VBA Rubberducking

Comintern

12:00 AM

Accounting for the annotation is a one-liner I think.

Duga

RELOAD!

[rubberduck-vba/Rubberduck] 3 commits. 3 opened issues. 1 closed issue. 1 issue comment. 3 additions. 3 deletions.

this

hmm but in this case, it stayed visible the whole time but lost icons anyway

Comintern

Or "make one line longer" would probably be more accurate.

Is it reverting when you select it afterward?

Duga

[rubberduck-vba/Rubberduck] bclothier added label difficulty-01-duckling to issue #4759: Interface nodes seem to lose their icons

[rubberduck-vba/Rubberduck] bclothier added label bug to issue #4759: Interface nodes seem to lose their icons

[rubberduck-vba/Rubberduck] bclothier added label user-interface to issue #4759: Interface nodes seem to lose their icons

[rubberduck-vba/Rubberduck] bclothier added label feature-code-explorer to issue #4759: Interface nodes seem to lose their icons

[rubberduck-vba/Rubberduck] bclothier opened issue #4759: Interface nodes seem to lose their icons

> **Rubberduck version information**
The info below can be copy-paste-completed from the first lines of Rubberduck's Log or the About box:

Version 2.4.0.14335
OS: Microsoft Windows NT 10.0.17763.0, x64
Host Product: Microsoft Office x64
Host Version: 16.0.11231.20120
Host Executable: MSACCESS.EXE

**Description**

It appears that when adding modules to a project can cause interfaces to temporarily lose their interface icons.

**To Reproduce**

1) open a folder that contains an in

this

@Comintern no, must expand it to revert.

Duga

12:04 AM

[rubberduck-vba/Rubberduck] comintern assigned comintern to issue #4759: Interface nodes seem to lose their icons

Comintern

Yep, debugger doesn't hit the breakpoint if the node is collapsed.

this

confused but it's not?

or did you mean the node for the interface itself?

Just tested w/ the interface's children expanded. Still loses the icon.

Comintern

It's a timing problem. In my test project, IsInterface is false when the tree refreshes.

Needs one of two things. Either defer rendering the TreeView until the parser state is ready, or do a second pass.

Mathieu Guindon

Would a second pass be weird? I think it would be preferred, but can already hear users' eyebrows raising

this

why can't it remember the icons it had?

Comintern

12:14 AM

@this Because it's a binding on the Declaration, not on the VM directly.

this

I guess the correct fix would be to maek Declaration raise events to VM.

BRB, going to get me a barrel of monkeys to go w/ that can of worms...

Comintern

I really don't like the idea of Declarations as event sources. I can hear the sucking sound of the memory leaks from here.

this

and now for something completely different... I'm a bit stumped on what is the correct architecture if I were to use MVVM patterns with a wizard-style dialog. Do I create a MVVM for each page, or do I use same model for all pages (which each will have their own VM)?

Comintern

@MathieuGuindon Anything that's annotated could get picked up in the first pass, so it would only effect interfaces that weren't tagged. It's already doing 2 passes to update the reference usages.

Mathieu Guindon

that seems legit

Q: why are icons changing? A: add an @Interface annotation!

this

12:19 AM

but... I had it already....

Comintern

The main reason it's working in 2 passes now is that it starts loading the Tree before references are resolved (which is apparently when the _subtypes are added).

I think that's legit. Especially if we add an inspection for InterfaceWithoutAnnotation.

Mathieu Guindon

^

this

I'm wondering if I'm missing something completely obvious because I always had the annotation....

Comintern

@this In WinForms, I'd have the pages serve up their own controls.

@this The annotation isn't currently being checked. It's a hard "is this class implemented anywhere" check.

this

The controls aren't really the issue. It's more of how to handle the data that's being collected from each step. I don't want to tie it to a specific model that might change.

One model that encompasses all the steps might work but then that makes it unwieldy to work with each page's VM.

whereas a model per page means I have to know something about each model when we are ready to build at the end of the process.

@Comintern :blinded by the obviousness:

Comintern

12:25 AM

@this Isn't a wizard just a structure way of collecting a common set of data? That makes me lean toward 1 VM.

Maybe with an Enum of WizardSteps that's bound to panel visibility.

this

I initially was going toward that but then thought that each page should have its own VM since it may have additional functions that are specific to this page.

BTW, I'm refactoring this wizard process - it's originally one god object

with smart UI to boot

Comintern

The main VM could simply act as an aggregator for the steps though.

this

Hmmm.

There's other thing - not all processes are required. There's conditions, so we may not end up visiting all the pages

I kind of can see an aggregator. I guess it means that each model for each page that do get visited would need a common interface for the aggregator to work with.

Comintern

@this Add visited pages to a collection.

You perform the aggregation of the wizard state by examining all steps that have been completed. Kind of like a builder.

this

yeah, that's what I was thinking - but the collection probably should be a collection of interfaces, so that way the aggregator doesn't have to know the details of each page's data.

Yeah. I think I do want a builder pattern here

Comintern

12:30 AM

That sounds entirely reasonable.

So Inspection Results. Start the expanders in collapsed state? That takes advantage of the group virtualization and lets each group render lazily. The flip side is more clicking.

this

so in short, each page gets their own model + VM. The model implements a common interface that the aggregator will accept, then at end, aggregator builds using the interface., done.

Alright, thanks!

IMO, I would prefer collapsed.

Because when I'm inspecting, I will want to see which inspections is causing me grief and act accordingly.

Comintern

Makes sense. If the IR window is docked, it kind of wants to be short and wide.

this

there is an exception, though.

If I'm using location grouping

e.g. I'm looking for problems in a particular module, I probably want them expanded.

Comintern

There's not really a non-trivial way to have a single expander expanded if that's where you're headed. Not when they're built dynamically.

That might possibly be more code than the rest of the VM combined.

this

wowza.

Comintern

12:34 AM

OK, gross exaggeration there.

Definitely not easy though.

this

Given the circumstances, I think a initial collapsed state is a preferable default over the "show me everything - oh wait, I can't read it all!"

because it's not often I get only 1 or 2 results.

I get in ballparks of 100s. :(

Comintern

OK, we'll go with that for now then.

Hosch250

1:11 AM

@this Yeah, bad joke. I knew that--I've used them a few times for playing audio.

@IvenBach Have you ever nuked your NuGet cache?

IvenBach

1:51 AM

@Hosch250 A couple times, along with the RD install folder.

Hosch250

@IvenBach The one in your user directory?

IvenBach

C:\Users\ivenbach\source\repos\Rubberduck and where the Nuget packages are.

Comintern

2:06 AM

Re #4028 - would right-click selecting be weird in the inspection results? The quick fixes are determined by the selected item, so there are a couple "dead zones" where you can get a context menu for an item half way across the window.

Tweeting Duck

2:28 AM

“Good code is its own best documentation. As you're about to add a comment, ask yourself, 'How can I improve the code so that this comment isn't needed?' Improve the code and then document it to make it even clearer.” - Steve McConnell

Tweeted by CodeWisdom on January 29, 2019 at 6:00 PM

2

IvenBach

2:48 AM

The only comment should be why it's done that way.

#BecauseIKnewNoBetterWayBackThen is no longer an excuse.

this

3:04 AM

@Comintern You mean the standard context menu (copy, paste, etc?)

Comintern

No - they're quick-fixes for the selected item.

Although I could toss copy in there I guess. Paste wouldn't work.

this

@TweetingDuck I understand where this is going but I agree w/ @IvenBach - code can only tell me what they do but not the why which usually is very important and not always readily deducted from the does within the code at all times.

@Comintern Then I guess I didn't follow the question. I visualized the menu as acting on the entire row.

Duga

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/ExtractInterface/ExtractInterfaceModel.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/Rename/RenameRefactoring.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

Comintern

@this It's not the menus. Gimme a sec and I'll help you visualize it.

Duga

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/RemoveParameters/RemoveParametersModel.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.CodeAnalysis/QuickFixes/EncapsulateFieldQuickFix.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldModel.cs of pull request #4072: Refactoring dialog refactor

Comintern

3:08 AM

^ Right clicking on an item gives a context menu of the available quick-fixes for it.

At least that's how I understood the issue.

this

That makes perfect sense to me.

My point was that the menu should work anywhere on the row

Mathieu Guindon

Yup. Could Ignore Once get an icon?

Comintern

The problem is that the stupid grid has dead spots.

this

even if I've right-clicked the other columns.

(is that what you meant by dead spots?)

or is the whitespace a dead spot?

Comintern

@MathieuGuindon Yep, I haven't added any icons yet.

Mathieu Guindon

3:10 AM

I understand dead spots as 1-2px areas that are clickable but aren't part of an item

Comintern

I think the dead spots are mainly on where the grid lines would be if they were shown.

this

Round the borders? not sure that's worth worrying over?

Mathieu Guindon

I don't see it as a problem if you get no ctx menu for rt-clicking between two items...

this

^

Comintern

Well, you can have the top item selected, then "miss" on the bottom item and get a context menu for something unexpected.

OK. I'll see if I can get the order right on the selection.

this

3:13 AM

Yeah, I'd rather get no context menu if this would be ambiguous

Comintern

So, right-click drill down to select, plus enabled binding on the context menu to the selected item (not null).

this

since no ctx menu => STOP FLIP FLOPPING AND MAKE UP YOUR DARNED MIND!

should work, I think.

Mathieu Guindon

@Comintern such a click should probably make the SelectedItem null

> Rubberduck.Setup.2.4.0.4488.exe (4.03 MiB) - downloaded 162 times. Last updated on 2019-01-28

Comintern

I think that's the default click behavior. If not I'll see what I can do.

Nope. It's tweaky. It seems to want the click dead center on the row. I'll see if I can fix that too.

Duga

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.Refactorings/Rename/RenameRefactoring.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.CodeAnalysis/QuickFixes/EncapsulateFieldQuickFix.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.CodeAnalysis/QuickFixes/EncapsulateFieldQuickFix.cs of pull request #4072: Refactoring dialog refactor

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

Comintern

3:27 AM

@Duga Damn you CW!

this

at one point i did considering writing the implementation for the factory

but the generics....

gonna pick yer poison.

Duga

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

this

@Comintern just so I'm clear -

This isn't static, right?

Comintern

Oh wait, did I misread that?

Yep. The diff view threw me off there.

this

ok, NP. I do not trust myself with things that are ~~WTF~~ WPF so wanted to be sure I wasn't missing something.

Comintern

3:33 AM

It looked like this:

        private static readonly DeclarationType[] ModuleTypes =
        {
            InterfaceName = $"I{TargetDeclaration.IdentifierName}";

            Members = declarations.Where(item => item.ProjectId == TargetDeclaration.ProjectId
            Members = new ObservableCollection<InterfaceMember>(declarations.Where(item =>

Duga

[rubberduck-vba/Rubberduck] bclothier commented on Rubberduck.Refactorings/RemoveParameters/RemoveParametersModel.cs of pull request #4072: Refactoring dialog refactor

this

oh yeah I see what you mean

Duga

[rubberduck-vba/Rubberduck] comintern commented on Rubberduck.Refactorings/RemoveParameters/RemoveParametersModel.cs of pull request #4072: Refactoring dialog refactor

this

@Comintern thanks for reviewing and feedback. I hope the rearchitecture makes sense.

Comintern

NP, and it makes a ton of sense.

this

3:38 AM

just to be fair, Vogel helped add some of the sense to the nonsense I may have introduced. ;)

Comintern

Nice.

That sentence gave me an idea for my BrainFuck IDE. I should have IntelliNonsense.

this

@Duga based on this, I am not sure it's always true everywhere but I can double-check. I'd rather be explicit at the construction, though.

lol

Since 4072 needs just a bit more work, we could just go ahead and merge the other 2 PRs?

probably start with Max's members attributes PR?

Comintern

ctor overload or argument works too I guess - it's just that if we're setting a specific state in a ton of places, it seems like the pattern should just be explicit.

Or implied (which was my first instinct).

this

Yeah - TBH, I can't remember why I didn't do that originally.

Comintern

I'm also go on the member attributes.

And the persistence service.

this

3:47 AM

I just checked - Max might want to address some of review comments first

Comintern

Oh FFS. Can you fix the spelling of "Persistance"? That's my long standing mistake...

Duga

[rubberduck-vba/Rubberduck] bclothier created comment on pull request #4686: Recovering member attributes

> To confirm, are you still seeing exceptions on exit as originally reported in the original post?

this

@Comintern Yes I can do that, though not tonight. :(

Comintern

Oh, doesn't have to be tonight. I can do that in a separate PR too. It's been wrong for going on 5 years, another year isn't going to hurt.

this

lol

Comintern

3:52 AM

Loving the VS spell checker, BTW.

this

yeah, earlier I used RD's rename and renamed my variable to Hightlight.

Then realized that I needed a spell checker.

Comintern

Hmmm... RD spellchecker...

this

the trouble is that the spellchecker needs to be more like VS'

the one that you get with Word or whatever won't work very well

Comintern

There are some NuGet packages.

Duga

8:26 AM

[rubberduck-vba/Rubberduck] MDoerner commented on Rubberduck.Refactorings/EncapsulateField/EncapsulateFieldRefactoring.cs of pull request #4072: Refactoring dialog refactor

Duga

8:56 AM

[rubberduck-vba/Rubberduck] Vogel612 edited issue #4759: Interface nodes seem to lose their icons

Vogel612

@this I could, but I won't be able to get to it before the end of February.

Also I really need to get a bit of a handle on cake before rewriting deployment, so there's that

Peter M Taylor

10:26 AM

Spotted some vba documentation on GitHub managed by MS. I note many Stack overflow labels.. Are they checking if used elsewhere? github.com/MicrosoftDocs/VBA-Docs/issues

FreeMan

12:31 PM

@IvenBach #BecauseICan'tFindABetterWayRightNow can be when you're under pressure #ItsUglyButItWorksForNow #TechDebt

@Comintern Yes!!!!

FreeMan

12:44 PM

monking, @all! -9°F and about 30MPH winds today. Lovely!

Mathieu Guindon

@PeterMTaylor they use the SO label for issues where the OP is asking about their broken code rather than providing feedback on the documentation

In other news, I finally got my SuggestionBox bindings to work.

M.Doerner

1:21 PM

A spellchecker might be nice, but not before our own code pane(s).

FreeMan

^ yup!

Comintern

^^

Feeds

1:47 PM

CommitStrip: Anti-freeze

posted on January 30, 2019 by CommitStrip

Tweeting Duck

SECURITY PEOPLE: You know how we make fun of users for not having “basic” security knowledge and blame them for not being interested in computers? This is what we sound like to everybody else. Like militant duck enthusiasts.

Tweeted by SwiftOnSecurity on January 30, 2019 at 4:55 AM

Hosch250

2:02 PM

@Feeds LOL, considering the temperatures...

this

@FreeMan Maybe consider moving there? It's warmer there:

accuweather.com/en/aq/amundsen-scott-south-pole-station/2258520/…

Hosch250

@FreeMan It was -27 when I left for work.

My car barely started. It turned the engine 3 times, almost started 3 times, and then turned 3 more times before it started.

It was so cold that running the heater dropped the engine temperature by half on the scale thing.

FreeMan

Yes, the local news has been running stories about how it's colder here than Anchorage and the antarctic. Just wonderful... I'm warmer than @Hosch250, though, so I can't complain too much.

Hosch250

I turned it off shortly before I got to work to run the engine up to its normal temperature so hopefully have it slightly warmer when I leave, but I doubt it'll do much.

FreeMan

@Feeds the plumber's crack/crotch frame was... uncomfortable...

Hosch250

2:09 PM

Apparently parts of north dakota are colder than even the north pole.

@TweetingDuck I'm not a duck fan, but if I found ducks were surreptitiously taking nude pictures of me, I'd sure as heck do my research on them.

I'd buy a hunting license too. Pity that doesn't work with some companies I'm thinking of.

Duga

2:35 PM

[rubberduck-vba/Rubberduck] MDoerner created comment on pull request #4686: Recovering member attributes

> I have not seen the exceptions for a while now. I guess they were related to some other problem fixed in the meantime.

I will address the review comments tonight and do a test run under the debugger again.

BlueEel

What is the difference between the LL parser and the SLL parser? - The reason I'm asking is we have a VB6 code base with extensive use of the bang notation and i'm experiencing the issue described in github.com/rubberduck-vba/Rubberduck/issues/4405

this

I don't know much but I understand that LL allows backtracking whereas SLL doesn't.

Hosch250

I probably know less than @this, but the way I view it is the SLL is more eager.

this

By backtracking, it means that when it is present with several possible alternatives, it will pick one, try to match the rest of tokens according to the alternative branch.

Hosch250

It does a best-guess, and if it doesn't match, it has to try the slower, more accurate way.

this

2:44 PM

if it fails, then it will go back up, and try the other alternative.

@Hosch250 That's actually us retrying -- SLL will simply faceplant, so we catch the error and then re-parse using LL.

Hosch250

Perhaps by looking at the first child token and saying "it's most likely to be X path".

@this OK.

this

Yeah something like that.

BlueEel

Is there any way to disable the SLL parser?

this

Anyway, the bang notation is a problem more because we need to know the context about the parent that isn't interred from the parsing alone.

So I don't think disabling SLL parser helps us.

Hosch250

The SLL parser probably goes so fast it wouldn't make a difference.

It can parse a module in about 20 milliseconds.

this

2:47 PM

Lacking a counterexample, I believe the uses of bang either fall into 2 categories -- implicit default member access or evaluate member (though I think that has less to do with the bang operator and more with the bracketed expression).

Hosch250

It's our resolving passes that take a long time, usually.

this

Therefore, expressions like Forms!myForm!myControl needs to be expanded into Forms.Item("myForm").Controls("myControl").Value

BlueEel

ok... so the reason my project takes 10 minutes to parse and then crashes VB6 is the LL parser and there is nothing I can do about it for now?

Hosch250

It shouldn't crash it.

Comintern

^

Hosch250

2:48 PM

It should just say "parser error" and the duck won't work.

this

but in order for us to know that, we need to know that the Item is a default member of the Forms interface, Controls is the default member of the Form interface, Value is the default member of the Control interface. That information we need to have in order to resolve the expression correctly. That information can't be had from parsing alone - we have to fill in the blanks from somewhere (e.g. type libraries)

@BlueEel I'm wondering - are you looking at the memory usage?

QHarr

\o

BlueEel

yup. the VB6.exe uses about 120mb with our project open - until i start parsing at it exceeds 1gb...

this

I think that is more likely the source of the crash

Mathieu Guindon

yikes.. how many modules and referenced type libraries are we looking at?

I bet the inspection results TW is hogging a good chunk of the resources all by itself

this

2:54 PM

@BlueEel if you are also willing, please share the logs & diagnostics at the issue #3347

BlueEel

about 225 forms, classes and modules

this

@MathieuGuindon It'll hog even if it's not initially visible? In my other Access project I mentioned few days, I hit 1 GB ceiling even though I don't have it visible.

It probably has around the same number of the objects.

Mathieu Guindon

@this yes

(AFAIK)

Comintern

1GB shouldn't be a problem for VB6.

this

Ok.

Hosch250

2:57 PM

I still think at least some windows should lazy load so we can just say "don't use X feature until we fix this.'

Comintern

I'd suspect that it has to do with something on one of the forms. Might be better as its own issue depending on the logs.

Mathieu Guindon

1GB where 85% is our stuff isn't really ok though

this

@Comintern Unless it has same weird arbitrary ceiling that I was talking about with Access? I honestly have no idea how they did memory management back then in '97....

^

Comintern

I'm not saying it's OK, but it shouldn't cause VB6 to crash.

VB6 should be closer in design to VS6 than it is to the VBE, and I know VS6 doesn't hit a memory cap at 1GB.

this

GTK.

Like I said, I absolutely have no idea how they did things back then.

I remember back in the old days System 6 had a settings for all applications where you could manually set the memory minimum and maximum. I guess sizing wasn't a thing back then.

BlueEel

3:06 PM

I don't know what's going on - now it doesn't crash :-S

2019-01-30 16:04:14.1680;DEBUG-2.4.0.4488;Rubberduck.Parsing.VBA.ParseCoordinator;Parsing run finished after 736,1032026s. (thread 12).;

Comintern

Oh, that's a decimal and not a thousands separator in 736,1032026s. Whew!

BlueEel

I'll see if I can upload some logs tomorrow - it's closing time.

Thanks!

this

that would be more than 10 minutes of just parsing. We probably haven't started inspecting yet.

(right?)

BlueEel

right

this

tomorrow, verify if crashing comes from inspecting.

thanks so much for coming and sharing! We appreciate it!

BlueEel

3:09 PM

it's a fairly big project - but I think it's the bang notation that's hogging it

this

Right - that'd be another thing - maybe screenshots of how parser is resolving those expressions would be good, too.

Mathieu Guindon

that's great feedback, thanks for sharing this!

this

Give us an idea of how parser is resolving and whether it's doing it correctly or incorrectly.

I know that bang notations and bracketed expressions has been problematic for the parser.

Mathieu Guindon

"screenshots of how parser is resolving those", you mean the "status bar"? helpful indeed, but rather hard if the VBE crashes ;-)

Comintern

Bang notations and bracketed expressions has been problematic for VBA in general...

Mathieu Guindon

3:13 PM

@BlueEel there's a checkbox in the inspection settings, checked by default, that you can uncheck to make inspections only run when you request a "refresh" from the inspection results toolwindow itself

this

@MathieuGuindon Yes. That was for the case where he did get it parsed without crashing.

Mathieu Guindon

won't help the bang notation, but should help identifying whether inspection results are contributing to that 1GB

this

@Comintern O RLY?

Comintern

[O!RLY]

Mathieu Guindon

I wonder if the Antlr guy could help us with getting bang notation to parse with SLL

this

3:16 PM

hm. Why do we need to use LL?

Mathieu Guindon

because we want that parse tree??

Comintern

The bracketed expressions seem like they'd be more difficult IMO.

this

No, I mean, what I said earlier was that we have no information about the context as part of parsing, so that is irrelevant to parsing?

that'd have to come in later as part of resolving.... right?

Comintern

Is the issue with bang notation that you can use keywords after? Foo!Input or Foo!If?

Mathieu Guindon

my understanding is that the long-running "parsing run" doesn't involve the resolver passes yet

@Comintern could be. ...no idea actually.

IIUC in order to understand what makes SLL fail, you have to think like Antlr does

this

3:20 PM

@Comintern kind of implies that the part after the bang should e treated as a string literal

though I think that you'd have to bracket if there's whitespace Foo![If Input]

Mathieu Guindon

on the bright side, the parse is completing. with the original VB6 grammar from 2015, it would have blown up well before that

this

^

Comintern

There is that.

this

so to my mind, bang notation should mean that the next token is an identifier, not a keyword.

Mathieu Guindon

yeah. I guess it's somehow ambiguous though, otherwise SLL would just zoom through.

this

3:23 PM

Right. woud have to look at the parse tree.

Anyone else want to fire up that eclipse?!

:D

Mathieu Guindon

lol

this

I probably have it somewhere but I think the .g4 files are outdated. :(

even more fun considering we invoke some .NET functions now

Mathieu Guindon

thank line continuations for that

M.Doerner

@BlueEel @this Let me give some background about the difference between the LL and SLL parsing mode and why the bang notation is a problem.

In each step, the SLL parser uses a finite context free automaton to determine which alternatives are viable. It then takes the first option found.

I am not completely sure about how the LL parser works. However, I think it back tracks and reverts previous decisions on subrules when it fails to resolve a grammar rule.

The problem with the bang notation is that the character ! is a type character in VBA.

Mathieu Guindon

ooooh

well there's the ambiguity

M.Doerner

3:36 PM

So, when the SLL parser descends into the rule for the identifier on the left, it consumes that character as part of the identifier.

Mathieu Guindon

do we define type hints as "must be followed by a space"?

this

Won't work - ?TypeName(1!)

Mathieu Guindon

ugh

M.Doerner

Since it is context free, it simply does not know that it will need that character in the rule one level higher.

It is the type character for Single.

this

To be clear, we're eating it by single character each? We are not eating it per token?

M.Doerner

3:40 PM

That character is its own token.

We eat it token by token.

this

Alright.

M.Doerner

@this For me ?TypeName(1!) returns Single in Excel's VBE's immediate window.

this

hmm. ?Forms! frmMain.Name does not seem to be legal. I get a compile error.

@M.Doerner Yes, same here - I actually typed that before Mat's question to find out what it was.

FreeMan

It's 0.4°F on my front porch now. #GettinToasty

M.Doerner

Ah, I missed the context.

this

3:44 PM

I assume we did already try to constraint the type characters to require either whitespace or end of line or non-identifier characters like )?

Mathieu Guindon

not sure we did..

this

After all, we would get a compile error if we tried a&b; it'd have to be a & b.

but...

?a&*b
 0

M.Doerner

So, the idea would be either to introduce a predicate checking that the next character is not a legal start character for an identifier or, maybe better, to check that the following token is not an identifier.

this

I think that seems reasonable.

M.Doerner

I prefer the token version. It will be simpler to implement.

this

3:54 PM

does predicate checking require use of .NET expressions?

M.Doerner

What exactly do you mean?

It requires .Net code, but no regex.

this

Start over. How do you implement a predicate check in a .g4 file?

M.Doerner

{predicate code}

this

like this? singleLetter : {_input.Lt(1).Text.Length == 1 && Regex.Match(_input.Lt(1).Text, @"[a-zA-Z]").Success}? IDENTIFIER;

M.Doerner

Yes

this

3:57 PM

so it is going to use .NET code in order to discover the next token.

Comintern

Is this correct?

lExpression mandatoryLineContinuation? EXCLAMATIONPOINT mandatoryLineContinuation? unrestrictedIdentifier # dictionaryAccessExpr

I didn't think you could use a line continuation there.

M.Doerner

I honestly don't know.

this

i just checked it doesn't seem to work

M.Doerner

Getting the next token is the _input.Lt(1).

Comintern

I don't think it works in either position.

this

4:00 PM

this is after prettifying:

Debug.Print Forms! _
          ; frmMain.Name

M.Doerner

I think La(1) provides the token type identifier, which probably more appropriate here.

Comintern

This is what 5.6.14 says, but I don't think that works in practice:

> dictionary-access-expression = l-expression NO-WS "!" NO-WS unrestricted-name dictionary-access-expression =/ l-expression LINE-CONTINUATION "!" NO-WS unrestricted-name dictionary-access-expression =/ l-expression LINE-CONTINUATION "!" LINE-CONTINUATION unrestricted-name

KySoto

have you guys ever heard of Hawaiian rolls?

this

Public Sub t()
    Dim s As String


    s = forms! _
        frmMain.Name

End Sub

that is a compile error

KySoto

i ask the question that way cuz my co worker... has never heard of hawaiian rolls

Comintern

4:03 PM

@KySoto Yes. It's kind of like a sweet bread, more like a form of challah that isn't woven.

KySoto

i was making sure i wasnt crazy for expecting him to know what they are

M.Doerner

To be honest, I did not know.

Hosch250

@KySoto TIL.

KySoto

-_-

Hosch250

And today he learns too.

KySoto

4:04 PM

@M.Doerner are you in north america?

M.Doerner

But I am german.

KySoto

ahh

yeah so that makes sense

i feel like its a bit of an american thing

Hosch250

The next step is to say "today is your lucky day!" and take him to lunch with Hawaiian rolls.

KySoto

might be in canada

i think it would be better to have my dad bake n butter some up

this

uh, when I hear hawaiian rolls, I think of those gross white fluffy thing they sell at your local grocery's bakery department.

Mathieu Guindon

4:10 PM

@Comintern depends if mandatoryLineContinuation includes a WS

@this #TIL

this

@MathieuGuindon do you have an example supporting this?

FreeMan

@KySoto they're delish!

KySoto

+1!

they made me hungrier

Mathieu Guindon

@this uhm, not knowing that foo! _\r\nBar was illegal

KySoto

i had 2 this morning to combat nausea from a migraine... and then nausea from the migraine meds

this

4:13 PM

just making sure there wasn't a counter example that I didn't think of.

Comintern

mandatoryLineContinuation : LINE_CONTINUATION WS*;, but IIR doesn't the VBE prettify it to always include the space?

Mathieu Guindon

no, it chokes and claims a syntax error

this

^

VBE is kind of stupid when it comes to those single character tokens.

M.Doerner

4:56 PM

Line continuations always have to include a leading space.

The leading whitespace gets consumed by the token.

Comintern

What do you mean by "self reference"? When then project is loaded it's implicitly in scope within that project. The worksheet document objects belong to the project, not the workbook. — Comintern 16 secs ago

Any clue what OP is after there?

M.Doerner

There is a situation where this is necessary to allow using a sane predicate in some rule.

Transcript for