VBA Rubberducking

12:00 AM

RELOAD!

[FreezePhoenix/Angels-And-Demons] 1 commit. 6 additions. 3 deletions.

[rubberduck-vba/Rubberduck] 4 commits. 1 opened issue. 38 issue comments. 472 additions. 34 deletions.

@Comintern I'm curious, what was the Professional feature you were missing earlier, that would have helped the problem at hand? I thought Community and Professional differences came down to licenses and MSDN access.

Comintern

@ThunderFrame Trying to remember what it's called. It's the references\test results\etc. display above the function names.

Gimme a sec and I'll screen-shot it.

In a real project it also shows things like number of unit tests, number of passing tests, etc.

Mathieu Guindon

I love VS Enterprise for having that

I love that MVP gets me VS Enterprise =)

Comintern

12:16 AM

Ah, it's called CodeLens. msdn.microsoft.com/en-us/library/dn269218.aspx

Mathieu Guindon

If we implement that in our panes, can we call it the same?

Comintern

I would doubt it, we'd probably have to call it DuckLens or something.

Mathieu Guindon

Ha! Awesome nonetheless

That means custom IntelliSense has to be ...QuackSense?

Comintern

It's addictive as hell - especially the source control integration.

Mathieu Guindon

Yup

Comintern

12:18 AM

That has a better ring to it than RubberSense.

RubberSense sounds like a brand of condom.

If we do a code coverage feature for unit testing, we absolutely have to call it "Duck and Cover".

Mathieu Guindon

lol

Comintern

OK, I swear I'm ready to completely reinstall VS. I can't get around the testing error at all. AFAICT it's an issue with MSBuild for .NET Core.

Mathieu Guindon

R# test runner doesn't seem to work, but VS does AFAICT

Tweeting Duck

Sometimes I wish I could downvote @UserVoice feature requests... The number of votes on this one is deeply worrying. Are that many people stuck in 1980? https://twitter.com/ITImpactInc/status/1017529807490502658

Tweeted by rubberduckvba on July 13, 2018 at 12:13 AM

Comintern

12:33 AM

For some reason it was loading the MSBuild for Core, but that references an older System.Xml.XPath.XmlDocument assembly than nunit.

Ahah! The .NET Core uninstaller left a bunch of residual crap.

this

12:53 AM

@MathieuGuindon Exactly! Also, when you think about it, even VS' stacktrace include the line number. Unlike VBA, the line numbers don't form a part of the code. The feature request need not mean that it be part of the code. But really easier if they just bought Wayne out.

ThunderFrame

@this any idea who's behind the uservoice request for line-numbers?

this

a "Greg". No idea who he is

ThunderFrame

Oh, clearly I have no idea how big IT Impact is.

this

No, that's not him.

The one that Mat's relying to is my boss.

ThunderFrame

Yeah sorry, the tweeter, not the OP. I assume you have different views to your boss WRT line numbers?

Oh wait, just read your chat to Mat properly.

Comintern

1:01 AM

In fact, I wrote a quick and dirty macro to line number a function for my boss just a couple months ago.

ThunderFrame

@Comintern paid by the number of lines?

Comintern

lol

this

Ah sorry, my bad. Yes the tweeter is my boss; I think the item got floated for promotion.

@Comintern why not use MZ Tools? Or even vbWatchDog? That would have been much cheaper than paying you, I think.

Comintern

@this It took less time to write than to download.

this

i'm sure.

it's until you edit stuff or whatever it goes out of the side. At least vbWatchDog's line numbering is done right.

ThunderFrame

1:04 AM

@Comintern there was an answer on SO last week where the user thought that an Integer used more memory than a Long because it used more characters to type. We assumed he must be paid by the character.

this

@ThunderFrame No, then he'd be arguing for the other way around. He would have to be paid for using less character, not more. (how does one measure that....)

Comintern

I dare somebody to pay me by the line.

148

A: Longest code to add two numbers

C - 2,739,341,494,945,868,415,002 (Not including whitespace) Brute force for the win. Only handles integers, and preserves overflow behavior. Here's a snippet of the code: int addnumbers(int x, int y) { if (x == -2147483648 && y == -2147483648) return 0; if (x == -2147483648 && y == -21...

ThunderFrame

@this makes more sense to show line numbers as a title row adjacent to the code. The toolbar is enough IMO, but I can see how having multiple line numbers visible might help in certain situations. Might also be useful to show physical/logical line numbers, depending upon what you're doing. Just keep them out of the code.

Comintern

Should I have a version 15.0 key in `HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\MSBuild\ToolsVersions`?

AFAICT, VS0217 is using the MSBuild from VS2015.

this

@Comintern but if you wrote a function to write the code, you would be paid for that function, not the output.

Mathieu Guindon

1:09 AM

@this except the feature request is completely talking about line-number labels

Comintern

@this Shhhhhhh....

Mathieu Guindon

@this sorry for the conversation you're going to be having with him tomorrow :)

this

@MathieuGuindon One would hope that MSFT would be a bit more smarter than to accept it at face value.

@MathieuGuindon I'll live.

but I saw why - in the feed he's promoting all VBIDE UV items - which isn't surprising since he wants them to invest more. Despite what they say, we're still making solutions with "1980" programming language today.

Mathieu Guindon

ugh

Comintern

Our use case is simply tech debt. We have production VBA code with a main working function that is ~1000 lines long (it's a huge loop with about 30 global state variables). It's used by non-technical customers, so we don't put any useful error details in the message boxes. The project is locked, and we open it in its own process so it won't assert a debugger. Line numbers and Erl :shudders: are the only thing that makes it remotely maintainable.

Mathieu Guindon

1:14 AM

if only there was a VBE tool with an extract method refactoring...

Comintern

@ITImpactInc If you need a line number to find an error, your function is doing too much. Refactor to the point where you don't need to number lines and you'll also be rewarded with much more maintainable code. Crutches are only useful if you can't walk yet.

Tweeted by CominternVBA on July 13, 2018 at 1:16 AM in reply to ITImpactInc

@MathieuGuindon Oh, I re-wrote the entire thing in my spare time. It runs about 10 times faster and it's completely OO. Management doesn't want to spend the money to run it through QA until it's broken to the point that they have to.

It was reminiscent of the legacy SmartIndenter code base.

Mathieu Guindon

damn, for a split-second I read that you had single-handedly rewritten the RD refactoring

@Comintern ouch

Comintern

If I had, I currently have no way to unit test it. :-/

Comintern

1:30 AM

OK, WTH does the VS2017 installer insist on installing the .NET Core toolchain?

Mathieu Guindon

1:41 AM

ok that's a tough one. the Call MsgBox( off-by-one crash? VBE is "helpfully" removing the empty parens when the autocompleted line is inserted, so I do module.ReplaceLine(pSelection.StartLine, "Call MsgBox()");, and still end up with Call MsgBox on the line

I mean, that's not the off-by-one (that one's fixed)... but the result is that there's no way to open parens when you use an explicit Call statement

...which... is ...interesting...

ideas?

Duga

[rubberduck-vba/Rubberduck] retailcoder created comment on issue #4154: Access crashes while writing in Code window

> Another work-around is to avoid `Call` altogether.

In fact, I've got the off-by-one under control, but a quirk of the VBE is literally preventing the insertion of the empty parentheses - they get "helpfully" removed as soon as they're inserted, making the `(` keypress feel like it's simply not working.

Mathieu Guindon

ok, I know what I'll do. I'll add bogus TODO content between ( and ), and select it

Comintern

What about sending something like (.) and then "deleting" the .?

Mathieu Guindon

the only way the VBIDE API lets me write to the code pane, is by replacing an entire line

that triggers "prettification"

so I can't delete that dot

might as well make it TODO, no?

Comintern

Does it set the caret?

Mathieu Guindon

1:48 AM

it being immediately selected won't impact typing at all

@Comintern it puts the caret on the next line, so AC needs to reposition the selection

...while accounting for any whitespace adjustment the VBE made

Comintern

If you can immediately select it, you could easily just PostMessage a delete keystroke.

Mathieu Guindon

except I'm already in the msgloop handling/eating a keystroke...

Comintern

Hmm....

I'm trying to remember - does changing the lparam in the message handler alter the keystroke?

Mathieu Guindon

didn't try, but then not eating the keystroke changes the logic quite a lot

Comintern

Yeah, I imagine it would.

Why does working with the VBE always involve hacks?

Mathieu Guindon

1:52 AM

because the VBE is a hack?

Comintern

Ah yes - hackishness is viral.

Mathieu Guindon

if extending the VBE was easy, we wouldn't need Rubberduck, right? ;-)

Comintern

OK, if this VS reinstall doesn't work, I'm seriously considering reinstalling Windows and starting from scratch. What a PITA.

You'd think it was the VBE or something.

Mathieu Guindon

oh shit

this is not going as planned

ok screw this. I need a plan B

Comintern

...and still no tests.

Mathieu Guindon

1:59 AM

OTOH, I kinda like that making a Call statement is impossible with AC on

Comintern

@MathieuGuindon Now that I at least have an IDE again, where's the code you're working with?

Mathieu Guindon

Rubberduck.Core.AutoComplete

Comintern

Pop up an obsolete call inspection result instead.

Mathieu Guindon

if (IsCallStmt(original)) { StripCallStmt(); }

seriously the only real fix is to remove the Call statement

the off-by-one is on AutoCompleteBase line 96 BTW; prettifiedTokenIndices is empty

Comintern

Can't you just suppress it autocomplete as a special case?

Mathieu Guindon

2:03 AM

that's what I'm thinking

but then foo = Recall( needs to work too :)

this

I hate to be that guy but -- if that's so problematic, should we be devoting our effort to making a AvalonEdit codepane?

Mathieu Guindon

@this yeah but I just want to deliver a 2.3 that doesn't crash all the time...

ideally by August

this

yeah, fair point.

Mathieu Guindon

it's really just patching - I have AC half-rewritten in another branch

Comintern

Toolchain problems suck. Doesn't help that the configuration looks completely different than it did in VS2015.

OK, that was dumb. I was trying to test some auto "correct" behaviors and just crashed Excel. Note to self, disable RD first...

What happens if you append a backspace to the end of a string that gets passed to ReplaceLine?

Mathieu Guindon

2:10 AM

what's the character for a backspace?

there's no WM_KEYDOWN message for backspace..

no wait

yeah there's no WM_CHAR message for backspace

I'm handling backspace through WM_KEYDOWN

Comintern

It's 0x8.

Mathieu Guindon

trying it. expecting a ? in the code pane

or some other junk

Comintern

I was thinking more along the lines of module.ReplaceLine(pSelection.StartLine, code + " \b");

Kind of like how you can update an existing line with Console.Write

Mathieu Guindon

same

..and the result is... sweet nothing. it's ignored.

Comintern

weird.

Now I'm curious if it would compile. I may have to try and break things.

Mathieu Guindon

2:18 AM

@Comintern IKR! Console AC was much easier!

19

Q: Autocompleting console input

Yesterday I stumbled upon a Stack Overflow question that asked if it was possible to implement Tab-triggered auto-completion in a console application. I thought that was an interesting idea, so I went ahead and wrote a small program that does just that, given a hard-coded string[] - I might even...

actually that was kinda lame

Comintern

Does CodeModule.AddFromString have the same behavior? You could call module.ReplaceLine(pSelection.StartLine, string.Empty); first, then call AddFromString.

Mathieu Guindon

hmm, everything points to the VBE immediately prettifying the inserted code

haven't tried though

also, losing member attributes on module rewrite is one thing, losing member attributes when typing code inside a procedure is... another

question: is a Call statement the only instance where the VBE would be so helpful as to remove trailing, empty parentheses?

Comintern

Yeah, just tested. AddFromString does the same thing.

Mathieu Guindon

and nukes all member attributes ;-)

Comintern

I'm assuming that it's Call and not MsgBox...

Mathieu Guindon

2:32 AM

yeah msgbox() works fine

but call msgbox() becomes Call MsgBox

Comintern

Do you have easy access to the references?

which refs?

From the parser.

not in AC

Damn. I was thinking of defaulting it to MsgBox(""), but to be a general solution you'd need to know the type of the first parameter (or whether it had one).

Mathieu Guindon

2:35 AM

yeah, no. I can't assume the initial parse has happened

Comintern

Right. That makes sense.

And you're already in the message pump... Hmmm...

Do we know what window messages the VBE sends to the code pane when it does it's prettify?

You could just set a flag and eat them.

It has to be sending a set text message or something like that in order to update the control.

Mathieu Guindon

@Comintern God I wish!

Comintern

Gimme a sec, I'll Spy++ it.

ThunderFrame

2:54 AM

@MathieuGuindon once we get declarations from TypeLib, we could have signatures worked out without a full parse. Or, we could just pre-parse the declararations rather than every line, and that would be faster. My preference would be using the TypeLib.

Mathieu Guindon

yeah

that'll be useful for ~~Intelli~~ QuackSense

Comintern

OK, the ReplaceLine triggers a WM_NOTIFYFORMAT exchange.

I can't find the control message though.

Mathieu Guindon

best I can do

typing the ( character nukes the Call statement. I can't add an extra whitespace without intercepting that WM_NOTIFYFORMAT

not exactly intuitive, but totally surprising enough to never want to make a Call statement again

ThunderFrame

if you replace Call MsgBox(| with Call MsgBox(1|) then send a {backspace}?

Mathieu Guindon

I can't send a backspace

at least not through the VBIDE API

Comintern

3:09 AM

I think that's reasonable.

ThunderFrame

sure, but with win32?

Mathieu Guindon

TBH, Call adepts can disable parenthesis AC if they want it that bad

ThunderFrame

but then, you could just send all of the keys

Comintern

I don't think the client area of the code pane is attached to the message pump at all. The wierd thing is that it asks itself "what character set do I want", and then replies to itself "ansi". And that's it.

ThunderFrame

add a few ms delay between keypresses, and you could have AC write the entire macro for you ;-)

Comintern

3:12 AM

I still like the idea of eating the keypress and triggering an immediate code inspection if they ever type Call.

Make it modal too.

Mathieu Guindon

@Comintern Rubberduck.VBEditor.WindowsApi.CodePaneSubclass.cs

ThunderFrame

@Comintern and an "Are you sure?" confirmation

Mathieu Guindon

//Stub for code pane replacement.  :-)

Comintern

Yeah, kind of like the auto syntax check.

@MathieuGuindon Oh yeah. That's the infrastructure for taking over the message pump.

Mathieu Guindon

that's what AC is using

Comintern

3:15 AM

That's exactly what it's there for.

Is it still configured to auto-subclass everything?

Mathieu Guindon

I think so

Comintern

I might take another look at that code. Isn't there an open issue with pre-existing windows not attaching?

Mathieu Guindon

yup, tagged autocomplete

Comintern

I seem to recall something about the VBE recycling it's hWnds.

OK, I really need to unf*ck my build environment.

Duga

[rubberduck-vba/Rubberduck] retailcoder created pull request #4177: Critical AC fixes to merge retailcoder/Rubberduck/rdnext into rubberduck-vba/Rubberduck/next

> Fixes the off-by-one bug that sends everything up in flames when you type this: Call Anything( So the good news is that it's fixed. The "bad" news... is that when parenthesis autocompletion is enabled, Call statement is special-cased and stripped - because the lovely VBE is being "helpful" and instantly removes empty parentheses in Call Something(), leaving the ( key appear to do exactly nothing. Short of preventing the VBE from "prettifying" the replaced line, stripping the explicit...

Call statement is really the only thing AC could do. See it in action

Mathieu Guindon

3:22 AM

ah, crap.. I made my branch with git checkout upstream/next -b rdnext, and now VS' Team Explorer thinks git push goes upstream and isn't letting me

git push origin works fine

how do I fix this?

(the half-rewritten AC is in my next - yes, I'm this dumb)

Comintern

Create a new clone of origin/next, then set that as your upstream and merge just the branch into it?

Mathieu Guindon

really?

Comintern

I think you can do that.

@Hosch250 might have a better idea. He's always been the git guru.

Mathieu Guindon

nah, @Vogel612 is :)

(is... well asleep at this time)

Comintern

I think you lose the history that it branched that way, but if you clone and then swap the config with the branch on the filesystem, in theory you should be able to just commit it.

The intermediary commits get lost though. I'm not sure what happens when you merge it back into your "real" next.

Mathieu Guindon

3:31 AM

yeah that's the thing... I don't want to lose the work I already put in getting AC testable :)

thank God you can't name a member "Call"

(stress-testing the PR)

I wish I could inject a tooltip at the caret position

Comintern

What about an enum or type member?

Mathieu Guindon

then it wouldn't be followed by an identifier and then a (, no?

foo.Call MsgBox( -> MsgBox

Duga

[rubberduck-vba/Rubberduck] [codecov[bot]](github.com/apps/codecov) created comment on pull request #4177: Critical AC fixes

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4177?src=pr&el=h1) Report
> Merging [#4177](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4177?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/abe03b04767868cce2672bff8c9289d0e1bbf2ad?src=pr&el=desc) will **decrease** coverage by `0.02%`.
> The diff coverage is `0%`.

```diff
@@ Coverage Diff @@
## next #4177 +/- ##
===========================

[rubberduck-vba/Rubberduck] build for commit 798c6b45 on unknown branch: 52.46% (target 0%)

[rubberduck-vba/Rubberduck] build for commit 798c6b45 on unknown branch: AppVeyor build succeeded

Comintern

Probably not. I'll try and break it after I'm synched.

BTW, I forgot this bit:

        //This is an output window firehose, leave this here, but comment it out when done.
        //if (idObject != (int)ObjId.Cursor) { Debug.WriteLine("Hwnd: {0:X4} - EventType {1:X4}, idObject {2}, idChild {3}", (int)hwnd, eventType, idObject, idChild); }

It's useful as hell for debugging the message pump.

Mathieu Guindon

yup

Mathieu Guindon

3:54 AM

lol there's a #partylikeits1998 hashtag trending on twitter

ThunderFrame

Does @Folder work atm?

I'm a little behind latest, but can't create folders

Mathieu Guindon

@ThunderFrame works here

Comintern

Huh. I apparently mis-remembered how that was wired up. We only call VBENativeServices.AttachWindow in response to a WinEvent.ObjectCreate. Apparently existing windows were never subclassed.

Mathieu Guindon

(latest+#4177)

Comintern

I think the only thing that needs to happen is to enumerate the child windows on startup.

Mathieu Guindon

3:56 AM

yeah, me too

ThunderFrame

@Comintern you need the floating ones too

Comintern

I'm curious if that would work in a static constructor for VBENativeServices...

Mathieu Guindon

wouldn't we rather want to have full control over the timing of the thing?

Comintern

That wouldn't be deterministic, but the first call on the class is always going to be attaching the event handlers to the VBE, and that's about the perfect time to run the ctor.

Mathieu Guindon

VBENativeServices is likely getting created while the splash screen is shown

hmm

Comintern

3:59 AM

We're creating the event handlers that early in the load process?

Mathieu Guindon

no, I don't think

but I'd explicitly do it near where it's happening

Comintern

Although actually, I don't think that lazy initialization of statics is to spec.

Mathieu Guindon

I suppose, in App.cs

Duga

[rubberduck-vba/Rubberduck] retailcoder merged pull request #4177: Critical AC fixes from retailcoder/Rubberduck/rdnext into rubberduck-vba/Rubberduck/next

Comintern

In theory the CLR could run it before anything. I just doesn't. So maybe not the best plan.

Duga

4:00 AM

[rubberduck-vba/Rubberduck] retailcoder pushed commit 798c6b45 to next: fixes #4141, closes #4154

[rubberduck-vba/Rubberduck] web-flow pushed commit ab68630a to next

Merge pull request #4177 from retailcoder/rdnext

Critical AC fixes

[rubberduck-vba/Rubberduck] retailcoder closed issue #4141: Off-by-one AC bug crashes host

[rubberduck-vba/Rubberduck] retailcoder closed issue #4154: Access crashes while writing in Code window

ThunderFrame

@MathieuGuindon damn - had an extra blank line between Option Explicit and @Folder foo so the annotation was above the Sub Foo. Removed the blank line, the annotation was treated as being in the declarations section, and then it worked. I thought you fixed that?

Mathieu Guindon

associating annotations to procedures is borked

Comintern

I'll just put it in HookEvents. Everything checks to see if it's stored in TrackedWindows already, so that should be a relatively simple fix.

ThunderFrame

seems the annotation is still only recognized if it's in the declarations section. The parser doesn't scan until the first Sub/Function/Property declaration statement, so comments above the statement can be treated as belonging to the procedure, and not to the declarations section

Duga

[rubberduck-vba/Rubberduck] comintern assigned comintern to issue #4080: CodePaneSubClass is only hooked on newly-created windows

[rubberduck-vba/Rubberduck] comintern created comment on issue #4080: CodePaneSubClass is only hooked on newly-created windows

> See chat.

Comintern

4:07 AM

@ThunderFrame Are the dockables the only ones that aren't under the child window tree of the VBE main window?

Tweeting Duck

"Bad news": because the VBE will "helpfully" turn 'Call Foo()' into 'Call Foo', parenthesis completion can't work for a 'Call' statement, so it's special-cased. Type this: 'call msgbox(' Get this: 'MsgBox|' Just type the " (no need for space!) to resume (i.e. to get 'MsgBox "|"')

Tweeted by rubberduckvba on July 13, 2018 at 3:50 AM in reply to rubberduckvba

Good news: Autocompletion critical bugfix coming up.

Tweeted by rubberduckvba on July 13, 2018 at 3:50 AM

Here's a link to a 9-second video showing the surprising new behavior in question: https://www.dropbox.com/s/i6hwqapja5aicew/RD%20AC%20-%20CallStmt.mp4?dl=0

Tweeted by rubberduckvba on July 13, 2018 at 3:50 AM in reply to rubberduckvba

I realize this is surprising behavior that's completely not ideal, and I apologize to all the hardcore fans of the explicit 'Call' statement. But if you had to choose between tripling your productivity with AC, or using obsolete 'Call' statements... I mean, right?

Tweeted by rubberduckvba on July 13, 2018 at 3:50 AM in reply to rubberduckvba

Comintern

I guess I could stub that part for later, but ATM we only attach to codepanes and designer windows.

ThunderFrame

@Comintern IIRC, there are 3 places to look. Child docked windows, Child windows that are undocked/MDI children, and windows that are floating (and floating windows can have a hierarchy too)

and maybe hidden windows also behave differently

Mathieu Guindon

@TweetingDuck that is not the order they were threaded in...

Comintern

The dockable windows are parentless IIR, but I'm pretty sure anything in the Windows collection will always be in the VBE main windows child hierarchy.

I'll make sure to play around with both docked and undocked windows.

ThunderFrame

4:11 AM

@TweetingDuck make the RD settings let you choose between: AC disabled for Call, AC enabled for Call, whereby typing Call will have the keyword removed, and a giant fish emerge from the monitor, to slap you in the face.

Comintern

We currently catch our docked windows explicitly - IIR the dockable window host inherits from the subclasswindow.

Duga

[rubberduck-vba/Rubberduck] retailcoder published prerelease Rubberduck-v2.2.0.3492

ThunderFrame

@Comintern There's a difference between docked and undocked, but if a window is made undockable, it becomes an MDI child. Currently, making an RD ToolWindow undockable makes it so that you can't make it dockable again without either killing the host before it saves, or obliterating your Dock registry key

Duga

[rubberduck-vba/Rubberduck] retailcoder created tag Rubberduck-v2.2.0.3492

[rubberduck-vba/Rubberduck] build for commit ab68630a on next: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] build for commit ab68630a on next: 52.46% (target 0%)

ThunderFrame

If you're poking around in the windows, it would be great if you could discover a way to rearrange them programmatically.

Comintern

4:15 AM

@ThunderFrame I think the MdiChildHack window is responsible for maintaining that relationship. I should play around with that too.

@ThunderFrame Rearrange where? Like where they're docked, or the MDI order?

Mathieu Guindon

@ThunderFrame hmm, alternatively I could simply skip AC given a Call statement. You want to #partylikeits1998, you stay in 1998.

but Call was already obsolete in 1998

Comintern

We could just start a FUD campaign about how using Call is dangerous and unstable. Then leave the crashy behavior as is.

ThunderFrame

@Comintern where they're docked and/or where they're floating.

Comintern

I remember trying that once. I could never figure out how the registry key was coded.

ThunderFrame

Right now, on first use, RD just floats the window. There's an issue somewhere for being able to have finer control over which window they're docked to, so we could create various workspaces. One say, optimized for debugging, and another, optimized for inspections.

@Comintern I've worked out the key, but IIUC, the key is only read on startup and written on shutdown. There's no way to have the key read when you want to change the layout, unless you restart the host.

Mathieu Guindon

4:21 AM

meh... I'd rather work on Avalon than anything that attempts to improve the doomed inspection results toolwindow

Comintern

I still want to own the immediate window.

Mathieu Guindon

then you need to own the debugger, no?

Comintern

Yes. I've pretty much come to the conclusion that it's a glorified terminal.

The debugger uses it kind of like stdin and stdout.

ThunderFrame

@Comintern that has its own project, we now know

Mathieu Guindon

IIUC it's literally an ITypeLib, kind of like an embedded project inside the project

Comintern

4:23 AM

The problem that I was running into is that there's not really a good way to detach another debugger unless you own its handle.

Visual Studio can't even do that.

Mathieu Guindon

googles 'how to tell Windows to STFU and let me write to any memory address and not explode'

Comintern

They keep throwing up roadblocks to that. Not all access is strictly a "violation".

:whistles innocently:

Mathieu Guindon

well, we're in-process, so...

Comintern

OK, that's a lie. I do want to violate the Immediate Window.

3

Mathieu Guindon

calling it a day here

'night folks!

Comintern

4:31 AM

Night.

...and still no unit tests. Note to self - never install a beta .NET Core toolset.

Comintern

4:53 AM

Does the build process set RD to load on startup? I can't seem to get the setting to stick.

Ah, I know how I can test it - assert into the VBE instead of opening it manually.

Vogel612

8:40 AM

@MathieuGuindon git push origin branch -u

U is short for set-upstream-to

There should also be unset-upstream

M.Doerner

9:14 AM

@MathieuGuindon The R# test runner works just fine since I removed the Deployment attribute.

this

12:29 PM

@Comintern FWIW: you can get to the Immediate's VBA project using type lib API. We have yet to try whether that lets you write code to the immediate window or whatever.

Comintern

@this I'm assuming I can just search it in chat?

this

uh, Wayne mentioned it only in passing.

Jan 18 at 16:48, by WaynePhillipsEA

Coming to a rubberduck near you soon... live ITypeLib support from the VBE, including reading and writing the conditional compilation arguments plus determine the types of objects (workbook/worksheet/form/report etc), and controls.

hmm, bummer it was a link on his website

and it's already gone

Comintern

Unfortunately my immediate problem looks like Visual Studio sorted out.

Hosch250

1:05 PM

@MathieuGuindon Yes, because it set the remote on that branch.

So, you need to either unset the remote and push normally, or change the remote to a different branch.

584

Q: How to remove remote origin from Git repo

I just did git init to initialize my folder as git repo and then added a remote repository using git remote add origin url. Now I want to remove this git remote add origin and add a new repository git remote add origin new-url. How can I do it?

git git-remote

Ah ha, Vogel beat me to it.

Malachi

1:49 PM

@IvenBach it is a good video. there are things in there that I forget so I need to watch it again as well. I only made it half way through yesterday

ThunderFrame

2:04 PM

@Comintern, some background material that Wayne has contributed via chat, that has either already been implemented, or should be. I may have missed a few, but SEchat search functionality is not as powerful as it could be. Thought you might find them interesting:

On getting the ITypeLib (despite the message saying ITypeInfo) from the References Object

On getting the ITypeLib (including protected projects) from the Project Explorer

On accessing the identifier pointers and table

On switching to the vTable based interface

On getting conditional compiler constants

On accessing IPeristStorage and getting the raw source code of a module, including attributes, without exporting to disk

Comintern

@ThunderFrame Perfect - thanks for tracking those down.

ThunderFrame

@Comintern took me 5 minutes. I hate to think how long it took Wayne to discover it all.

Comintern

True that.

this

2:28 PM

and just in case @Comintern didn't see it already there's a project: github.com/rubberduck-vba/Rubberduck/projects/15

FreeMan

@Malachi @IvenBach - video on Git? Would one of you repost the link? I can use all the help I can get...

Hosch250

2:58 PM

@FreeMan No, on Excel. Follow the chain, if you aren't on mobile.

FreeMan

ah, thanks!

I get the purpose of using Right$() to avoid the implicit type conversions. What was the point of creating Right() with the implicit type conversion, and is there a time when one might actually, intentionally, want to use it?

Duga

[rubberduck-vba/Rubberduck] BZngr added label PR-Status: WIP to pull request #4127: Unreachable Case Inspection: follow-up

Mathieu Guindon

@FreeMan if there is, I don't know what it is. if anything we should hijack _B_VAR_RIGHT calls and redirect them to _B_STR_RIGHT

this

@FreeMan I would say it's primarily for interoperability.

FreeMan

@this with what?

this

3:11 PM

As an example, a control's Value is usually Variant

Therefore, myLameTextBox.Value = Right("foo") would be more inline with what you are storing (a Variant data type).

Mathieu Guindon

so? you get the implicit conversion inside the function rather than at the call site, no?

this

Yes, I still would use the $ anyway myself.

Mathieu Guindon

(or rather, inside the Value mutator)

this

I know for sure that it's not available inside context of an Access query but that's not really the same thing as doing it in within VBA.

Mathieu Guindon

which? Right$?

this

3:13 PM

I'm also thinking about the case where you can get ByRef mismatch error.

oh you can do this in Access query: SELECT DoNotThis FROM AtHome WHERE That = Right("foo", 1);

but Right$ won't work there.

Mathieu Guindon

right but then that's Access-SQL, not VBA

this

Correct - even though it's capable of using VBA to evaluate expressions. Same problem w/ Excel formulas.

Hmm, is there a way to get a byref mismatch error when the expected data type is a variant?

Mathieu Guindon

what's a byref mismatch error?

this

oh you know when you have something like Public Function Foo(ByRef bar As String)

and you try to pass in Foo someVariant

in that case, implicit conversion won't happen and you get a error

Mathieu Guindon

I only ever recall seeing that error with arrays

this

3:17 PM

i do see that more often in the immediate window

since i'm going to be v = "foo": ?Foo(v)

Mathieu Guindon

why can't I have a refactor/rename @variable in SSMS?

FreeMan

Ctrl-h

:)

this

confirmed a MCVE:

Public Sub foo(bar As String)
    Debug.Print bar
End Sub

in immediate: v = "bar": foo v

Malachi

@FreeMan youtu.be/0nbkaYsR94c Joel

Mathieu Guindon

@this same here

this

3:24 PM

What I am not sure is if one could ever get a byref mismatch error for a Variant

Mathieu Guindon

Sub test(ByRef bar As String)
    Debug.Print bar
End Sub

Sub t(ByVal v As Variant)
    test v
End Sub

sheet1.t "bar" -> byref arg type mismatch

Right.

@Malachi thanks

yw

So in those use cases, you might need to use the variant returning types rather than string-typed alias. At least that's what I thought why we default to using variant-typed functions.

FreeMan

3:30 PM

So, if you're writing good, strongly-typed code, you have to think long and hard to contrive an example of when you'd want to use the variant-typed functions. i.e. they're there as a crutch for those who don't write good code because VBA makes it so easy to do so

this

Basically, yes.

FreeMan

well, it's good to get that cleared up!

:)

this

The original idea was to "not worry about data types, just get on with what you are trying to solve"

which is all good and well until you start writing nonsense like 1 + "1"

or even #5/3/80# - 3.25

FreeMan

which created a generation of poor coders who really weren't coders, and really, that created the entire demand for SO, so, I guess, it's good in the long run.

/always looking for the silver lining

this

According to Edgsar, I should never be a coder, so....

FreeMan

3:32 PM

meh, what's he know, anyway...

Googles "Edgsar". Gets no satisfaction.

this

> I think of the company advertising "Thought Processors" or the college pretending that learning BASIC suffices or at least helps, whereas the teaching of BASIC should be rated as a criminal offence: it mutilates the mind beyond recovery.

en.wikiquote.org/wiki/Edsger_W._Dijkstra

oh also this gem:

> It is practically impossible to teach good programming to students that have had a prior exposure to BASIC: as potential programmers they are mentally mutilated beyond hope of regeneration.

FreeMan

The subtle difference between Edgsar and Edsgar

also:

> FORTRAN's tragic fate has been its wide acceptance, mentally chaining thousands and thousands of programmers to our past mistakes.

this

I speel gooder, ok?

FreeMan

otay

> I mean, if 10 years from now, when you are doing something quick and dirty, you suddenly visualize that I am looking over your shoulders and say to yourself "Dijkstra would not have liked this", well, that would be enough immortality for me.

Transcript for