hidden methods it might need that can be completely hidden from the outside world.

If your still using VB style syntax and creating a lot of structures I’d suggest you give using classes a try.

Once you get the hang of it you will appreciate how classes let you organize your code and data in a nice single item.


Ever wanted to define a class, maybe a vector class, and make it possible to write code like

dim v1 as new Vector(1,1)
dim v2 as new Vector(2,2)

if v1 > v2 then
  // do something when v1 > v2
elseif v1 < v2 then
  // do something when v1 < v2
  // do something when v1 = v2
end if

Perhaps you have some other kind of class you’d like to define that has some kind of custom mechanism to compare itself to other instances – or even other data types.

If so you need to know how to implement operator_compare

First off operator_compare can ONLY be implemented in a Class. You cannot extend existing classes and add it to them.

It’s a method, or several methods, you add to a class that will be called whenever you use the comparison operators =, <, >, <=, >= and <> are used. Note that this has implications for your code if you are using = to check is one instance is the same instance as another. In that case you should probably use IS rather than =.

Beyond that operator_compare is pretty straight forward.

You define the method with that name, and the parameter is whatever type you wish to compare your instance, the “self” instance, to. So you can have many as each signature would be different. The return value is and integer that is

  • < 0 means SELF is “less than” the passed parameter
  • = 0 means SELF is “equal to” the passed parameter
  • > 0 means SELF is “greater than” the passed parameter

and you get to define what less than, equal to, and greater than mean.

So suppose our Vector class was defined as follows

Class Vector
   protected x as double
   protected y as double

   Public Sub Constructor(x as double, y as double)
      Self.x = x
      Self.y = y
   End Sub

   Function Operator_compare(other as vector) as Integer
      Dim a, b As Integer
      a = Self.x ^ 2 + Self.y ^ 2
      b = other.x ^ 2 + other.y ^ 2
      If a > b Then Return 1
      If a = b Then Return 0
      If a < b Then Return -1
    End Function
End Class

And now our original code from way back would work

dim v1 as new Vector(1,1)
dim v2 as new Vector(2,2)

if v1 > v2 then
  // do something when v1 > v2
elseif v1 < v2 then
  // do something when v1 < v2
  // do something when v1 = v2
end if

Now you could make Vectors compare themselves to Strings if you really wanted. I’m not sure what exactly that might mean – but you could. You might try implementing this as

Public Function Operator_compare1(other As string) as Integer
  Dim myRepresentation As String = "(" + Str(x,"-######.00") + "," + Str(y,"-######.00") + ")"
  If myRepresentation = other Then Return 0
  return -1
End Function

Again I’m not sure what this might truly mean but we’ll indicate that unless they are the exact same string the “self” is < the string value. Suppose you try to test this with

Dim v1 As New Vector(1,1)

If v1 = "(foo)" Then
End If

And now you Vectors can be compared to Strings. With extra overloads of Operator_compare you can make your classes compare themselves to all sorts of your other classes & types.

Enjoy !