VBA Rubberducking

FreeMan

8:00 PM

@Vogel612 just to be sure!

^ #LateToThePartyAsUsual...

Vogel612

~sigh. 10 open issues assigned to me

Duga

[rubberduck-vba/Rubberduck] build for commit ca691d55 on unknown branch: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] codecov[bot] edited comment on pull request #4924: Remove dymanic calls from resolver

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4924?src=pr&el=h1) Report
> Merging [#4924](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4924?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/63e529a2bd25f73822bcea037c83b4001b5aa800?src=pr&el=desc) will **increase** coverage by `<.01%`.
> The diff coverage is `86.32%`.

```diff
@@ Coverage Diff @@
## next #4924 +/- ##
=======================

[rubberduck-vba/Rubberduck] build for commit ca691d55 on unknown branch: 64.53% (target 0%)

FreeMan

Is there any particular penalty for using Iif() in VBA (not WorksheetFunction)?

Vogel612

@FreeMan yes.

both expressions are evaluated

Duga

[rubberduck-vba/Rubberduck] codecov[bot] edited comment on pull request #4924: Remove dymanic calls from resolver

> # [Codecov](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4924?src=pr&el=h1) Report
> Merging [#4924](https://codecov.io/gh/rubberduck-vba/Rubberduck/pull/4924?src=pr&el=desc) into [next](https://codecov.io/gh/rubberduck-vba/Rubberduck/commit/63e529a2bd25f73822bcea037c83b4001b5aa800?src=pr&el=desc) will **increase** coverage by `<.01%`.
> The diff coverage is `86.32%`.

```diff
@@ Coverage Diff @@
## next #4924 +/- ##
=======================

FreeMan

8:11 PM

ah. I knew there was something. This is fairly simple, foo = iif(bar<0,0,bar), so it's divide by zero and invalid use of null safe.

Mathieu Guindon

@FreeMan same as any other function call :)

FreeMan

if yer stupid, it'll bite?

Mathieu Guindon

if yer givinit expressions, VBA will evaluate them before passing their respective result as arguments

IIf isn't "evaluating both expressions" - it doesn't even get to see the expressions

FreeMan

Can I not wildcard white listed identifiers?

Vogel612

@FreeMan time for you to pick up C#?

FreeMan

8:17 PM

I still get loads of Consider renaming 'Sheetxx' inspections. I can't rename these because the sheets are dropped & recreated on every execution.

@Vogel612 I'm getting my head wrapped around git. That's my first step...

TBH, I know everyone appreciates my sarcastically given title as head of QA, but I really do feel like all I do is complain about stuff. I get that finding the bugs is important, but I do want to do my share of fixing them, too.

@FreeMan OK, I could rename them, but next time I run the code, the warnings will come right back...

IvenBach

@Vogel612 You got hot a minute to help me unFCK my git folder?

Vogel612

sure

Mathieu Guindon

@FreeMan create them with a better name?

FreeMan

@MathieuGuindon I thought that code name could only be changed by hand in the VBE...

Mathieu Guindon

hmm yeah, or through the VBIDE API

IvenBach

8:21 PM

@Vogel612 I wanted to remove a file that was removed in a directory. I used git rm . -r and that wasn't the correct command. That removed everything that had already been pushed to my repo.

I have commits that I want to keep that need to be pushed. Is the correct way to stash them and pull from my repo?

Vogel612

run git checkout -- .

M.Doerner

Making the resolver non-dynamic has one more positive aspect: analyzing the performance in a profiler is a lot more straight forward.

FreeMan

and now that I've stumbled & bumbled my way into the VBE API to export/import modules, I'm a certified expert in finding my way around docs to figure out how to change worksheet code names!

Vogel612

wait ... @IvenBach the working directory is now empty except for the .git folder, right?

FreeMan

@IvenBach ouch! BTDT, lost 11TB of data!!

IvenBach

8:24 PM

@Vogel612 the .git directory and the new directories that I had added and want to push to the repo.

Vogel612

what does git status say?

Mathieu Guindon

@FreeMan github.com/rubberduck-vba/Rubberduck/issues/4504

IvenBach

$ git status
On branch master
Your branch is ahead of 'origin/master' by 5 commits.
(use "git push" to publish your local commits)

nothing to commit, working tree clean

Vogel612

that's confusing ...

IvenBach

looking at git log I see the commits I want to push to my repo.

Vogel612

8:25 PM

if git checkout -- . does not fix it, do git reset HEAD and retry

this

Just out of curiosity - before pattern matching was a thing, would have Visit<T> worked?

IvenBach

git checkout -- . is saying to checkout everything from the index (committed&pushed repo)?

Vogel612

@this no, because we still wouldn't have known the correct closed type

@IvenBach yes

FreeMan

@MathieuGuindon patiently waits for that to hit someone's priority list and/or for him to learn 'im some C#

Vogel612

@FreeMan that's marked as [ignored], so I don't think it will hit anyone's priority list any time soon

IvenBach

8:28 PM

:barf: refusing to merge unrelated histories.

FreeMan

details... :/

I'll let you focus on Iven - much more important topic than mine...

Vogel612

@IvenBach hmm ...

aight, let's just nuke the thing

git reset --hard HEAD

you should be able to push just fine, btw.

IvenBach

will that delete the directories & files I created?

Vogel612

@IvenBach but you deleted them with rm did you not?

IvenBach

That only removed those that had already been pushed to my repo.

I still had my directories and files I'd created, committed and wanted to push.

Vogel612

8:31 PM

and checkout did not restore them?

IvenBach

git checkout -- . did restore them.

Vogel612

so what is the issue left now?

IvenBach

My branch and origin/master have somehow diverged.

$ git status
On branch master
Your branch and 'origin/master' have diverged,
and have 6 and 3 different commits each, respectively.
  (use "git pull" to merge the remote branch into yours)

nothing to commit, working tree clean

$ git pull
fatal: refusing to merge unrelated histories

Vogel612

could you graph them for me? git log --oneline --decorate --graph master origin/master

IvenBach

$ git log --oneline --decorate --graph master origin/master
* 597ae31 (HEAD -> master) Perimeter and Area practice sheet, Part2
* f664d11 Fractions practice sheet, Part1
* 72e05e1 Shapes and modifier keys post
* 493fb58 Remove Thumbs.db
* 44d1cc6 Add .gitignore file
* 9088076 Initial commit
* 8e6bf8c (origin/master) Home page images
* c0b7dcc Published blog posts
* dbada7c Initial commit

Vogel612

8:35 PM

wat

wow, wtf did you do?

IvenBach

Unlocked the easter egg?

:shrug:

I don't know what I'm looking at.

Vogel612

this looks like your origin doesn't even have anything to do with the repo you're currently in

Mathieu Guindon

O_O

this

what if the origin isn't the origin you think it should be?

(not sure if that is even possible)

IvenBach

origin is indeed where it should be pointing to.

Vogel612

8:38 PM

if you run git show --format=%P 9088076 I assume that it will be empty

IvenBach

Nope it's populated with lots info.

this

try git reset --unIvenBach?

Vogel612

sigh... that's because git partially ignores the format specifier

add an -s to the options, please

IvenBach

Laymans terms git has disconnected my working directory from origin? I thinks they're separate distinct things?

Vogel612

I'm trying to confirm that right now ...

note how you have two initial commits

IvenBach

8:44 PM

$ git show --format=%P -s 9088076 shows nothing

Vogel612

yea, that confirms my suspicion

there's literally no link between origin/master and master

they're completely unrelated histories

IvenBach

I'm so good at Funucking stuff up.

Vogel612

when you try to merge (and pull always includes a merge), git searches for the "merge base" between the commits you merge by going back through history

IvenBach

Since they don't have a common ancestor they are considered distinct?

Vogel612

exactly

The only way this can happen is if you didn't clone your repo from origin

instead you probably ran git init and git remote add.

Then you started working and commited some stuff

Duga

8:47 PM

[rubberduck-vba/Rubberduck] daFreeMan opened issue #4926: '@Ignore ImplicitByRefModifier doesn't work with a line continuation

IvenBach

I don't think that's what I did.

Duga

> **Rubberduck version information**
Version 2.4.1.4666
OS: Microsoft Windows NT 10.0.15063.0, x64
Host Product: Microsoft Office 2016 x64
Host Version: 16.0.4822.1000
Host Executable: EXCEL.EXE

**Description**
Adding `'@Ignore ImplicitByRefModifier` to a function declaration does not work if there is a line continuation (`_`) in the declaration.

**To Reproduce**
Steps to reproduce the behavior:
1. Write a function declaration with an implicit `ByRef`
2. Edit in a line continuatio

Vogel612

voilá, histories are diverging

Duga

[rubberduck-vba/Rubberduck] daFreeMan added label bug to issue #4926: '@Ignore ImplicitByRefModifier doesn't work with a line continuation

Vogel612

git clone will always do a pull

IvenBach

8:48 PM

mkay. I guess best option for me to then is clone my gh repo and manually move over the directories to that new one. Push to repo and see what happens.

Vogel612

and because the commit messages are all different, I don't think you inadvertently rewrote the commits

@IvenBach that would be one option. The other would be to manually merge the histories

you said your wd is clean?

IvenBach

git its saying wd is clean.

Vogel612

you could do git checkout -b fix-mess origin/master and then manually merge the histories

IvenBach

That's probably going to be harder than copying them over to a newly cloned repo.

Vogel612

that would entail running git cherry-pick 44d1cc6 493fb58 72e05e1 f664d11 597ae31

git might shout at you a bit about conflicts, which you'd need to resolve

FreeMan

8:51 PM

There's a fix-mess parameter??? How awesome is that???!?!?!?!111!/!?

Vogel612

@FreeMan there isn't

it's a branchname

IvenBach

@FreeMan It's an alias for the Iven-Done-Fucked-It-Up-AGAIN command.

Vogel612

remember git checkout -b is a shorthand for git branch && git checkout

OBTW @MathieuGuindon will you create a preview image?

for the new github feature thingy?

Mathieu Guindon

yeah sure

todo++;

Vogel612

ugh... the template recommends leaving a "40 pt border" ...

Mathieu Guindon

8:57 PM

wut

i.e. over half an inch??

Duga

[rubberduck-vba/Rubberduck] Vogel612 created comment on issue #4926: '@Ignore ImplicitByRefModifier doesn't work with a line continuation

> Yea, I assume that's because a part of the logic for ignore annotations may fall back to checking the annotation based on the module's lines as retrieved from the VBE, which only respects physical lines, not logical lines...

Vogel612

@MathieuGuindon I'm mostly miffed that they didn't use ems, (px would've been the best metric tho)

IvenBach

@Vogel612 I ran through it manually as that was easier. I do need to use cherry-pick in more than just a example scenario. Thanks for the help.

Vogel612

how was this an example scenario?

IvenBach

This wasn't. I've only ever done them as part of examples.

Vogel612

9:00 PM

oh well, cherry-picking is usually not needed for most work with git anyways

either use rebase or have a workflow that doesn't require moving commits around

IvenBach

I know it exists. Have done it as part of examples. I just knew moving them manually and adding them was simpler (for my at least) this time.

Duga

[rubberduck-vba/Rubberduck] MDoerner created comment on issue #4926: '@Ignore ImplicitByRefModifier doesn't work with a line continuation

> Quite exactly. If there is any physical line with code in between, an annotation does not apply. This is how annotations currently work.

mansellan

10:22 PM

@FreeMan TBF, they probably didn't block it. They just bought a list that included it. Enterprises...

FreeMan

@Vogel612 d'oh! I should know that

mansellan

@M.Doerner Very interesting to hear that. Contrast to the LSP approach, which (iiuc) sends keystrokes across to the server, which can then choose when to send updates. Typically, it would wait for a certain dwell time (for example to wait for the user to stop typing), then request a build.

this

@mansellan I'm trying to envision how Liskov's Substittion Principle applies. ;-)

mansellan

10:38 PM

Heh no, Language Server Protocol. New standard that VSCode uses, and increasingly others too.

#AcronymConflict

langserver.org

If we were ever to do out-of-process, we should [consider/do] this.

Not that Liskov doesn't rock, ofc :-)

tl;dr - the server keeps an in-memory representation of the source code, independent from how it is represented on disk. It then uses that to service requests for symbols, documentation, locations...

oh btw 'server' in this context just means 'another process', not an actual bit of iron.

this

Makes sense.

I would point out that if we have our custom code pane, there would be no reason to not just export the attribute pass files (for VBA project) and just work with them directly and import into the VBE lazily.

hopefully importing is much quicker than exporting.

mansellan

Yeah but OOP solves more than just that - like not eating the Office memory limit for 32-on-64...

this

is there a LSP implementation we can build on?

mansellan

sorry, OOP - Out Of Process not Object Oriented Programming #AcronymConflict :-)

this

I figured. :)

mansellan

10:52 PM

@this there are many implementations. Ones that we can build on, uh, maybe?

IMHO, not a problem for right now anyway, just something to consider if we decide that OOP is the way forward.

this

Even if we do the OOP way, it still can't happen without a custom code panes...

mansellan

@MathieuGuindon do you have a view on whether going out-of-process is a good idea?

@this gah clashed. why?

could we not hook ketstrokes?

this

you really want to do IPC w/ VBIDE?!?

Yes we already do that (hooking the keystroke) - we're just doing it with a host that really does not like having its bits touched.

2

mansellan

Why would VBE get in the way of IPC?

had to star that!

this

The real problem is that it has lot of quirks and some hacky workarounds that makes it hard for us to just do a straight implementation.

mansellan

10:57 PM

Hook the keystoke, dispatch to a server (async) - what could go wrong (genuine question)?

@this Yes, agreed. But in theory...

Vogel612

@mansellan VBE prettifying behaviour is the smallest edge case to consider there

we want the VBE as authoritative source on what the state of the code is.

mansellan

@Vogel612 eesh, yes. retreats to box

Vogel612

then there's stuff like the break mode

basically OOP has a host of complications

mansellan

@Vogel612 c.f. AC

gah

Vogel612

the clear advantage is the lift of memory restrictions and an avenue for using a proper DB

mansellan

11:00 PM

WTF is the VBE so damn unpleasant to work with...

this

because it's a product of 90s?

Vogel612

the disadvantage is that we still need something to hook into the guts of the VBE that's more than "send keystrokes to [this named pipe]"

Also OOP doesn't alleviate the mess around window messages and dockable window hosts

mansellan

@this I think you had the right idea.. replace the damn thing outright.

Vogel612

which we can't do because of bracketed expressions

like ... [A1].Value is perfectly valid VBA (albeit clearly using an Excel host)

and completely replacing the VBE implies completely breaking any form of bracketed expression

I fear going OOP is just going to increase the communication overhead without providing enough benefits to be worth the effort

mansellan

@Vogel612 But... the VBE must have solved for this. There isn't an Excel VBE, Access VBE, AutoCAD VBE...

Vogel612

11:04 PM

it will also complicate the setup significantly because we need to maintain a language server that behaves correctly

@mansellan yes and no... VBA SDK is the magic keyword here

mansellan

@Vogel612 But VSCode proves that it's sufficiently efficient?

Vogel612

@mansellan VSCode doesn't deal with the VBE though...

IvenBach

56K sloc app.js file. That's way worse than any VBA module I've seen...

Vogel612

it's not "send this URL to the language server, that will parse an initial representation, then keep stuff synchronized by sending key events over"

again: we probably need the VBE as authoritative code source

which means: every parse must start from scratch (to some extent)

mansellan

@Vogel612 mulls

this

11:07 PM

@Vogel612 You seem to be saying that even if we had a custom code pane where we work with the exported file directly, this still wouldn't fly?

I figured that with the custom code pane in control, we can just work with the exported file directly, cut out the code pane pass altogether, and simply import changes on-demand.

Vogel612

@this if we want to be able to correctly deal with bracketed expressions, bracketed identifiers and the general quirks of the VBE, I think it's impractical

not impossible, just waaaaayy too much effort

mansellan

@Vogel612 Sorry if I'm being thick (not unusual...) The VBE is not a unicorn, therefore its special sauce must be replicable. Is the problem that too much of its operation is undocumented?

Vogel612

"undocumented" is putting it mildly ...

this

I'm not sure I follow why bracketed expression is a problem.

mansellan

@Vogel612 well yes, but....

Vogel612

11:08 PM

@this pretty-printing around them is ... weird

this

Ah, you want the pretty-printed version

which we can't have w/ a attribute file.

Vogel612

of course. That's the canonical one

mansellan

IIUC (unlikely), pretty-printing is a side-effect of the operation of the VBE. Hit enter, code gets compiled, which then gets round-tripped to the IDE.

Vogel612

that's my understanding as well, yes

mansellan

So, we "just" need a p-code parser...

ANTLR... (joking!)

Vogel612

11:11 PM

if you give me a proper grammar for p-code that's host-agnostic ... sure. Why not?

though p-code is a binary format, IIRC

mansellan

p-code has been REd, so...

(can find links if need be)