So, 2.0.12 is late… what’s cooking?

Recently I tweeted this:

The release of Rubberduck 2.0.12, due 5 days ago, is being delayed because we have something awesome cooking up. Give us 2-3 more weeks ūüôā

TL;DR: if awesomeness can be cooked, that’s what’s cooking.

The amount of work that went into the upcoming release is tremendous. We’ve been trying to figure out exactly what was blowing up when the VBE dismantled itself and the host was shutting down, causing that pesky crash on exit… ever since we’ve introduced WPF user controls in dockable toolwindows. And at last, solved it.

We’ve been working on improving performance¬†and thread safety of the entire parsing engine, and fixed a few grammar/parser bugs on the way, including a long-standing bug that made redundant parentheses trip a parse exception, another with the slightly weird and surely redundant¬†Case Is =¬†syntax, and @Magic annotations can now legally be followed by any comment, which is useful when you want to, well, annotate an annotation:

'@Ignore ProcedureNotUsed; called by [DoSomething] button on Sheet12
Public Sub DoSomething()
End Sub

We’ve enhanced the COM reference collector such that the resolver has every bit of useful information about everything there is to know in a type library referenced by a VBA project. This allows us to enhance other features, like the¬†context-sensitive¬†commandbar that tells you what Rubberduck is your selection as, e.g. a TextBox¬†control in a UserForm:


(don’t mind that “Serialize” button – it’s only there in debug builds ;^)

Oh, and then there’s the interactions with the website – we’ll be running the inspections and the indenter on the website, and we’ll have the ability to (optionally) have Rubberduck know when a new version is available!

2.0.12 is going to be epic.

The 2.0 build

And then there’s even more: we’re going to make¬†the inspections a concern of the parser engine, and turn them into parse tree node annotations – which means the code that currently finds the Declaration¬†that’s currently selected (or one of its references), can also be used to find inspection results associated with that particular Declaration;¬†this will probably prompt a redesign of how we present inspection results, and will definitely improve performance and memory footprint.

One of the best 2.x features is probably going to be the add/remove references dialog, which is currently merely prototyped. Beefing up unit testing with data-driven tests is also going to be a big one.

And when you see where we want to be for 3.0 (code path analysis & expression resolution, plug-in architecture, a subclassed CodePanethat actually tells us what’s going on, perhaps even with our own enhanced IntelliSense, more host-specific behaviors, TONS of new inspections), …this project is so awesome, I could just keep going on and on.

Not coming soon enough? I know, right!


To be continued…


Nothing to declare

Somewhere in¬†the first batch of issues/to-do’s we created when we started Rubberduck on GitHub (Issue# 33 actually), there was the intention to create a tool that could locate undeclared variables, because even if you and I use Option Explicit and declare all our variables,¬†we have brothers and sisters that have to deal with code bases that don’t.

So we tried… but¬†Rubberduck simply couldn’t do this with the 1.x resolver: identifiers that couldn’t be resolved were countless, running an inspection that would pop a result for every single one of them would have crippled our¬†poor little¬†duckling… so we postponed it.

The 2.0 resolver however, thinks quite literally like VBA itself, and knows about all available types, members, globals, locals, events, enums and whatnot, not just in the VBA project, but also in every referenced COM library: if something returns a type other than Variant or Object, Rubberduck knows about it.

The role of the resolver is simple: while the parse tree of a module is being traversed, every time an identifier is encountered it attempts to determine which declaration¬†is being referred to. If the resolver finds a corresponding declaration, an¬†IdentifierReference is created and added to the¬†Declaration instance. And when¬†the resolver can’t¬†resolve the identifier (i.e. locate the exact¬†declaration¬†the identifier is¬†referring to), a¬†null reference was returned and, unless you have detailed logging enabled, nothing notable happens.

As of the last build, instead of “doing nothing” when a reference to variable can’t be resolved to the declaration of that variable, we create a¬†declaration on the spot: so the first appearance of a variable in an executable statement becomes the “declaration”.

We create an implicit Variant variable declaration to work with, and then this happens:


With a Declaration object for an undeclared variable, any further reference to the same implicit variable would simply resolve to that declaration Рthis means other Rubberduck features like find all references and refactor/rename can now be used with undeclared variables too.

Rubberduck is now seeing the whole picture, with or without Option Explicit.

The¬†introduce local variable¬†quick-fix simply inserts a “Dim VariableName As Variant” line immediately above the first use in the procedure, where¬†VariableName¬†is the unresolved identifier name. The variable is made an explicit Variant, …because there’s another inspection that could fire up a result if we added an¬†implicit¬†Variant.

The quick-fix doesn’t assume an indentation level – makes me wonder if we should run the indenter on the procedure after applying a quick-fix… but that’s another discussion.

To be continued…

2.0 Beta is here!

A little while ago, we issued an alpha release of Rubberduck 2.0, just because, well, v1.4.3 had been around since July 2015, and we wanted to say “look, this is what we’ve been working on; it’s not nearly stable yet, but we still want to show you what’s coming”.

Time flies. 6 whole weeks, 353¬†commits (plus a few last-minute ones), 142* pull requests from 8 contributors, 143* closed issues, 60* new ones, 129,835¬†additions¬†and 113,388 deletions in 788* files later, Rubberduck still has a number of known issues, some involving COM interop, most involving COM reflection and difficulties in coming up with a host-agnostic way of identifying the exact types we’re dealing with.

It might seem obvious, but knowing that¬†ThisWorkbook is a¬†Workbook object is anything but trivial – at this point we know that¬†Workbook¬†implements a WorkbookEvents¬†interface; we also know what events are exposed: we’re¬†this close to connect all the dots and have a resolver that works the way we need it to.

So what does this mean?

It means a number of false positives for a number of inspections. It means false negatives for a number of others.

Other than that, if the last version you used was 1.4.3, you’re going to be blown away. If the last version you used was 2.0.1a, you’ll appreciate all the work that just went into this beta build.

There are a number of¬†little minor issues here¬†and there, but¬†the major issues we’re having pretty much all revolve around¬†resolving identifier references,¬†but I have to admit I don’t like unit test discovery working off the parser – it just doesn’t feel right and we’re going to fix that soon.

Speaking of unit testing… thanks to @ThunderFrame’s hard work,¬†Rubberduck 2.0 unit tests now work in Outlook, Project, Publisher and Visio.

@Hosch250 If you get unit testing to work in outlook I’ll eat my hat.

– @RubberDuck 2016-05-13

So Chris, how’s¬†the hat?

Stay tuned, things are going to snowball from this point on – we’ll be releasing much more often than we have been.

*From the GitHub “Pulse” page between May 7 and June 7, 2016.

VBA Rubberducking (Part 4)

This post is the fourth in a series of post that walk you through the various features of the Rubberduck open-source VBE add-in.

  • Part 1¬†introduced¬†the¬†navigation¬†features.
  • Part 2¬†covered the¬†code inspections.
  • Part 3 featured the¬†unit testing¬†feature.


At first we were happy to just be able to inspect the code.


Quickly we realized “inspection quick-fixes” could be something else;¬†some of the inspections’ quick-fixes are full-fledged automated refactoring¬†operations. Renaming an identifier – and doing it right¬†– is very different than just Ctrl+H/replace an identifier. Manually removing an uneeded parameter in an existing method breaks all call sites and the code no longer even compiles; Rubberduck sees all call sites, and knows which argument to remove everywhere to keep the code compiling.. and it’s much faster than doing it by hand!

Rubberduck 1.3 had Rename and Extract Method refactorings; v1.4.3 also had Remove Parameters and Reorder Parameters refactorings.

Rubberduck 2.0 introduces a few more.


The context menu commands are enabled depending on context; be it the current parser state, or the current selection.


That’s a pretty well-named refactoring. It deals with the impacts on the rest of the code base, of renaming pretty much any identifier.

Extract Method

Pretty much completely rewritten, v2.0 Extract Method refactoring is becoming pretty solid. Make a valid selection, and take that selection into its own member, replacing it with a call to the extracted code, all parameters and locals figured out for you.

Extract Interface

VBA supports interface inheritance; Rubberduck makes it easy to pull all public members of a module into a class that the original module then¬†Implements. This is VBA’s own way of¬†coding against abstractions. Unit tests love testing code that’s depending on abstractions, not concrete implementations, because then the tests can provide (“inject”)¬†fake dependencies and test the applicative logic without triggering any unwanted side-effects, like displaying a message box, writing to a file, or to a database.

Implement Interface

Implementing all members of an interface (and all members of an interface must be implemented) can be tedious; Rubberduck automatically creates a stub method for every member of the interface specified in an Implements statement.

Remove/Reorder Parameters

Reworking a member’s signature is always annoying, because then you have to cycle through every single call site and update the¬†argument list; Rubberduck knows where every call site is, and updates all call sites for you.

Move Closer to Usage

Variables should have the smallest possible scope. The “scope too wide” inspection uses this refactoring to move a declaration just above its first usage; it also works to rearrange “walls of declarations” at the top of a huge method you’re trying to cut into more manageable pieces.

Encapsulate Field

Fields are internal data, implementation details; objects shouldn’t expose public fields, but rather, encapsulate them and expose them as properties. Rubberduck turns a field into a property with only as much effort as it takes to name the new property.

Introduce Parameter/Field

Pretty much the antagonist of¬†move closer to usage, this refactoring promotes a local variable to a parameter or a field, or a parameter to a field; if a new parameter is created, call sites will be updated with a “TODO” bogus argument that leaves the code uncompilable until an argument is supplied for the new parameter at all call sites.

More refactorings are planned for 2.1 and future versions, including Inline Method (the inverse of Extract Method), to move the body of a small procedure or function into all its call sites. Ideas for more refactorings and inspections? Suggest a feature!


VBA Rubberducking (Part 3)

This post is the third in a series of post that walk you through the various features of the Rubberduck open-source VBE add-in.

  • Part 1¬†introduced¬†the¬†navigation¬†features.
  • Part 2¬†covered the¬†code inspections.

Unit Testing

If you’ve been following Rubberduck since its early days, you already know that this is where and how the project started. Before Rubberduck was a VBE add-in, it was an Excel add-in completely written in VBA, that started with this Code Review post; before Rubberduck was even¬†named “Rubberduck”, it was a C# port of this¬†VBA code – the idea being to enable writing and running unit tests¬†beyond Excel, in Access and Word VBA as well, without having to replicate all that code in multiple add-in projects.

Zero Boilerplate

There¬†are other VBA unit testing solutions out there. A lot require quite a bit of boilerplate setup code; those written in VBA require programmatic access to the VBIDE object model, which may be a security concern (you’re allowing VBA to execute code that can¬†generate and run VBA code after all). Rubberduck unit tests require neither. Because it’s a VBE add-in, Rubberduck¬†already¬†has programmatic access to the code in the IDE, and the ability to scan, modify, generate and execute VBA code – without requiring a dent in your corporate security policy.

Rubberduck requires pretty much zero boilerplate. This is a fully working test module:

 Private Assert As Rubberduck.AssertClass
 Public Sub FooIs42()
     Const expected As Integer = 42
     Dim actual As Integer
     actual = Module1.GetFoo
     Assert.AreEqual expected, actual, "Nope, not 42."
 End Sub

Okay, it’s just an example. But still, it shows how little is¬†required for it to work:

  • A @TestModule annotation in the¬†declarations section of a standard module.
  • A¬†Rubberduck.AssertClass¬†instance, which can be late or early-bound.
  • A¬†@TestMethod annotation to formally identify a test method.

That’s all. And up until recently, the¬†@TestMethod annotation was optional – in Rubberduck 1.x, if you had a public parameterless method with a name that starts with “Test”, in a standard module, Rubberduck treated it as a test method.¬†This is changing in 2.0, as we are¬†making the¬†@TestMethod annotation mandatory, favoring explicitness over implicit naming conventions. Test methods still need to be public and parameterless, and in a standard module though.


Now, let’s say GetFoo¬†returning 42 is a business requirement, and that something needs to change in¬†Module1 or elsewhere and, inadvertently,¬†GetFoo starts returning 0. If you don’t have a unit test that documents and verifies that business requirement, you’ve introduced a bug that may take a while to be discovered. However if you¬†do have a test for it, and that you’ve made it a habit to run your test suite whenever you make a change¬†just to be sure that all the business requirements are still met…


Then you have a failing test, and you know right away that your modification has subtly introduced a change in behavior that will be reported as a bug sooner or later.

If you’ve already written unit tests, I’m probably preaching to the choir here. If you’ve only ever written VBA code, it’s possible you’ve¬†heard¬†of unit testing before, but aren’t quite sure how you could make¬†your code work with it.

Luckily, the key concepts are language-agnostic, and VBA definitely has support for everything you need for full-blown Test-Driven Development.

Code Against Abstractions

Whether you’re writing C#, Java, PHP, Python, Ruby, or VBA, if your code is¬†tightly coupled¬†with a UI, accessing the file system, a Web service, a database, …or a worksheet, then it’s not¬†fit¬†for a unit test, because a¬†unit¬†test…

  • Should be fast
  • Should¬†not have side-effects
  • Should not depend on (or impact) other tests
  • Should have all dependencies under control
  • Should test¬†one thing, and¬†have¬†one reason to fail

Wait. My code¬†is¬†accessing a worksheet. Does that mean I can’t write tests for it?

Yes and no. I’ll tell you a secret. Quite a lot of VBA posts I see on Code Review are asking for tips to get their code to run faster with large data sets. Something I often say in my reviews, is that¬†the single slowest thing you can do in VBA is access a worksheet.

Don’t code your logic against the worksheet, code your logic against¬†an abstraction¬†of the worksheet. An array is often all you need: refactor your logic to work with an array instead of a worksheet, and not only you’ll be able to write a test that gives it any array you want, your code will also perform better!

Encapsulate your logic in class modules, test the public interface; if the logic brings up a UI (even a message box!),¬†extract that piece of code elsewhere¬†– make it the responsibility of¬†something else, get it out of the way so your tests can concentrate on the actual important things that they’re testing for.



A whole book could be written about reducing coupling in code, increasing cohesion, and writing tests in general. Poke around, research a bit. You’ll see where Rubberduck wants to take your VBA code.

The Test Explorer

Rubberduck’s¬†Test Explorer¬†offers two main “sets” of commands: “Run”, and “Add”.

The “Add” menu lets you easily add a¬†test module to your project, and from there you can just as easily add a¬†test method, one of two templates:

  • Test Method is the standard Arrange-Act-Assert deal, with error handling that ensures the test will correctly fail on error¬†and report¬†that error.
  • Test Method (Expected Error) is the same AAA deal, except this template is for writing tests that are¬†expected¬†to raise a specific runtime error; such tests¬†fail if the expected error isn’t raised.

The “Run” menu lets you easily run all, or a subset of the tests – e.g. you might want to only run the tests that failed the last time you ran them.

Results can be regrouped either by outcome or by location (project/module), and again can be copied to the clipboard with a single click.

Test settings let you control the contents of the test module template:


Binding mode determines whether the¬†AssertClass instance is going to be declared “As Object” (late-bound, default) or “As New Rubberduck.AssertClass” (early-bound).

Type safety determines whether the Assert variable is going to be a Rubberduck.AssertClass (strict) or a Rubberduck.PermissiveAssertClass (permissive); the permissive asserts differs with the strict (original and default) version in that equality checks are more closely modeled on VBA equality rules: with a permissive assert, an Integer value of 254 can be compared to a Byte value of 254 and deemed equal. Strict equality requires the types to match, not just the value.

Test Module Template checkboxes determine whether the @TestInitialize, @TestCleanup, @ModuleInitialize and @ModuleCleanup method stubs are going to be generated, and also whether creating a new test module creates a test method by default.

All these settings only affect new test modules, not existing ones.

The Assert Class

Tests¬†assert¬†things. Without assertions, a Rubberduck test can’t have a meaningful result, and will simply pass. The IAssert¬†interface (implemented by both¬†AssertClass and¬†PermissiveAssertClass) exposes a number of members largely inspired by MS-Tests in Visual Studio:

Name Description
AreEqual Verifies that two specified objects are equal. The assertion fails if the objects are not equal.
AreNotEqual Verifies that two specified objects are not equal. The assertion fails if the objects are equal.
AreNotSame Verifies that two specified object variables refer to different objects. The assertion fails if they refer to the same object.
AreSame Verifies that two specified object variables refer to the same object. The assertion fails if they refer to different objects.
Fail Fails the assertion without checking any conditions.
Inconclusive Indicates that the assertion cannot be verified.
IsFalse Verifies that the specified condition is false. The assertion fails if the condition is true.
IsNothing Verifies that the specified object is Nothing. The assertion fails if it is notNothing.
IsNotNothing Verifies that the specified object is not Nothing. The assertion fails if it isNothing.
IsTrue Verifies that the specified condition is true. The assertion fails if the condition is false.

To be continued…

VBA Rubberducking (Part 2)

This post is the second in a series of post that walk you through the various features of the Rubberduck open-source VBE add-in. The first post was about the navigation features.

Code Inspections


Back when the project started, when we started realizing what it meant to¬†parse¬†VBA code, we knew we were going to use that information to tell our users when we’re seeing anything¬†from possibly iffy¬†to this would be a¬†bug¬†in their code.

The first one to be implemented was¬†OptionExplicitInspection. The way Rubberduck works, a variable that doesn’t resolve to a known declaration simply doesn’t exist. Rubberduck is designed around the fact that it’s working against¬†code that VBA compiles; is also needs to assume you’re working with code that declares its variables.

Without ‘Option Explicit’ on, Rubberduck code inspections can yield false positives.

Because it’s best-practice to always declare your variables, and because the rest of Rubberduck won’t work as well as it should if you’re using undeclared variables, this inspection defaults to Error severity level.

OptionExplicitInspection was just the beginning. As of this writing, we have implementations for 35 inspections, most with one or more one-click quick-fixes.

35 inspections?

And there’s a couple more left to implement, too. A lot of inspections rely on successful parsing and processing of the entire project and its references; if there’s a parsing error, then Rubberduck will not produce new inspection results. When parsing succeeds, inspections run automatically and the “status bar” indicates¬†Ready¬†when it’s completed.

  1. AssignedByValParameterInspection looks for parameters passed by value and assigned a new value, suggesting to either extract a local variable, or pass it by reference if the assigned value is intended to be returned to the calling code.
  2. ConstantNotUsedInspection looks for constant declarations that are never referenced. Quick-fix is to remove the unused declaration.
  3. DefaultProjectNameInspection looks for unnamed projects (“VBAProject”), and suggests to refactor/rename it. If you’re using source control, you’ll want to name your project, so we made an inspection for it.
  4. EmptyStringLiteralInspection finds “” empty strings and suggests replacing with vbNullString constants.
  5. EncapsulatePublicFieldInspection looks for public fields and suggests making it private and expose it as a property.
  6. FunctionReturnValueNotUsedInspection locates functions whose result is returned, with none of the call sites doing anything with it. The function is used as a procedure, and Rubberduck suggests implementing it as such.
  7. IdentifierNotAssignedInspection reports variables that are declared, but never assigned.
  8. ImplicitActiveSheetReferenceInspection is Excel-specific, but it warns about code that implicitly refers to the active sheet.
  9. ImplicitActiveWorkbookReferenceInspection is also Excel-specific, warns about code that implicitly refers to the active workbook.
  10. ImplicitByRefParameterInspection parameters are passed by reference by default; a quick-fix makes the parameters be explicit about it.
  11. ImplicitPublicMemberInspection members of a module are public by default. Quick-fix makes the member explicitly public.
  12. ImplicitVariantReturnTypeInspection a function or property getter’s signature doesn’t specify a return type; Rubberduck makes it return a explicit Variant.
  13. MoveFieldCloserToUsageInspection locates module-level variables that are only used in one procedure, i.e. its accessibility could be narrowed to a smaller scope.
  14. MultilineParameterInspection finds parameters in signatures, that are declared across two or more lines (using line continuations), which hurts readability.
  15. MultipleDeclarationsInspection finds instructions containing multiple declarations, and suggests breaking it down into multiple lines. This goes hand-in-hand with declaring variables as close as possible to their usage.
  16. MultipleFolderAnnotationsInspection warns when Rubberduck sees more than one single @Folder annotation in a module; only the first annotation is taken into account.
  17. NonReturningFunctionInspection tells you when a function (or property getter) isn’t assigned a return value, which is, in all likelihood, a bug in the VBA code.
  18. ObjectVariableNotSetInspection tells you when a variable that is known to be an object type, is assigned without the¬†Set keyword – this¬†is¬†a bug in the VBA code, and fires a runtime error 91 “Object or With block variable not set”.
  19. ObsoleteCallStatementInspection locates usages of the Call keyword, which is never required. Modern form of VB code uses the implicit call syntax.
  20. ObsoleteCommentSyntaxInspection locates usages of the Rem keyword, a dinosaurian syntax for writing comments. Modern form of VB code uses a single quote to denote a comment.
  21. ObsoleteGlobalInspection locates usages of the Global keyword, which is deprecated by the Public access modifier. Global cannot compile when used in a class module.
  22. ObsoleteLetStatementInspection locates usages of the Let keyword, which is required in the ancient syntax for value assignments.
  23. ObsoleteTypeHintInspection locates usages of type hints in declarations and identifier references, suggesting to replace them with an explicit value type.
  24. OptionBaseInspection warns when a module uses¬†Option Base 1, which can easily lead to¬†off-by-one bugs, if you’re not careful.
  25. OptionExplicitInspection warns when a module does not set¬†Option Explicit, which can lead to VBA happily compiling code that uses undeclared variables, that are undeclared because there’s a typo in the assignment instruction. Always use Option Explicit.
  26. ParameterCanBeByValInspection tells you when a parameter is passed¬†ByRef (implicitly or explicitly), but never assigned in the body of the member – meaning there’s no reason not to pass the parameter by value.
  27. ParameterNotUsedInspection tells you when a parameter can be safely removed from a signature.
  28. ProcedureCanBeWrittenAsFunctionInspection locates procedures that assign a single ByRef parameter (i.e. treating it as a return value), that would be better off written as a function.
  29. ProcedureNotUsedInspection locates procedures that aren’t called anywhere in user code. Use an¬†@Ignore annotation to remove false positives such as public procedures and functions called by Excel worksheets and controls.
  30. SelfAssignedDeclarationInspection finds local object variables declared¬†As New, which (it’s little known) affects the object’s lifetime and can lead to surprising/unexpected behavior, and bugs.
  31. UnassignedVariableUsageInspection locates usages of variables that are referred to before being assigned a value, which is usually a bug.
  32. UntypedFunctionUsageInspection recommends using String-returning functions available, instead of the Variant-returning ones (e.g. Mid$ vs. Mid).
  33. UseMeaningfulNamesInspection finds identifiers with less than 3 characters, without vowels, or post-fixed with a number – and suggests renaming them. Inspection settings will eventually allow “white-listing” common names.
  34. VariableNotAssignedInspection locates variables that are never assigned a value (or reference), which can be a bug.
  35. VariableNotUsedInspection locates variables that might be assigned a value, but are never referred to and could be safely removed.
  36. VariableTypeNotDeclaredInspection finds variable declarations that don’t explicitly specify a type, making the variable implicitly¬†Variant.
  37. WriteOnlyPropertyInspection finds properties that expose a setter (Property Let or Property Set), but no getter. This is usually a design flaw.

Oops, looks like I miscounted them… and there are even more coming up, including host-specific ones that only run when the VBE is hosted in Excel, or Access, or whatever.

The Inspection Results toolwindow

If you bring up the VBE in a brand new Excel workbook, and then bring up the inspection results toolwindow (Ctrl+Shift+I by default) you could be looking at something like this:


Most inspections provide one or more “quick-fixes”, and sometimes a quick-fix can be applied to all inspection results at once, within a module, or even within a project. In this case¬†Option Explicit can be automatically added to all modules that don’t have it in the project, using the blue Fix all occurrences in project¬†link at the bottom.

Or, use the Fix drop-down menu in the top toolbar to apply a quick-fix to the selected inspection result:


Each inspection has its own set of “quick-fixes”¬†in¬†the¬†Fix menu. A common one is¬†Ignore once; it inserts an¬†@Ignore annotation that instructs the specified inspection to skip a declaration or identifier reference..


The bottom panel contains information about the selected inspection result, and fix-all links that always use the first quick-fix in the “Fix” menu.¬†Disable this inspection turns the inspection’s “severity” to¬†DoNotShow, which effectively disables the inspection.

You can access inspection settings from the Rubberduck | Settings menu in the main commandbar, or you can click the settings button in the inspection results toolwindow to bring up the settings dialog:


If you like using the Call keyword, you can easily switch off the inspection for it from there.

The Copy toolbar button sends inspection results into the clipboard so they can be pasted into a text file or an Excel worksheet.

As with similar dockable toolwindows in Rubberduck, the way the grid regroups inspection results can be controlled using the Grouping menu:


The¬†refresh button causes a re-parse of any modified module; whenever parser state reaches “ready”, the inspections run and the grid refreshes – just as it would if you refreshed from the¬†Rubberduck¬†command bar, or from the¬†Code Explorer toolwindow.

To be continued…



VBA Rubberducking (Part 1)

The VBE editor was last updated in 1998 – back when VB6 was all the rage, and the .NET framework was probably just a little more than a nice idea.

Vanilla VBA editor

The VBE was just slightly less full-featured than its standalone counterpart, Visual Studio 6.0; however years went by, and the latest Visual Studio versions make the VBE look like an odd beast from another century.

Enter Rubberduck.


There are other VBE add-ins out there. For many years, VBA (and VB6) devs have loved using MZ-Tools and Smart Indenter Рperhaps the two most popular add-ins ever written for the VBE. One has a lightning-fast analyzer that is capable of finding unused declarations, and even locates references in commented-out code; the other offers a highly configurable indenter that lets you instantly apply an indenting style to an entire module, or more surgically to a single procedure. What does Rubberduck bring to the table?

Lots, lots, lots of things.

This post is the first in a series of post that walk you through the various features of the Rubberduck open-source VBE add-in.

Navigation Tools

One of the most frustrating aspects of the VBE, is its limited set navigation tools. Let’s recap what the¬†vanilla VBE¬†gives us:

  • Ctrl+F / “Find” is a little more than a basic text search, that lets you search and replace text in the current procedure, module, project, or selection. Although VBA isn’t case-sensitive, you can match case, and use pattern matching,¬†which isn’t exactly a regex search, but better than nothing.
  • Shift+F2 / “Go to Definition”, is actually fantastic: you can right-click any identifier and jump to its declaration – and if it’s an identifier defined in a referenced library, it takes you to its entry in the¬†Object Browser.
  • Ctrl+R / “Project Explorer” is a dockable toolwindow that lists all opened projects and the modules under them, in a convenient TreeView where you can double-click on a node and navigate there.
  • Ctrl+Shift+F2 / “Last Position”¬†is also fantastic: the VBE keeps a little stack of recent places you’ve been, and works like a “back” browser button that takes you back to where you were before. Quite possibly my personal favorite of all.
  • Bookmarks¬†is under-used… for a reason. You can toggle any line as a bookmark, and cycle through them, but there’s no place to see them all at once.

And… that’s about it. Let’s see what Rubberduck has to offer.

Code Explorer


This isn’t the final¬†version (we haven’t released it in 2.0 yet).¬†When it grows up, it wants to be a full-fledged replacement for the¬†Project Explorer. Its default hotkey even hijacks the Ctrl+R shortcut. Here’s what it does that the¬†Project Explorer¬†doesn’t do:

  • Drill down to module members, and then further down to list enum and user-defined type members.
  • See¬†constant values as they appear in code.
  • Navigate not only¬†to any module, but any field, enum member, constant, procedure, property get/let/set accessor, function,¬†imported library functions and procedures.
  • Rename¬†anything.. without breaking the code that references what you’re renaming.
  • Find all references to anything.
  • Indent¬†an entire project, or a selected module.

But the coolest thing is that Rubberduck’s¬†Code Explorer¬†takes special comments like this:


And then renders the module like this:


That’s right.¬†Folders. In VBA. Sure, okay, they’re not¬†real folders – it’s a trick, an illusion… but that¬†trick now means that with a simple annotation in every module, you can organize your VBA project¬†the way you want to; you’re no longer forced to search for a class module among 80 others in a large project, you’re free to regroup forms together with their related classes!

This feature alone is a game changer: with it, class¬†modules¬†can become first-class citizen; you don’t have to fear drowning in a sea of modules, and you don’t have to give them funky prefixes to have them sorted in a way that makes it anywhere near¬†decent to navigate.

Find Symbol

One of my favorite ReSharper features, is Ctrl+T / “go to anything”. When I realized we could have this feature in the VBE, I went ahead and did it. This simple feature lets you type the name of any identifier, and locate a specific instance of it:


This includes any variable, constant, type, enum, procedure, function, property, library function/procedure, parameter, …even line labels can be navigated to.

Just Ctrl+T, type something, hit ENTER, and you’re there. Or browse the dropdown list and click that “go” button.

Find all references

Whether you’re looking for all call sites of a procedure in your code, or you’re just curious about how many times you’re using the¬†vbNullString¬†built-in constant, you can right-click any identifier¬†(at the declaration, or any of its references) and¬†Find all references will give it to you, in a convenient tabbed¬†search results¬†toolwindow:


Double-click any result to navigate there.

Find all implementations

Similar to¬†find all references¬†(its results use the same toolwindow), this one is also one of my favorite ReSharper features, that Rubberduck simply¬†had¬†to implement. It’s only useful when you’re coding against abstractions and implementing interfaces (if you didn’t know… yes, VBA code can do that!) – but then, it’s the best way of navigating to implementations of an interface class or member.

For example, here I added two new class modules, added this line in each, and then implemented the members:

Implements ProgressIndicator

After refreshing the parser state, I can right-click the¬†Execute method in my¬†ProgressIndicator¬†class, select “Find all implementations”, and get this:


TODO Explorer

Rubberduck can (well, does actually) spot special markers in comments, and lets you navigate them in a dockable toolwindow Рagain, double-click navigates there:


Take that, bookmarks! You can group them by marker or by location.

By default, Rubberduck will mark NOTE, TODO and BUG as interesting, but you can always configure it to whatever suits your needs in the Todo Settings tab of the settings dialog:


Regex Search & Replace

Okay, that one’s not really there yet. But it’s totally on the roadmap, and definitely coming in a future version of Rubberduck. Take that,¬†search with pattern!

Whew! That covers Rubberduck’s navigation features. What do you think?

To be continued…

Preview Build is Here!

We couldn’t hold it much longer. (was about time!)

So we issued a preview build.¬†Keep in mind that this is a¬†preview¬†build –¬†we know it’s not completed, there are little glitches and issues left and right, things to polish a bit more, it¬†may¬†hang or crash your host application.

And it’s missing features: the¬†Code Explorer¬†(and its “virtual folders”) aren’t part of this release – that’s¬†really¬†not ready.

The Smart Indenter port¬†is there though, and although the preview box doesn’t show it in the settings dialog, it works pretty well.

IDE-Integrated Git Source Control is there too, and works nicely as well.

A brand new shiny WPF UI for all docked toolwindows, a new settings dialog, “why?” rationale¬†for¬†every inspection result, “fix ’em all” across the project/module, or disable inspections with a single click.

Our parser/resolver is much more powerful now – all known parser issues (as of v1.4.3) are now fixed, and new ones you never encountered too.

Unit testing works in AutoCAD and CorelDRAW now, and although the settings have no effect, the API is there and you can use a new PermissiveAssertClass that evaluates Equals in a less strict type-safe way, to work more like VB6 would do it.

There’s a lot to say and a lot will be said I’m sure.

Poke around, play with it – by all means,¬†break it –¬†and if you¬†do¬†break it, make sure you let us know how, so we can fix it by the time we ship the “real” 2.0 build.

Apparently there’s¬†still¬†a few¬†inspection¬†false positives (although it should be much better than in 1.x), so make sure you double-check before¬†you¬†fix all occurrences in project.


To be continued…

How we solved the most annoying COM Interop issue

I first noticed this strange behavior when I refactored Rubberduck’s menu system last summer: although I was 100% certain that there only ever was a single instance of a given menu item, in the Click event handler the¬†sender object’s¬†GetHashCode method returned a different value every time the handler ran.

GetHashCode, in the “normal” .net world, is tightly related to the¬†Equals implementation – whenever you override¬†Equals, you¬†must also override¬†GetHashCode. The value returned by this method is used by data structures like¬†Dictionary and¬†HashSet to determine some kind of “object ID” – the rules are simple:

  • if two things are equal (Equals(...) == true) then they must return the same value for GetHashCode()
  • if the GetHashCode() is equal, it is not necessary for them to be the same; this is a collision, and Equals will be called to see if it is a real equality or not.


In the “normal” .net world, it’s usually safe to assume that an object’s¬†hash code doesn’t change throughout the lifetime of the object – because a correct implementation relies on immutable data.

Apparently COM Interop has different rules.

When Rubberduck parses the projects in the VBE, it generates a plethora of Declaration¬†objects – thousands of them. There’s a¬†Declaration object not only for every declared variable, procedure, property, parameter, constant or enum member, but also one for every¬†VBComponent, and one for each¬†VBProject – anything that has an identifier that can appear in code has a¬†Declaration object for it. Declarations are related to each others, so any given¬†Declaration instance knows which¬†Declaration is its “parent”. For example, a module-level variable has the declaration for the module as its parent, and the declaration for that module has the declaration for the project as its parent.

On the first pass, there’s no problem: we’re just collecting new data.

Problems start when a module was modified, and is now being re-parsed. The parser state already has hundreds of declarations for that module, and they need to be replaced, because they’re immutable. And to be replaced, they need to be identified as declarations that belong under the module we’re re-parsing.

A module’s name must be unique within a project –¬†we can’t just say “remove all existing declarations under Module1”, because “Module1” in and by itself isn’t enough to unambiguously qualify the identifier. We can’t even say “remove all existing declarations under Project1.Module1”, because the VBE has no problem at all with having two distinct projects with the same name.

In Rubberduck 1.x we used object equality at project level: if a declaration’s¬†Project was the same reference as the¬†VBProject instance we were processing, then it had to be the same object. Right? Wrong.

And this is how we got stumped.

We couldn’t use a project’s¬†FileName, because merely¬†accessing¬†the property when the project is unsaved, throws a beautiful COMException – and we could be facing 5 different unsaved projects with the same name, and we needed a way to tell which project that modified “Sheet1” module belonged under. We couldn’t use a project’s¬†hash code, because it was now known to be unreliable. We couldn’t use… we couldn’t use anything.

This COM Interop issue was threatening the entire Rubberduck project, and shattered¬†our hopes of one day coming up with an efficient way of mapping a parse tree and a bunch of¬†Declaration objects to a¬†VBComponent instance: we were condemmned to constantly scrap everything we¬†knew¬†hadn’t changed since the last parse, and re-process everything,¬†just to be 100% sure that the internal representation of the code matched with what the IDE actually had.

Enter @Comintern and his pure awesomeness:

What about hi-jacking one of the R/W properties that are never going to end up user facing? Like .HelpFile? Just copy the original hashcode to the property and then search for it.


Rubberduck 2.0 will hijack the VBProject.HelpFile property, and use it to store a ProjectId value that uniquely identifies every project in the IDE.

Problem solved! Nobody ever writes anything to that property, right?

Stay tuned, we’re just about to announce something very, very cool =)

A Reflection on VBA Reflection

The idea has always been floating around, and a number of feature requests had been made to support something like it.

Not to mention all the Stack Overflow questions asking how to iterate the members of a module, with answers explaining how to use the VBIDE API to achieve some level of “reflection”.

The VBIDE API works: it gives you the members of a module rather easily. But if you need anything more granular, like iterating the local variables in¬†a given member, you’ll have to write code to manually parse that member’s code, and if you’re trying to programmatically access all references to a global variable, you’re in for lots of “fun”.

Truth is, the VBIDE API is bare-bones, and if you want to do anything remotely sophisticated with it, you basically need to write your own parser.

Rubberduck 2.0 will expose [some of] its guts to VBA clients, through COM reflection – using the¬†ParserState¬†API, you’ll be able to write code like this:


..and leverage the richness of Rubberduck’s own API to iterate not only module members, but any¬†declaration¬†Rubberduck is aware of – that includes everything from modules and their members, down to variables, constants, events,¬†Declare¬†statements, and even¬†line labels. You can also iterate the¬†references¬†of any declaration, so in theory you could implement your own code inspections with VBA and, in conjunction with the VBIDE API, you could even implement your own¬†refactorings

This API can easily grow TONS of features, so the initial 2.0 release will only include a minimalist, basic functionality.

One can easily imagine that this API will eventually enable surgically parsing a given specific class module, and then iterating its public properties to create form controls for each one, e.g. a Boolean property maps to a checkbox, a String property maps to a textbox, etc. Рtakes dynamic UI generation to a whole new level.

What will you do with that power?