VBA Rubberducking

this

5:00 PM

use conventions

e.g. a plugin always gets its own entry with a submenu

KISS

Mathieu Guindon

so Rubberduck > SomePlugin > EveryCommand

this

ditto for hte Ce's context menu

Mathieu Guindon

makes sense.. Rubberduck > Inspections > Disable in this module

this

what are you thinking what should go into the extensibility framework? thinking better and simpler if each plugins were their own islands with only access to 1) ability to describe menus (as discussed above), 2) ability to describe toolwindows (e.g. implement the IDockableWindowHost and we do the rest), 3) ability to see the parse tree and state and the SCWs? They don't need ot see the CE or the menu system or the addin framework itself... right?

Mathieu Guindon

correct

this

5:09 PM

The SCW part scares me the most, though.

Mathieu Guindon

I think the easiest way to figure out what plugins need, is to take some of the current functionality and make it a plugin

this

like inspections.

Mathieu Guindon

yeah

what's SCW?

this

safe com wrappers

Mathieu Guindon

oh

this

5:10 PM

hmm. but inspections should never interact with SCW.

only with the parse tree

correct

SC?

SC would

yeah it has to interact with the SCWs.. :(

TBF, i'd rather provide an abstract view

Mathieu Guindon

but proper usage can (should) be documented anyway

@this a facade on top of the facade? :)

this

5:11 PM

the problem is that if plugins are their own VS project... no analyzer.

aren't we already standing on 1000 facades already? What's another facade?

Mathieu Guindon

lol

@this that's a showstopper right there

a plugin that wouldn't cleanup their stuff could take us down badly

and we wouldn't have any control over it

this

exactly - and i realized --- they woudln't have to have AV and CodeCov

i definitely don't want to be the next Eclipse.

their plugin system is abysmal

Mathieu Guindon

so we've just killed plugins right there then

this

damn

and what does that leaves SC & PM then?

those too contribute to a good IDE....

Mathieu Guindon

I guess we gotta pick our battles

this

5:20 PM

TBH, I think PM is #WorthIt. Sharing VBA code is way too hard than it should, IMO/E

and if we don't share, then it's hard to help people raise the level of standard.

Dain Ironfoot III

SC and PM?

this

SC=Source Code --- PM =Package management

Dain Ironfoot III

Ah.

FWIW, I think we should pull our facades out into a separate project.

Then anyone can use them without going through the pain we did.

For example, PM could be a completely separate project. It needs nothing but the SCWs and the import/export functionality, right?

Mathieu Guindon

I guess

(and yeah, it would be a separate project anyway)

this

agreed RE: project

not sure I follow the facade part. put exactly what in a project of its own?

the SCWs?

Mathieu Guindon

5:26 PM

isn't that already in its own project? not sure I like making additional dll's for the heck of it

I'm already not too warm about Rubberduck.Resources.dll

this

yeah, VBEEditor

Dain Ironfoot III

FWIW, I don't like millions of projects either.

this

@MathieuGuindon sorry about that.

Mathieu Guindon

@this lol, that was my PR!

this

recall that it came because of my idea about separating the guid/progid because that was a cross-cutting concern.

mansellan

5:29 PM

duck check: the risk is in handing out SCWs, but without doing that a plugin wouldn't be able to do anything useful...

?

this

that said, when Core does lean down to just UI, I totally can see resx files going back. Not sure where that leaves the guid/progid

@mansellan no if they have a SCW and they don't clean it up properly... memory leak galore

right now, SCW is still fragile (e.g. chained member access => memory leak)

mansellan

yep that's what i meant

this

even if we did solve this, what about the code analyzer? AV? CodeCov?

a separate repo wouldn't have none of that.

mansellan

and there's no way do abstract it? "Tell us what you want and we'll find \ do it for you"

this

unless rubberduck-vba owns them but that means MORE management.

not sure how one abstracts the code quality checks....

mansellan

5:31 PM

@this if we can isolate properly, does it matter? worst case, the plugin is flakey but RD keeps on truckin'

(for 3rd party plugins i mean - any factored out from current RD would get proper CI)

this

Honest? I don't want a 3rd party plugin at this point.

maybe later later LATER but irght now, I'd prefer that the duck own it all

mansellan

yes agreed - so we could set that up with AV, analysis etc

Mathieu Guindon

^

this

and TBH, managing AV,etc. is big PITA

2nd repo doubles that.

so... I'm not ready for a 2nd repo. I doubt anyone else is.

mansellan

ok makes sense

this

5:37 PM

so if PM (and SC and whatever else) is going to happen, they'll have ot be a project withiin the Rubberduck repo.

again, recall we booted SC because it was causing so much pain

mansellan

but for the 1st party plugins, could we not keep those in the main repo but still using a plugin system

Mathieu Guindon

OTOH, what does SC/PM need? They don't need the whole wrapper API

this

not without not making it a plugin.

Mathieu Guindon

Just some API

mansellan

yep

this

5:39 PM

PM and SC mainly needs SCWs, basically. I don't think they give a hoot about the parse tree.

we can describe the menu and toolwindow like we discussed earlier

Mathieu Guindon

But if we expose an API that cleanly handles adding/removing/replacing modules, what do SC/PM need of SCWs?

this

ooooh

mansellan

not sure they need the SCWs... surely just a "here's a module, please import it" API (PM)

^^

this

that's definitely clean

keep their dirty grubby paws off the SCWs

mansellan

or a "Please let me know when a module goes dirty" API (SC)

#Simplifying

this

5:41 PM

exactly. so really a notification system + a facade (on top of facade ;-) ) to handle the import/export

mansellan

yeah

Mathieu Guindon

@mansellan that could be exposed via parser state in theory

this

hmm, and we can use parser to tell us what modules there are, too

mansellan

Rx maybe... events are problematic

this

but what about content hash for PM?

IvenBach

5:42 PM

Roslyn is the VS built-in equivalent of R#?

Mathieu Guindon

@IvenBach Roslyn is the compiler

IvenBach

1 hour ago, by Mathieu Guindon

long-term is to kill the inspection results dialog, and instead have R#/Roslyn-like popup-buttons in the code panes; makes sense to have the results as parse tree state then, no?

AFAIU we want red squiggly capabilities for code pane.

Mathieu Guindon

Yup

Or blue dotted underline... whatever

IvenBach

Roslyn being the compiler helps by compiling what is type in and then that is what get's parsed? I'm having a disconnect.

this

I think roslyn is more than just ocmpiler

it provides analysis and other stuff, too

Mathieu Guindon

5:46 PM

The compiler has multiple parts. There's the parser, and then a semantic pass and analyzers building on top of ot

this

so it's similar to R# but not equivalent (e.g. R# doesn't actually compile any code; it only analyze code like RD does but cannot compile)

IvenBach

Not "just" a compiler. It's able to do other things in addition to it.

Mathieu Guindon

R# can tell you if your code will build. Roslyn analyzers can make the build fail, but run as part of the build process

IvenBach

How can it make it fail but still run?

--/ /-- disconnect again

this

well it has to keep running to get the rest of code analyzed, no?

otherwise, you're back to VBA's nefarious "fix compile error one by one MWAHAHAHA!"

IvenBach

5:49 PM

VBA Says: "My code doesn't always compile, and when it doesn't, you get to fix the compile-time errors one by one."

this

(I absolutely hate that situation, never knowing how many compiler errors there are. Am I one typo away, or am I 10000000 bad stuff away? I WILL NEVER KNOW UNLESS I FIX THEM ONE BY AGONIZING ONE!!1!"

IvenBach

Nope, never been in than situation before. Not never.

this

it's worse when you inherit the codebase

if i wrote it, I at least can expect to ahve few compiler error and I'll be alright. But someone else's code... well....

Mathieu Guindon

@this just typos I guess ;-)

IvenBach

Not being able to test your own code's compilability because of someone else's work isn't pleasant.

this

5:58 PM

and there are lot of VBa projects out there that has never been "compiled" (explicitly via the menu or keyboard) for years.

they keep adding and adding and adding and adding and it's "it runs fine!"

Gaaaah

Mathieu Guindon

@IvenBach if the code doesn't compile, analyzers don't need/get to run. if the code compiles, analyzers can still send the build up in flames if they're specified to do that

Dain Ironfoot III

@MathieuGuindon Actually, not quite.

Analyzers get run as the parse tree is created.

Mathieu Guindon

but not too far off either

Dain Ironfoot III

Which is a heck of a lot more often than a compile happens.

Mathieu Guindon

if the parse tree isn't valid, what's the point of running analyzers?

Dain Ironfoot III

6:01 PM

Speed.

Mathieu Guindon

...

Dain Ironfoot III

And guess what?

They can run analyzers and all nodes but invalid ones.

this

running more code makes it go faster? confused

Dain Ironfoot III

@this Makes it look like it to the user.

Mathieu Guindon

@DainIronfootIII ah, right

this

6:03 PM

ok that makes sense. OH that reminds me. Didn't we enable that for RD, too?

Dain Ironfoot III

Well, I did a bunch of prelim work on it, but I broke most of the parser :(

So, I think it's rotting in a branch somewhere.

Our error system ATM relies entirely on getting one error at a time.

Mathieu Guindon

some 2000 commits behind lol

Dain Ironfoot III

Yeah, that was back at Christmas :)

TBH, it's not that hard--until the tower falls around your ears.

IvenBach

I get the idea. I just need to code more C# to eventually better understand all your discussions.

Dain Ironfoot III

@IvenBach Don't we all :)

IvenBach

6:05 PM

I'm lost on most of what gets discussed.

Dain Ironfoot III

@this Use Paket if you don't want NuGet.

Ah man, VS is going down hard.

> Paket is a dependency manager for .NET and mono projects, which is designed to work well with NuGet packages and also enables referencing files directly from Git repositories or any HTTP resource. It enables precise and predictable control over what packages the projects within your application reference.

this

IDK. Does it have a DLL for backend like ZeroInstall?

and also, it won't be .NET - it's VBA code sooo....

Dain Ironfoot III

Oh, you found something already? Carry on.

this

nuget package pretty assumes that it'll be .NET

Dain Ironfoot III

Yeah. It works with JS files, though, so...

this

6:17 PM

asking you to specify a framework version, a specific language, etc.

cos those are "content"

in theory we can just build a content-only package but IDK. Feel really wrong.

it'd be really annoying if the nuget package with just VBA code was listed in public nuget repository and they put it in VS and go "WTF"

Dain Ironfoot III

@this /shrug

Mathieu Guindon

who'd be searching for VBA nugets in VS though?

this

my point is that it wouldn't be filtered out

because we can't say "this ain't C#" (or .NET)

Mathieu Guindon

right. so the whole .net world would be like "hey wtf is going on with VBA these days?"

could be a good thing

"wait VBA has nuget packages now??!"

and every one of them would say "Rubberduck" somewhere

this

mindshare is a good thing.

TBF never heard of ZeroInstall (or Paket for that matter) until yesterday. Everyone knows what nuget is.

Mathieu Guindon

6:21 PM

^

I love the play with "nuggets" though

makes me wonder if the .net team @MS would love it as much

this

HOW DARE YOU POLLUTE OUR SACRED REPOSITORY WITH PROFANE CODE THAT IS VBA

Dain Ironfoot III

@MathieuGuindon They wouldn't care.

The legal team might, though.

Mathieu Guindon

who owns nugget.org?

Dain Ironfoot III

Nobody.

Nugget.com is registered with GoDaddy.

this

i'm amazed that it's not taken already

QHarr

6:30 PM

allo allo

this

just to point it out, tho--- if we end up using nuget package... why not just use nuget.org?

Mathieu Guindon

if we end up using nuget, of course we'll use nuget.org

...if they let us :)

Vogel612

@DainIronfootIII IIRC that was also before the upgrade to ANTLR 4.6

Dain Ironfoot III

@Vogel612 Twas.

Duga

8:18 PM

[rubberduck-vba/Rubberduck] tdalon created comment on issue #4009: Excel 2016: unexpected error

> If you need anytime to have a debugging Skype session with me just contact me (Matthieu got my email).

Duga

8:56 PM

[rubberduck-vba/Rubberduck] mansellan created comment on issue #4019: Rubberduck Nuggets (discussion)

> I agree that the ability to browse and easily download / install trusted collections of modules has great value of itself. But I'd like to think there was a pathway from there to being later able to implement managed dependencies, as that allows the community to build ever upwards (think of how many NuGets take Json.Net as a dependency, for example).

IMO, one of the issues with SC was the use of external files to track project state - this had implications for portability and shared develop

[rubberduck-vba/Rubberduck] bclothier created comment on pull request #3975: Reflection API & COM management

> Just in case anyone needs a quick guide to aid PR....

Most of files are simply reogranization and thus can be skipped.

1) COM stuff

[RubberduckGuid](https://github.com/rubberduck-vba/Rubberduck/pull/3975/files#diff-9fe778a6e26c2eed55788ad235bbcb15)
[RubberduckProgId](https://github.com/rubberduck-vba/Rubberduck/pull/3975/files#diff-5cb4aac2a636218103f5b621b8c796a7)
[ApiProvider](https://github.com/rubberduck-vba/Rubberduck/pull/3975/files#diff-1e37877fec6b8f8019b0f71fc0015960)
[Par

this

#MemoToMyself stop making mega-PRs. Nobody like them....

Vogel612

@this that's not only you there ...

Dain Ironfoot III

LOL. We've all had our share.

Especially when laying the groundwork for an architectural shift.

this

I know. I did not exactly plan that out well. "this is just a little clean up..... WTF where I am now?!?"

3

Mind, i'm pretty stoked with how it went, mainly because that solves a long standing problem with COM interop in .NET (e.g. renaming enums and customizing the type library).

Vogel612

Mar 14 at 17:50, by Vogel612

it's just that "while I was at it", I decided to do some more cleanup as well, and that's when everything went to hell and back

this

9:04 PM

LOL

but #worthit, I think

Vogel612

probably, yes

Duga

[rubberduck-vba/Rubberduck] mansellan added label has-workaround to issue #4008: VB6 COM context menus missing

Dain Ironfoot III

"... A scholarly father may be of some use in exploring ..." - even better, an old codger of a father in a ring might provide loads of snarky and acerbic comments :). — Edheldil 7 hours ago

Wow...

> lol, an extension method would do it. as for extension properties... we're not on C# 6, so even if you implement it and it ships with C# 8...

And turns out it is being implemented to ship with C# 8.

this

extension properties?

that feels .... wrong.

Dain Ironfoot III

Yes.

Extension anything.

Even static extension methods.

mansellan

9:11 PM

'bout time

Dain Ironfoot III

So you can do Type.Foo().

Vogel612

soo .... "define anything anywhere and let the compiler figure out what the heck to call"?

Dain Ironfoot III

Well, not quite. It uses the extension keyword.

mansellan

sure, why not?

Vogel612

@mansellan retraceability and cohesion

mansellan

9:12 PM

Meh

:-)

this

O HAI JAVASCRIPT! (wannabe)

Vogel612

I get that you can navigate to definition, but....

mansellan

I love extension methods, and it really grinds my gears that props and statics (especially) are not there.

this

but.... why?

you're making C# more like javascript

Dain Ironfoot III

github.com/dotnet/roslyn/issues/11159

mansellan

9:14 PM

trivial example - class has FirstName {get;}, LastName {get;}. Extension has FullName {get;}.

Dain Ironfoot III

@this Because you don't own hardly any code compared to what you use...

mansellan

and yeah, i know names #Depend, but 5s example

the static one is really annoying. end up with StringHelper, EnumHelper... ad naseum

Mathieu Guindon

TTQW

Dain Ironfoot III

TTQW.

mansellan

i think they finally got around to allowing "where T : Enum" too

Dain Ironfoot III

9:16 PM

@mansellan Yes, C# 7.3

Also, Delegate and unmanaged, IIRC.

mansellan

another nice one would be allowing params in the new() constraint. Must have a public ctor that looks like this

Dain Ironfoot III

There is talk about defining something they call "shapes" so you can specify ctors other than new and methods by signature, etc.

mansellan

hmm, ima vote on that

this

what does unmanaged does? (google gave me snow)

Dain Ironfoot III

@this unsafe stuff.

this

9:18 PM

sorry , let me rephrase -- what unmanaged has to offer over unsafe ?

Dain Ironfoot III

@this No, it's a generic constraint.

docs.microsoft.com/en-us/visualstudio/releasenotes/…

this

oh so I can make a generic out of C++ non-ref class then?

Dain Ironfoot III

Basically.

Generic constrains are getting powerful.

But, I don't know where the mentioned "shapes" are on the roadmap. They were going to be a TON of work.

Might be C# 9/10.

Duga

9:39 PM

[rubberduck-vba/Rubberduck] Vogel612 synchronized pull request #4006: Splice Off Rubberduck.Refactorings from Rubberduck.Core (WIP)

[rubberduck-vba/Rubberduck] Vogel612 edited pull request #4006: Splice Off Rubberduck.Refactorings from Rubberduck.Core

[rubberduck-vba/Rubberduck] build for commit 1709e3a4 on unknown branch: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] [codecov[bot]](github.com/apps/codecov) created comment on pull request #4006: Splice Off Rubberduck.Refactorings from Rubberduck.Core

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4006?src=pr&el=h1) Report
> Merging [#4006](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4006?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/f86f7be13d15c8ba7610c1684c4576ce43784598?src=pr&el=desc) will **decrease** coverage by `0.13%`.
> The diff coverage is `32.84%`.

```diff
@@ Coverage Diff @@
## next #4006 +/- ##
=======================

[rubberduck-vba/Rubberduck] build for commit 1709e3a4 on unknown branch: 53.54% (target 0%)

QHarr

10:10 PM

Anyone know why [^.] wrapping causes a regex to suddenly have an unclosed ) ?

[^.(0[1-9]|1[012])[- \/.](0[1-9]|[12][0-9]|3[01])[- \/.](19|20)[0-9]{2}]

pattern error of unclosed ) on first )

Where as
(0[1-9]|1[012])[- \/.](0[1-9]|[12][0-9]|3[01])[- \/.](19|20)[0-9]{2}

is fine.

Example string
xxx xxx 10/08/2008 xxxx

Used on https://regex101.com/

Vogel612

10:26 PM

@QHarr that seems like a bug in their regex-parser

if you put that through RDs RegexAssistant, does it exhibit the same problem?

QHarr

Ooh...I don't, yet, know how to do that.... new toy!

this

@M.Doerner just to quickly check -- I'm addressing your feedback on SuspendParser and already ahve moved the bulk of code to the ParseCoordinator -- would you agree it's appropriate to use _cancellationSyncObject to lock the access for thread safety?

(this same token is used to manage the cancellation though it's not used to start parsing; that happens outside the lock block).

this

11:13 PM

Also, in the RunAllTestCommand we have this:

private static readonly ParserState[] AllowedRunStates = { ParserState.ResolvedDeclarations, ParserState.ResolvingReferences, ParserState.Ready };

this smells because a command is saying what state we can run in, while the engine only look for Ready state.

Duga

11:42 PM

[rubberduck-vba/Rubberduck] bclothier synchronized pull request #3956: Introducing Busy state and SuspendParser action

[rubberduck-vba/Rubberduck] [codecov[bot]](github.com/apps/codecov) edited comment on pull request #3956: Introducing Busy state and SuspendParser action

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/3956?src=pr&el=h1) Report
> Merging [#3956](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/3956?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/024eecde21aeca61da2cd9d66e2dfadfe0ed2be8?src=pr&el=desc) will **increase** coverage by `3.97%`.
> The diff coverage is `14.81%`.

```diff
@@ Coverage Diff @@
## next #3956 +/- ##
=======================

[rubberduck-vba/Rubberduck] Vogel612 added label PR-Status: Review Requested to pull request #4006: Splice Off Rubberduck.Refactorings from Rubberduck.Core

[rubberduck-vba/Rubberduck] bclothier created comment on pull request #3956: Introducing Busy state and SuspendParser action

> Addressed most of @MDoerner 's comments and hopefully the revised version is now thread-safe.

However, what I could not come up with a clean way of specifying allowable state other than `Ready`. Originally, the test engine (actually the command) would allow running tests when state is in `ResolvingReferences`/`ResolvedDeclarations` in addition to `Ready` -- however if I pass it, I cannot make sense of what state the `SuspendParse` should set back as it would be an incorrect thing to do and

Vogel612

soo @QHarr how do you like your new toy?

Duga

[rubberduck-vba/Rubberduck] Imh0t3b created comment on issue #3870: Rubberduck stuck on "Pending" Status after downgrade from O2016 to O2007

> First, sorry for the late reply, but I was disabled for a couple of weeks. And sorry for expecting you translation my log (just read the first 2 lines and expected the rest to be english too).

I just managed to fix the error, but this one is not the initial.

The Error "The file or assembly ... or a dependency of it was not found" was caused by using an extracted source .zip (not my GIT-Version, because I wanted clean source files to prevent errors in my repository files distracting me fr

[rubberduck-vba/Rubberduck] Imh0t3b edited comment on issue #3870: Rubberduck stuck on "Pending" Status after downgrade from O2016 to O2007

> First, sorry for the late reply, but I was disabled for a couple of weeks. And sorry for expecting you translation my log (just read the first 2 lines and expected the rest to be english too).

I just managed to fix the error, but this one is not the initial.

The Error "The file or assembly ... or a dependency of it was not found" was caused by using an extracted source .zip (not my GIT-Version, because I wanted clean source files to prevent errors in my repository files distracting me fr

[rubberduck-vba/Rubberduck] build for commit c82b58e0 on unknown branch: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] [codecov[bot]](github.com/apps/codecov) edited comment on pull request #3956: Introducing Busy state and SuspendParser action

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/3956?src=pr&el=h1) Report
> Merging [#3956](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/3956?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/024eecde21aeca61da2cd9d66e2dfadfe0ed2be8?src=pr&el=desc) will **decrease** coverage by `0.07%`.
> The diff coverage is `24.03%`.

```diff
@@ Coverage Diff @@
## next #3956 +/- ##
=======================

[rubberduck-vba/Rubberduck] build for commit c82b58e0 on unknown branch: 53.74% (target 0%)

[rubberduck-vba/Rubberduck] Vogel612 edited comment on issue #3870: Rubberduck stuck on "Pending" Status after downgrade from O2016 to O2007

> First, sorry for the late reply, but I was disabled for a couple of weeks. And sorry for expecting you translation my log (just read the first 2 lines and expected the rest to be english too).

I just managed to fix the error, but this one is not the initial.

The Error "The file or assembly ... or a dependency of it was not found" was caused by using an extracted source .zip (not my GIT-Version, because I wanted clean source files to prevent errors in my repository files distracting me fr

Vogel612

^^ fixed formatting error

Duga

11:59 PM

[rubberduck-vba/Rubberduck] Imh0t3b edited comment on issue #3870: Rubberduck stuck on "Pending" Status after downgrade from O2016 to O2007

> First, sorry for the late reply, but I was disabled for a couple of weeks. And sorry for expecting you translation my log (just read the first 2 lines and expected the rest to be english too).

I just managed to fix the error, but this one is not the initial.

The Error "The file or assembly ... or a dependency of it was not found" was caused by using an extracted source .zip (not my GIT-Version, because I wanted clean source files to prevent errors in my repository files distracting me fr

Transcript for