VBA Rubberducking

I find git hard to understand as I could not find the meaning of the words used for the actions. I have checked the dictionary for the meaning of 'stage' and none of the meanings were related to source control concepts. What does 'stage' mean in the context of git?

It's basically a way for you to split your working directory into a commit, and a future commit.

IvenBach

Lonche time!

this

From what I am seeing, that's not a part of git's "language" per se. It seems to be strictly conceptual. This confirms, too: githowto.com/staging_changes

Aside: I can't understand why there are people who are perfectly fine with coding their thing without any indentation. I'm seeing this all over:

Set foo = Foo
 Set = bar = Bar

That sets my OCD off.

Cyril

@this i'm on the 1NF and the gist is my db is shit... lol

this

Everyone goes through that stage. I know I've made a non-1NF'd databases before.

Cyril

7:48 PM

with this being the first time seeing these standards... i've got a long way to go. so much repeated data

this

I blame Excel. :)

Cyril

i mean, VBA is the reason i have so much trouble wtih my JS...

";" is very difficult to remember

this

Sep 24 at 23:35, by this

well, because as I said, excel is quite brain-damaging.

Hosch250

You don't need ; in JS.

Newlines work just as well.

this

uh, I don't think so.

maybe you mean braces?

that would not require a ;

Cyril

7:51 PM

if(blah=blah){
kill_all_human;
}

the break after kill_all_human needs to exist or it doesn't run

or i should say, it just skips it

Hosch250

No.

JS doesn't need ;.

It's a serious pain, because sometimes it interprets linebreaks wrong.

You have to be careful that the linebreak doesn't occur after a valid statement.

TypeScript doesn't require it either, although they strongly recommend it and reformat it for you if you set it up that way.

davidwalsh.name/javascript-semicolons

Cyril

in several cases i have had javascript not execute as written because i do not use a break, so it was my understanding that javascript would require that

Hosch250

LOL.

> The concept is known as JavaScript Semicolon Insertion or "Automatic Semicolon Insertion".

90

A: Do we need semicolon at the end?

The concept is known as JavaScript Semicolon Insertion or "Automatic Semicolon Insertion". This blog post: JavaScript Semicolon Insertion: Everything you need to know outlines the concept well in an understandable manner using examples under the headings: Where Semicolons are Allowed Where S...

Cyril

well that's nice and confusing lol

this

JS...

because C and Perl wasn't bad enough, apparently!

Cyril

7:57 PM

this straight up didn't work without the ";", but looks like it can be omitted:

<script type="text/javascript">
var count = 1;
function adjust_zIndex(ele_id) {
document.getElementById(ele_id).style.zIndex = count++;
}
</script>

should specify, the count++; didn't work

@Hosch250 that movie holds up ridiculously well today... really wish something would happen to change that

between @this and @Hosch250 i've got several days worth of reading...

@MathieuGuindon i haven't seen on SE meta if they're still talking about allowing a checkbox/answer from a comment. last i heard it wasn't going to be implemented, despite SO wanting it, because it would be a global SE change which may be inappropriate. you have any visibility if that is still being discussed at mod level?

Mathieu Guindon

8:12 PM

I would evaluate the probability of this ever being implemented, at ~0.000000000000000000000000000001%

this

O_O I'm not implemented?!?

Cyril

his parents are real outliers

statistically speaking

IvenBach

@IvenBach A local commit is conflicting with a remote commit when merging. Yep I needed that food and to step away.

mansellan

8:46 PM

@this Nope, you're abstract.

2

Mrblackey

If not even a pointer to abstract...

IvenBach

I spy a kitty cat.

Uh oh... VS is giving me the metadata file not found errors again.

Mrblackey

It did the same for me last week.

IvenBach

How did you resolve it?

I had this same issue a while back and couldn't ever get it resolved. Moving to a different workstation became the fix.

Duga

[rubberduck-vba/Rubberduck] BZngr created pull request #5205: Incorporate InspectionTestsBase progress - part 1 to merge BZngr/Rubberduck/5191_AthruE into rubberduck-vba/Rubberduck/next

> This PR incorporates InspectionTestsBase for Inspection unit test names beginning with the letters A thru E. So...it is progress on #5191 - but not complete. I thought I would generate a PR at this point of the process in case there were comments that would apply to the content of this PR as well as the next set of Inspection unit test updates.

Mrblackey

9:00 PM

I didn't, actually. Or maybe yes?
I checked out to Master, Build, then rebuild again from Next.

I can't really call it a solution, but it worked... :D

IvenBach

Just tried and didn't work on my dev branch.

M.Doerner

Is there a particular reason why we always access the references on the IVBProject of a project when we create the AddRemoveReferencesPresenter?

We save them on the ProjectDeclaration and update them each parse.

this

I don't think it has to do w/ parsing

What if I add or remove references via the VBIDE's built-in dialog

and never parse, and then proceed to open RD's references dialog?

At that point, the project's declarations would be stale.

Duga

[rubberduck-vba/Rubberduck] build for commit 63fca903 on unknown branch: AppVeyor build succeeded

this

If we completely hijacked the references dialog (e.g. the Tools -> References now open our own dialog, then that might be unnecessary.

but some may not like us hijacking in this manner.

Mrblackey

9:15 PM

This could be optional behavior.

I would actually prefer it this way.

this

So would I, actually. I keep forgetting to use the CE to open the RD's dialog.

M.Doerner

Actually, the IVBProject on the ProjectDeclaration is stale after the next parse.

You should no longer access it.

this

But given the example above, how do we know the references was changed?

I rememer comintern commenting that he was having problems with add/remove references showing out of date data, which necessitating him adding that extra check

M.Doerner

The basic problem is that the Project member on the ProjectDeclaration should not exist.

We currently guarantee that the IVBProject gets released by providing it from the ProjectsRepository.

However, this means that it will always be released after the next parse and thus somewhere between useless and dangerous.

this

Strangely, that seem to have came from this PR: github.com/rubberduck-vba/Rubberduck/pull/2551

M.Doerner

9:30 PM

Before, it was on the QMN, which was even worse.

The declarations relly should not store COM wrappers.

this

IMO, what we really need is a COMWrapperFinder.

where one can provide a QMN or a Declaration and get back a wrapper.

M.Doerner

I removed all the components, but not the project, because that is used in so many places.

We have that.

this

you mean the ProjectsRepository? That requires an DI

M.Doerner

It is called ProjectsRepository

this

I think the problem is that there are several places where DI is not practical

M.Doerner

9:31 PM

Yes, it does.

this

Not sure that applies in the case of the presenter, though.

But I seem to remember there may be other places where DI would add extra complications.

M.Doerner

In those rare cases, you will have to get your appropriate project from the VBE.

What I really do not like is that we tend to store COM wrappers in a lot of places.

IvenBach

Any way in which I merge upstream/next into my dev branch I get matadata file not found errors.

Tried 4 times now with same result.

this

I agree we shouldn't store COM wrappers. They need to be treated as a transitional.

I was thinking that we should force access via a lambda parameter

e.g. GetSomething(qmn, w => (w.DoSomething();))

@IvenBach I'm curious --- what diff do you see in the merge?

Duga

[rubberduck-vba/Rubberduck] MDoerner opened issue #5206: Remove Declaration.Project

> I propose to remove the member `Project` from `Declaration` and especially from `ProjectDeclaration`. This will have to consequence that `ProjectDisplayName` will also have to be removed from these.

The problem with this member is that it returns a stored `IVBProject` instance. As managing all references to declarations is near impossible, we hand a project from the `ProjectsRepository` to the declaration. Ufortunately, this means that the project will get released on the next parse and thu

[rubberduck-vba/Rubberduck] MDoerner added label development-issue to issue #5206: Remove Declaration.Project

[rubberduck-vba/Rubberduck] MDoerner added label technical-debt to issue #5206: Remove Declaration.Project

IvenBach

9:49 PM

@this Are you asking what files are conflicting Team Explorer>Resolve Conflicts pane. Or are you asking for the actuall diff output?

this

the actual diff output for the merge commit

IvenBach

ivenbach@workstation MINGW64 ~/source/repos/Rubberduck (Issue5109_Consolidate_copy_command_logic|MERGING)
$ git diff
diff --cc Rubberduck.Core/UI/CodeExplorer/CodeExplorerWindow.cs
index 6ae3fb601,1c6d89027..000000000
--- a/Rubberduck.Core/UI/CodeExplorer/CodeExplorerWindow.cs
+++ b/Rubberduck.Core/UI/CodeExplorer/CodeExplorerWindow.cs
@@@ -20,7 -20,7 +20,11 @@@ namespace Rubberduck.UI.CodeExplore
          public CodeExplorerWindow(CodeExplorerViewModel viewModel) : this()
          {
              ViewModel = viewModel;

I recognize all of my edits and think I'm choosing the correct Source/Target/Merge where appropriate.

M.Doerner

Have you manually resolved the conflicts?

this

Yeah, I don't see anything that looks out of place but then again I am not sure whether the diff will show other changes from the merge that wasn't yours.

IvenBach

@M.Doerner Yes. A couple times now to ensure it wasn't something I borked. Each time a conflict has arisen.

this

9:54 PM

What I'm wondering is that the act of merging from the upstream/next somehow causes a modification or deletion of some metadata that you never edited and isn't reported as a conflict.

M.Doerner

Could you double check whether the tags in the XML balance?

IvenBach

In trying for a solution I've taken source for all merge conflicts and it errors out.

In the conflict files?

M.Doerner

Yes, I have seen at least one instance where taking one side will cause a missing closing data tag.

That is always a pain.

Hm, no. That isnstance looks OK after scrolling further.

However, I get such problems regularly with the inspection resources.

mansellan

@this Actually I've been thinking that too - never give out a COM wrapper, just act on expressions on them. A lot of the Core 3 extensions work exactly like that (e.g. SomeMethod(Action<ISomeInterface> action))...

You can't store something you're never given...

"Tell me what to do with the toxic waste", rather than "Here's the toxic waste you ordered, have fun!"

this

10:14 PM

^ @M.Doerner what do you think?

IvenBach

@M.Doerner It looks like it's closed everything properly. Using N++ and Ctrl+End to the bottom shows </root> to be the last without any hanging collapse icon in the lefthand gutter.

What's the easiest way to check?

M.Doerner

I find passing aound lambdas all the time not very helpful for readability, to be honest.

mansellan

it takes a little getting used to, but the way they've implemented it in Core 3 is impressively clean

and besides, you can always use method groups instead of lambdas

this

Granted that it does clutter things up but I don't know what else to make it abundantly clear that this thing is be treated as ephemeral and thus not safe to store.

M.Doerner

Moreover, how do you imagine to implement this halfway performantly?

this

10:17 PM

wait, lambda hurt performance? I haven't considered that aspect.

M.Doerner

No

mansellan

well it's an extra indirection...

this

then what do you mean by performant?

M.Doerner

But turning everything the other way around.

mansellan

it would mean that (perhaps) the VBEditor project would own all COM wrapper instances, which would never leave it, and it would execute all operations on them and manage their lifetimes.

M.Doerner

10:20 PM

I already started to implement useful wrappers like ISelectionService, but to get all functionality of the VBE API requires tons of wrappers.

You would basically need to know all types you will want to get from it.

mansellan

not if you just accepted (eg) an Action<IVBProject>, or a Func<IVBProject, TReturn> ?

M.Doerner

If you make it generic, I bet somebody will use a Func<IVBComponent,IVBProject>.

mansellan

heh... yeah...

bah

you could add a runtime guard, but that's not compile-time...

this

my thought was that with a lambda taking only Action, we can easily write an analyzer to disallow storing references outside the lambda expression and returning it.

That way, all checks are compile-time.

M.Doerner

Currently, my main pain point is really the IVBProject on the ProjectDeclaration. Apart from that and the VBE itself, we already treat next to all wrappers either as transient or store them in the ProjectsReporitory.

this

10:26 PM

Earlier I considered using namespace as a guard but that would not be suffice to ensure that the references are used temporarily

M.Doerner

Action is not sufficient.

You have to return stuff like the selection.

And anyway, even Action is not safe.

mansellan

@M.Doerner why?

M.Doerner

(IVBE vbe) => _myProject = vbe.ActiveProject;

Captured variables.

mansellan

yeah...

M.Doerner

Closures can leak everything.

mansellan

10:29 PM

ok, not gonna work...

M.Doerner

At least, you have to make quite an effort in this case.

Usually, wrappers are accessed because they are readily available.

The remaining stored wrappers are really technical dept from the past.

Two years ago, the QMNs still carried IVBCombonent references around.

Or is it three years by now?

Duga

[rubberduck-vba/Rubberduck] MDoerner assigned MDoerner to issue #5206: Remove Declaration.Project

[rubberduck-vba/Rubberduck] philiplaureano starred us

3

mansellan

@M.Doerner Does the closure issue still arise if you accept Expression<Action<IVBProject>> ?

M.Doerner

I have no experience with that.

It does not matter anyway.

mansellan

yeah I'm on shaky ground there... I know it's the basis of IQueryable

M.Doerner

10:39 PM

We need to return information and also pass other parameters.

mansellan

yeah

ah well

hmm, actually, iiuc you can wrap any Action or Func in an Expression.. not sure if that helps us though...

afaiui, it changes it from being a function pointer to the AST needed to construct a function

like I say though, I'm on shaky ground...

on the upside though, we could invent a new branch of Linq - Linq-to-Toxic-Waste :-)

ttgtb

IvenBach

11:59 PM

Made no headway on merging next. I don't know what to do about it. Self note: github.com/rubberduck-vba/Rubberduck/issues/2777 is my next PR to work on tomorrow.

Home time.</iven>

Transcript for