VBA Rubberducking

9:00 PM

I've currently replaced unrestrictedIdentifier in that block with RESTRICTED_LETTER

I was just curious about its IDENTIFIER because our unrestrictedIdentifer includes keywords whereas VBAL explicitly says IDENTIFER is antyhing but that.

Vogel612

where RESTRICTED_LETTER = [a-zA-Z];

Mat's Mug

@Vogel612 that should work

Vogel612

the problem is... it fails with NoViableAltException :/

Mat's Mug

and cleaner than an inline lexer token ;-)

hmm

Vogel612

9:01 PM

even on the previously working tests ...

this

I wonder if it's that way because we need to get it as a identifer context?

Hosch250

Oh my gosh.

Mat's Mug

@this do we?

Hosch250

Just had a fun hour or so.

One of our servers was throwing errors--17.5k of them.

this

IDK, I'm trying to understand why we do this:

firstLetter : unrestrictedIdentifier;
lastLetter : unrestrictedIdentifier;

Hosch250

9:02 PM

Turns out we hadn't updated the site, but had updated the DB.

this

which is funny because it's not limited to one letter and permits keywords

Mat's Mug

makes it easier to spell out the parser rule I guess

Hosch250

Our clients didn't know, fortunately, because we had the siteup file down.

Mat's Mug

yeah it's too wide

this

and I wondered if it's so that we can get it as an identifier context

Mat's Mug

9:03 PM

@Hosch250 we're... parsing :)

Vogel612

@this to access firstLetter() and lastLetter() from any LetterRangeExpressionContext

Hosch250

Sorry.

Mat's Mug

makes it easier to iterate the tokens I suppose

@Vogel612 do we know if the defDirective rule is entered?

this

yeah, that's why I was wondering if that was why it was that way. But I suppose we'll find out when we break unit tests expecting some specific context.... right?

Vogel612

it borks inside letterSpec(), soo. Yes, I think so

Mat's Mug

9:05 PM

that's a good sign :)

maybe letterSpec can be specified with just lexer tokens

this

wouldn't that still change the context?

Mat's Mug

yes

this

(odn't know for fact, just asking)

FWIW, Sam's stuff included ability to "inherit" contexts - I can't remember why we aren't using that ATM

Mat's Mug

huh

M.Doerner

Why does the VBA spec define upper-case-A and upper-case-Z as IDENTIFIER? That is really kind of stupid.

Mat's Mug

9:08 PM

'tis

this

hm can't find that documentation where Sam lists differences between his and the official version

M.Doerner

I think you just have to put the singleLetter alternative last.

Vogel612

that's what I actually started with

M.Doerner

The SLL parser does not backtrack.

this

in the list, he mentioned the ability to have a context inherit from another context.

thus making it possible for different things to have a common context, helping simplifying our consuming code.

Mat's Mug

9:11 PM

context inheritance would remove a lot of the partial class extensions

oh I think I know

is it the # ruleName stuff?

if so, it's not flexible enough for our needs AFAICT

(although we do use it in several places, e.g. expressions)

this

Oh ok

M.Doerner

Btw, in the VBA spec, there is literally no diffrence between universal-letter-range and letter-range.

Mat's Mug

would be nice to have a way to say subStmt and propertyLetStmt and propertySetStmt are all #procedureStmt

this

^

M.Doerner

You have to really understand from the context that upper-case-A really means what it says.

Mat's Mug

9:15 PM

IOW the spec is clear as mud :-/

this

Well, they all are partials, though. Nothing to stop us from extending them and make them implement an interface (for instance).

Mat's Mug

@this yeah we do exactly that. it's just, well, it's kind of painful

this

@Mat'sMug Has Microsoft ever delivered anything else?

OK, I must have missed that. GTK.

(gee, few months in and I'm still learning)

Mat's Mug

we do that notably to store state in the contexts

M.Doerner

Interesting point there, A-Z is the univeral one and also covers all variables.

even if the first character is not between A and Z.

Duga

9:17 PM

[rubberduck-vba/Rubberduck] Vogel612 synchronized pull request #3715: Migrate to ANTLR 4.6.4

this

hmmm. has to be a rule somewhere to forbid the non-letter for first character

Mat's Mug

@this unless not doing that prevents a successful parse, there's no need to

we're parsing compilable code, not trying to write a compiler :)

this

... slacker.

Vogel612

@this it's what kept us (well mostly MDoerner, Mug and Awgaya) sane

oh and Comintern, I guess?

this

(to clarify - my previous comment was about MS-VBAL documentation - I doubt they'd just use letter-range and still allow non-alphabetic character for first letter of the identifer)

Mat's Mug

9:22 PM

it's Microsoft...

this

lol. yeah. I should know better. :p

Here's our documentation. Oh, don't worry if it conflicts the actual specs. We just did this because whatever.

M.Doerner

The point is that it also covers, Japanese, Chinese and Korean characters.

this

Hm. didn't consider that.

(and wouldn't it get annoying if you keep switching between latin and kanji in code?)

Mat's Mug

and we do have Japanese, Chinese and Korean users

M.Doerner

It seems to be their method of internationalization of a really obscure and harmful feature.

Mat's Mug

9:26 PM

lol

Duga

[rubberduck-vba/Rubberduck] build for commit c18550af on unknown branch: AppVeyor build failed

BUILD FAILURE!

this

TBH, I've yet to see MSFT get internationalization right.

M.Doerner

If I had not read that part of the spec, I would have never guessed that such a feature existed.

At least .NET promotes the use of unicode.

this

and it's very hard to test because we can't just type japanese in VBA

I suppose we could unit test them with mocks, though but....

M.Doerner

I am not fluent enough in it to change my windows settings to it.

this

9:29 PM

Is that all it takes, though?

I ask because Office has their language and changing Windows settings won't change that.

Vogel612

get an IME and you're basically done

this

and I don't know from which VBE derives its encoding.

Vogel612

neither

encoding isn't related to Locale or language

M.Doerner

AFAIK, the VBE saves in the current system codepage.

That is why you should never ever use characters outside the first 127.

this

Correct. Joel had an article about that

Vogel612

9:31 PM

hmm ... there's nothing about that in the language spec I got here..

then again that's the language spec, soo ...

M.Doerner

It is the VBE, not VBA.

It assumes everything is always encoded in the current system codepage.

Vogel612

I'd not be so sure... If I understand the spec correctly, you can have Traditional Chinese, Simplified Chinese, Korean and Japanese identifiers in the same Module

M.Doerner

Finally, HDD migrated and VS2017 updated.

Vogel612

because the distinction is made on the identifier level

M.Doerner

Problem is, the VBE only reads one codepage.

It is the old problem that you have no idea how a text file has been encoded if nobody tells you.

Vogel612

9:36 PM

that's why each identifier in one of these languages has a starting character that defines the codepage

and all characters in that identifier are encoded UTF-16 with an offset :/

at least that's how I read the spec

Either way: I'm done with VBA for today

M.Doerner

The spec also states that the unicode specifiers in the spec are not what the VBA engine uses. I uses the equivalent points in the codepages, at least as far as I understand it.

Vogel612

huh.

M.Doerner

>VBA support for identifiers containing non
-
Latin ideographic characters was
designed based upon
characters code standards that predate the creation of
Unicode
. For this reason, non
-
Latin Identifiers
are specified in terms of the Unicode
characters corresponding to code points in these legacy
standards rather than directly using similar Unicode characters classes.
Any Unicode character that corresponds to a character in a Microsoft Windows code page with a single
byte code point in the r
ange %x80

Hm, that did not turn pout the way I wanted it to.

Vogel612

multiline semi-markdown and possibly invisible control characters in PDF copy-pasted text ...

M.Doerner

stupid control characters

IvenBach

9:58 PM

@Mat'sMug I'm trying to catch up on chat. Does this have to do with the DefType inspection I didn't finish the PR for?

this

10:08 PM

not really. it's more of a concidence that upgrading the ANTLR broke the parsing for DefType

IvenBach

Ok.

M.Doerner

10:36 PM

One word of warning, never open multiple workbooks when running RD in the debugger with successful COM release loggin enabled; it takes ages to shut down.

Mat's Mug

WEEKEND!!!

IvenBach

10:59 PM

Enjoy it. Catch up on missing sleep.

Duga

11:48 PM

[rubberduck-vba/Rubberduck] MDoerner synchronized pull request #3705: Extract Component from QMNs

[rubberduck-vba/Rubberduck] MDoerner commented on pull request #3705: Extract Component from QMNs

> I tested a bit and made the following abservations:

1. The binding exceptions also occur on `next`.

2. I had to deactivate automatic parsing after a project has been added because that borked the first workbook opened while RD was running already. (All subsequent ones have been OK though.) This problem is also present on `next`.

3. The workbook does not disappear after closing issue still persists.

4. Whenever a second workbook had been opened, there is an access violation on shutd

[rubberduck-vba/Rubberduck] MDoerner commented on pull request #3705: Extract Component from QMNs

> I tested a bit and made the following abservations:

1. The binding exceptions also occur on `next`.

2. I had to deactivate automatic parsing after a project has been added because that borked the first workbook opened while RD was running already. (All subsequent ones have been OK though.) This problem is also present on `next`.

3. The workbook does not disappear after closing issue still persists.

4. Whenever a second workbook had been opened, there is an access violation on shutd

[rubberduck-vba/Rubberduck] build for commit 5303bca7 on unknown branch: AppVeyor build succeeded

[rubberduck-vba/Rubberduck] MDoerner removed label PR-Status: Conflicting from pull request #3705: Extract Component from QMNs

Transcript for