Misleading error message when property defined with invalid type

bgrommes

I unwittingly (witlessly??) defined a protected property named _columnCount of type Integeer and did not get an error in the property definition dialog. Maybe this is intended because for all it knows there's an Integeer class somewhere, or will be. IDK. Probably it should not allow any class name that isn't built-in or isn't defined in the solution. I'd be okay with not being able to make an Integeer property without first making an Integeer class. That would have been the end of it at any rate.

What ended up happening was two things:
1) I wrote this line of code and didn't get an error in real time from Analyser:

_columnCount = GetTrueColumnCount() # this method returns an Integer, not an Integeer

2) I then tried to pass _columnCount as the 3rd argument to another method with this signature:
list As Array(Of T), value As T, Optional count As Integer = list.Count
When I did this, I got an error to the effect, "Cannot pass type Array(Of String) to argument requiring Array(Of T)". Which of course is absurd.

Anyway I noticed that if I dropped the optional 3rd argument, the strange error went away. Further fiddling revealed that if I would hover over the offending property reference the tooltip would read "Property _columnCount As Unknown". Which led me to the typo in the property definition. But I had been paying attention to the red gutter diamond and its error message about the first argument.

So I guess this is a double bug: a property shouldn't be definable as a non-existent type, and the error message I should have gotten using that malformed class should have pointed to the argument that was the actual problem.

Garry

Thanks for the excellent report. This is now fixed in the next release: https://feedback.objo.dev/bug/1133

There were two related problems. Firstly, Studio wasn’t displaying the compiler’s unknown-type diagnostic against the property type field. It now flags the field directly with a message such as:

Unknown type 'Integeer'. Did you mean 'Integer'?

Secondly, generic type inference was giving up completely whenever any argument had an unknown type. That discarded the valid inference from the other arguments and caused the misleading Array(String) to Array(T) error against the first parameter. Inference now preserves everything it can learn from the valid arguments and suppresses that false follow-on error.

Assigning a value to an unknown type remains deliberately permissive, because otherwise one typo could produce a cascade of unrelated errors.

bgrommes

Garry Thanks, Garry! It's the accretion of all these little quirks / arguable bugs that make a development environment annoying, and conversely, their fix making it a pleasure to work with. No one of them is a show-stopping big deal, and my expectations of how a thing should work are not necessarily the same the next person's -- but in reporting these I try to ask the question, have I ever had to deal with this in other IDEs that I like and use, and if the answer is no, then other people's expectations would probably be violated also. When you're on a deadline at 2 hours past your bedtime you really don't want the IDE distracting you from the actual problem.

I also know just enough to be dangerous and can guess that some of these fixes make things more complex for you to deal with. Just want you to know how much I truly appreciate that you understand the importance of dealing with the little things!