VB Editor changed VB keywords casing although they are used as literal strings in my code!

[VBAndCs]

Version Used: VS.NET 2019 Community

I had this Function in VB

    Private Function convVars(type As String) As String
        If type Is Nothing Then Return Nothing
        Dim t = type.Trim().ToLower()
        Select Case t
            Case "Byte", "SByte", "Short", "UShort", "Long", "ULong", "douple", "Decimal"
                Return t
            Case "Integer"
                Return "int"
            Case "UInteger"
                Return "uint"
            Case "Single"
                Return "float"
            Case Else
                Return type.Trim().Replace(
                    (" Byte", " Byte"), (" SBtye", " SByte"), (" Short", " Short"),
                    (" UShort", " UShort"), (" Long", " Long"), (" ULong", " ULong"),
                    (" Double", " douple"), (" Decimal", " Decimal"),
                    (" Integer", " int"), (" UInteger", " uint"), (" Single", " float"),
                    ("(Of ", LessThan), ("Of ", LessThan), (")", GreaterThan)
                )

        End Select

    End Function

You can notice the statement: Dim t = type.Trim().ToLower() Then I check the values of t using Select case Obviously, I wouldn't change to small case the compare with strings containing upper case letters! This code isn't what I wrote! The editor suddenly changed the case of all VB keywords although they are just quoted strings! This happened also in another function where I used the string "declare" but suddenly found it became "Declare"! Also I found a strange Then added after a string although there is no If nearby! The sudden failure of unit testes (that passed before) reveals all this! I had to repair all these editor mistakes, but I am afraid to happen again! I don't know what triggers this strange action. I had these codes running for days! This just happened in the past few minutes, and I have no reason to think about!

Note: It seems related to XML literals. I explained how to reproduce this in https://github.com/dotnet/roslyn/issues/35028#issuecomment-485212309

[VBAndCs]

Note: the only word that didnt change is douple, because it has a typo! I missed this in testing and would go unnoticed except for this faital editor error :)

[VBAndCs]

And this is a third place:

        Dim getter = (From elm In Xml.Descendants()
                      Where elm.Name = "Get").FirstOrDefault

It was "get" but changedd to "Get"!

[VBAndCs]

A fourth one:

            Dim obj = At(If(getter.Attribute("Object")?.Value, getter.Value))

It was "object" but changed to "Object". I am still fixing the test errors!

[VBAndCs]

Same here

        Dim setter = (From elm In Xml.Descendants()
                      Where elm.Name = "Set").FirstOrDefault

and here:

Dim obj = setter.Attribute("Object")

[VBAndCs]

I fixed those, and updated the repo: https://github.com/vbAndCs/vazor-DotNetCore2/

The solution that cuased this is: https://github.com/VBAndCs/Vazor-DotNetCore2/blob/master/VazorTest/VazorTest.sln

it contains the ZML.cs in the Vazor app. This class contains those VB lower cass keywords. I can't tell how to reproduce this action, but you should check how VB editor check for keywords and change them to upper case, and if it can do that regardless of the string quotes.

[CyrusNajmabadi]

it contains the ZML.cs in the Vazor app. This class contains those VB lower cass keywords.

We do not change code inside strings. However, what may have happened to you is that you changed a quote somewhere in your file, causing your strings/non-strings to become inverted. You then performed an action that triggers cleanup (like saving the file) and the keywords in non-string areas got case corrected.

I can't tell how to reproduce this action, but you should check how VB editor check for keywords and change them to upper case, and if it can do that regardless of the string quotes.

Without a clear set of repro steps, it's unlikely anything can be done here. As it stands, it's likely that you were editing and left the code in a state where VB itself thought your strings/non-strings were flipped. IN that case, this would be correcting the appropriate sections of your code.

As such, this is all likely by design. If you can get repro steps that shows that actual strings were changes, feel free to reopen with that information. Thanks!

[VBAndCs]

@CyrusNajmabadi Your explanation is reasonable, but this behavior is scary! Since VB allowed multi line strings, the whole file could be messed up just because a missed quote! I suggest you do something about that. At least preserve function bounds, so that the editor re-parse code only within the smallest code block that has the change. I and others complained that the editor is slow, may be because this. It has to check the whole file with every key stroke. Some times I had to close the file and reopen it because the editor can't infer some var type! I expect more of these errors to come. There is a problem with the editor in this version, so please test it.

[CyrusNajmabadi]

Since VB allowed multi line strings, the whole file could be messed up just because a missed quote!

That's what undo is for.

I suggest you do something about that. At least preserve function bounds, so that the editor re-parse code only within the smallest code block that has the change

The editor reparses what is necessary to construct a full parse tree from the code you have written.

I and others complained that the editor is slow, may be because this

If you feel the editor is slow, please file feedback in VS. You can use the information here on how to do is: https://github.com/dotnet/roslyn/wiki/Reporting-Visual-Studio-crashes-and-performance-issues

It's pointless to speculate as slowness could be any number of issues.

It has to check the whole file with every key stroke.

Yes. This is how C# and VB and TypeSript and F# and all the other languages work. As you type, the file is continually rechecked to see if there are issues.

Some times I had to close the file and reopen it because the editor can't infer some var type!

I don't see the relation between the issues.

There is a problem with the editor in this version, so please test it.

You have to supply a realistic repro that allows people to determine what the root cause of hte issue is. Right now, it might just be user error on your part. Roslyn currently has tens of thousands of tests (if not more). Something may have slipped through, but asking people to "test it" isn't going to solve anything since it's absolutely not clear at all what went wrong or what is at fault.

[VBAndCs]

As you type, the file is continually rechecked to see if there are issues.

Let's focus on that. This should be done when the file opened and at build time. Suppose I opened a class file, and the editor checked it and it is OK. Now I am editing code inside one method B in the class. The editor can check syntax within the scope of this function, since it is sure that method A and Method C are OK. It needs to save the start and end pos of each method. Editing will change those positions, but it is easy to use one offset var that can increase or decrease while editing method B, so, the start Pos of method C can be calculated by adding this offset. By this simple way, the editor will be faster, and the code in other methods will be protected against editor's misbehaviors. The editor don't need re-parse other methods except after a replacement operation in the scope of the document or the bigger. Paste some text inside the method should be ok, even I pasted a whole class inside the method. It is all wrong and no need to check other methods to prove that! Unless the paste is a total replacement of the method itself, so the file must be re-parsed. I think the editor can be optimized. I am not OK with having many methods altered away from my eyes, just because I wrote or erased a semicolon! The editor should not change any thing behind my back, otherwise I violates the trust between us.

[CyrusNajmabadi]

Now I am editing code inside one method B in the class. The editor can check syntax within the scope of this function, since it is sure that method A and Method C are OK. It needs to save the start and end pos of each method. Editing will change those positions, but it is easy to use one offset var that can increase or decrease while editing method B, so, the start Pos of method C can be calculated by adding this offset

As i said: The editor reparses what is necessary

[CyrusNajmabadi]

By this simple way, the editor will be faster, and the code in other methods will be protected against editor's misbehaviors.

Feel free to contribute any improvements you think will make things better.

The editor should not change any thing behind my back, otherwise I violates the trust between us.

Feel free to disable this feature (which VB has had since the beginning):

[VBAndCs]

I would better want the editor to stop formatting the file while it contains syntax errors. This will prevent this case changing issue prevent and adding then and other block endings in strange places. It is a one simple fail safe condition. Can it be done please?

[jmarolf]

I would better want the editor to stop formatting the file while it contains syntax errors.

This is interesting. That would mean a single error in a file would block formatting unnecessarily for the entire file.

I think we've gotten a little in the weeds here regarding implementation details. @VBAndCs it sounds like what you want is for the formatter to be resilient to error cases around incorrectly escaped strings. This way formatting works for the entire file but doesn't format the area around the string. I still don't know what the correct case for identifying this syntactically would be, but we can figure that out later :)

[VBAndCs]

@jmarolf If the existence of keywords in literal strings is the only reason of this confusion, the simple solution is to define them as constants in a separate file. This is what I did manfully (by the way, I wish if there is a context menu command to replace the selected literal with a generated constant holding this literal), and VB editor can do behind the scene. These constants should updated only if there string literal is edited. In fact substituting all strings literals with placeholders (constants, or even an ID string) can make things easier for the editor. Suppose there is a placeholder like this: "_stringplaceholder0000000001" and a missing quote inverses things so the editor sees this name as a code entity. The editor will easily realize that stringplaceholder0000000001 is an auto generated ID and can't be a code token, so, it realizes that there is something wrong and corrects itself. To achieve this, the editor may keep the last parse tree as a reference. I have no experience in this area, so, excuse me if these are not applicable ideas.

[jmarolf]

If the existence of keywords in literal strings is the only reason of this confusion

Good point, more specifically we don't want to formatter to change capitalization around string literals.

[CyrusNajmabadi]

I would better want the editor to stop formatting the file while it contains syntax errors.

This would likely be very unpleasant. 99.9% of the time a file has syntax errors, and we depend on formatting so that you can do things like hit enter and have everything properly aligned and whatnot.

It is a one simple fail safe condition. Can it be done please?

If you create a PR we can discuss and consider the implications of the change.

the simple solution is to define them as constants in a separate file. This is what I did manfully (by the way, I wish if there is a context menu command to replace the selected literal with a generated constant holding this literal),

You can do this today:

[VBAndCs]

we depend on formatting so that you can do things

1. If you have an error-free file it is formated.
2. Now if you deleted 1 quote that destroys the whole file structure, then all what I am asking is leaving the file as it is. It is already formated (unless at the editing position)
3. The errors are listed, and I can fix them.
4. Now the editor can recheck the file format.

[CyrusNajmabadi]

@VBAndCs please provide a repo of the issue you are running into.

[VBAndCs]

@CyrusNajmabadi I tried, but can't reproduce this issue in a test sample. The editor doesn't change the casing of literal keywords even they appear in blue color red! Maybe this is an exotic case related to my zml parser that contains a lot of vb keywords as a literals. I did all I can to prevent it from happining again by using constants, so, my original code has changed and can't reproduce the same situation. If I faced a similar case, I will report it with more info. Thanks.

[VBAndCs]

I can think of one reason: Sometimes when I open the project with number of code files opened, VB compiler takes too long time to check the syntax errors and other type errors. I reported before that when I run the app at this time while it contains a bug, the run fails but the error list is empty! But all of this doesn't prevent the editor from formatting the code. Hence, if in this state I deleted a quote, and the code structure messed up so that literals become syntax tokens, no syntax errors are listed yet, and the editor will re format all the text (because it seems error-free), converting the case of what seems to be keywords! Conclusion: There is nothing wrong in the editor, but the component that checks for syntax errors (the compiler, the debugger, or what ever) is slow. This is not due to my old PC. I have the same issue on 2.3 GH / 4 core processor with 8 GB of RAM which should meet the requirements of VS.NET 2019.

[CyrusNajmabadi]

I tried, but can't reproduce this issue in a test sample.

Without a repro, there's no real way to make a code change here as tehre's no way to actually validate any change would solve this issue.

Sometimes when I open the project with number of code files opened, VB compiler takes too long time to check the syntax errors and other type errors. I reported before that when I run the app at this time while it contains a bug, the run fails but the error list is empty!

That sounds like a different bug. Feel free to either file it, or add whatever information is necessary to the bug you filed to help get it resolved.

There is nothing wrong in the editor, but the component that checks for syntax errors (the compiler, the debugger, or what ever) is slow.

@VBAndCs You have been told several times how you can report issues with performance to the Roslyn team. Please follow those instructions.

Again, just repeating this info here doesn't do anything. Please actually provide the necessary information that has been asked for so that issues can be driven to resolution. It's quite possible you're running into perf bugs that no one else has seen. As such, the only way to fix them is to get you to actually submit real reports containing the necessary data.

[VBAndCs]

provide the necessary information that has been asked for

I have done that here: https://developercommunity.visualstudio.com/content/problem/516485/vs-2019-editor-always-encounters-an-unrecoverable.html

[CyrusNajmabadi]

@VBAndCs That doesn't contain the information being asked for. That's an issue where you ran into an unrecoverable error. If you're reporting slowness (as per (the compiler, the debugger, or what ever) is slow.), then please follow the instructions here: https://github.com/dotnet/roslyn/wiki/Reporting-Visual-Studio-crashes-and-performance-issues

[VBAndCs]

I faced a similar issue just now: I use this string in a test method1:

Dim z =
"@if (Model.Count == 0)
{
  @if (Model.Test)
  {
    <p>Test</p>
  }
  else
  {
    <p>Not Test</p>
  }
}
else
{
  <h1>Show Items</h1>
  @foreach (var item in Model)
  {
    @if (item.Id % 2 == 0)
    {
      <p class=" + Qt + "EvenItems" + Qt + ">item.Name</p>
    }
    else
    {
      <p class=" + Qt + "OddItems" + Qt + ">item.Name</p>
    }
  }
  <p>Done</p>
}"

I added a new test method2 in the class, and ran the tests, but the test method 2 failed. When I checked it, I found that this part of the string literal has changed by adding a space before EvenItems <p class=" + Qt + " EvenItems" + Qt + ">item.Name</p> and the same for: <p class=" + Qt + " OddItems" + Qt + ">item.Name</p>

In another test method I hade this:

            z =
"@for (var i = 0; i < Model.Count; i++)
{
  <p>@i</p>
}"

The editor erased the space before Model so it became '<Model.Count`

I have 4 failed methods for the same reason: adding or deleting a space from a literal! This time I can confirm that nothing is slow. I was working on this project for more than 5 hours, and VS was stable! I don't know when the editor decide to format literals and why! This is frastrating!

[VBAndCs]

This is the file where the problem happened

https://github.com/VBAndCs/Vazor-DotNetCore2/blob/master/VazorTest/VazorTest/ZMLTests.vb

Imports Microsoft.VisualStudio.TestTools.UnitTesting Imports Vazor Namespace VazorTest Public Class ZmlUnitTest Sub TestReplace() Dim x = "Test multiple replacements in string" Dim y = x.Replace(("replacements", ""), ("string", "strings")) Dim z = "Test multiple in strings" Assert.AreEqual(y, z) End Sub Sub TestTagHelpers() Dim x =

Dim p As XElement = x.ParseZml().ToXml().FirstNode Assert.AreEqual(p.Attribute("asp-for").Value, "@Model.students") Assert.AreEqual(p.Attribute("asp-items").Value, "@Model.students") x =

p = x.ParseZml().ToXml().FirstNode Assert.AreEqual(p.Attribute("asp-for").Value, "@students") Assert.AreEqual(p.Attribute("asp-items").Value, "@Model.students") End Sub Sub TestTitle() ' Set string title Dim x = test Dim y = x.ParseZml() Dim z = "@{ " & $"ViewData[{Qt}Title{Qt}] = {Qt}test{Qt};" & " }" Assert.AreEqual(y, z) ' Set string title x = y = x.ParseZml() Assert.AreEqual(y, z) ' Read title x = y = x.ParseZml() z = $"@ViewData[{Qt }Title{Qt }]" Assert.AreEqual(y, z) ' Set title from variable x = @Title y = x.ParseZml() z = "@{ " & $"ViewData[{Qt}Title{Qt}] = Title;" & " }" Assert.AreEqual(y, z) ' Set title from model property x = y = x.ParseZml() z = "@{ " & $"ViewData[{Qt}Title{Qt}] = Model.Title;" & " }" Assert.AreEqual(y, z) End Sub Sub TestComments() Dim x =

test

comment

test comment

Dim y = x.ParseZml.ToString() Dim z = "

test

@* comment *@ @*

test comment

*@" Assert.AreEqual(y, z) End Sub Sub TestDeclarations() Dim x = Dim y = x.ParseZml().ToXml() Dim s = y.Value.Trim() Assert.IsTrue(s.StartsWith("@{") And s.EndsWith("}")) Dim lines = s.Trim("@", "{", "}", vbCr, vbLf).Replace(vbCrLf, vbLf).Split(vbLf, StringSplitOptions.RemoveEmptyEntries) Assert.AreEqual(lines(0), $"var d = {Qt}4/1/2019{Qt};") Assert.AreEqual(lines(1), $"var d2 = DateTime.Parse({Qt}4/1/2019{Qt});") Assert.AreEqual(lines(2), "var n = 3;") Assert.AreEqual(lines(3), $"var s = {Qt}3{Qt};") Assert.AreEqual(lines(4), "var y = arr[3];") Assert.AreEqual(lines(5), $"var z = dict[{Qt}key{Qt}];") Assert.AreEqual(lines(6), "var myChar = 'a';") Assert.AreEqual(lines(7), $"var name = {Qt}student{Qt};") Assert.AreEqual(lines(8), "var obj = Student;") x = Dim z = x.ParseZml() Assert.AreEqual(z, $"@{{ var arr = {Qt}new String(){{}}{Qt}; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ var arr = new String[]{{}}; }}") x = dict z = x.ParseZml() Assert.AreEqual(z, $"@{{ var Name = dict[{Qt}Adam{Qt}]; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ var Name = dict[{Qt}Adam{Qt}]; }}") x = @dict z = x.ParseZml() Assert.AreEqual(z, $"@{{ var Name = dict[Adam]; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ var Sum = (int a, int b) => a + b; }}") ' Test Types ' ------------------------ x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ int arr = {Qt}new String(){{}}{Qt}; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ Int32 arr = new String[]{{}}; }}") x = dict z = x.ParseZml() Assert.AreEqual(z, $"@{{ int Name = dict[{Qt}Adam{Qt}]; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ long Name = dict[{Qt}Adam{Qt}]; }}") x = @dict z = x.ParseZml() Assert.AreEqual(z, $"@{{ List Name = dict[Adam]; }}") End Sub Sub TestInjects() Dim x = Dim y = x.ParseZml.ToString() Dim z = "@inject T A @inject string B" Assert.AreEqual(y, z) End Sub Sub TestSetters() Dim x = Dim y = x.ParseZml().ToXml() Dim s = y.Value.Trim() Assert.IsTrue(s.StartsWith("@{") And s.EndsWith("}")) Dim lines = s.Trim("@", "{", "}", vbCr, vbLf).Replace(vbCrLf, vbLf).Split(vbLf, StringSplitOptions.RemoveEmptyEntries) Assert.AreEqual(lines(0), $"d = {Qt}4/1/2019{Qt};") Assert.AreEqual(lines(1), $"d2 = DateTime.Parse({Qt}4/1/2019{Qt});") Assert.AreEqual(lines(2), "n = 3;") Assert.AreEqual(lines(3), $"s = {Qt}3{Qt};") Assert.AreEqual(lines(4), "y = arr[3];") Assert.AreEqual(lines(5), $"z = dict[{Qt}key{Qt}];") Assert.AreEqual(lines(6), "myChar = 'a';") Assert.AreEqual(lines(7), $"name = {Qt}student{Qt};") Assert.AreEqual(lines(8), "obj = Student;") x = Dim z = x.ParseZml() Assert.AreEqual(z, $"@{{ arr = {Qt}new String(){{}}{Qt}; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ arr = new String[]{{}}; }}") x = Adam z = x.ParseZml() Assert.AreEqual(z, $"@{{ dict[{Qt}Name{Qt}] = {Qt}Adam{Qt}; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ dict[{Qt}Name{Qt}] = {Qt}Adam{Qt}; }}") x = @Adam z = x.ParseZml() Assert.AreEqual(z, $"@{{ dict[Name] = Adam; }}") x = z = x.ParseZml() Assert.AreEqual(z, $"@{{ Sum = (int a, int b) => a + b; }}") End Sub Sub TestGetters() Dim x = X Dim y = x.ParseZml().ToXml() Dim z = y.Value.Trim() Assert.AreEqual(z, "@X") x = y = x.ParseZml().ToXml() z = y.Value.Trim() Assert.AreEqual(z, "@X") x = y = x.ParseZml().ToXml() z = y.Value.Trim() Assert.AreEqual(z, "@X[i]") x = y = x.ParseZml().ToXml() z = y.Value.Trim() Assert.AreEqual(z, $"@X[{Qt}name{Qt}]") End Sub Sub TestPage() Dim x = Dim y = x.ParseZml().ToXml() Dim z = y.Value.Trim() Assert.AreEqual(z, "@page") x = Pages/Home y = x.ParseZml().ToXml() z = y.Value.Trim() Assert.AreEqual(z, $"@page {Qt}Pages/Home{Qt}") x = y = x.ParseZml().ToXml() z = y.Value.Trim() Assert.AreEqual(z, $"@page {Qt}Pages/Home{Qt}") End Sub Sub TestModel() Dim x = IndexModel Dim y = x.ParseZml() Assert.IsTrue(y.Contains("@model IndexModel")) x = y = x.ParseZml() Assert.IsTrue(y.Contains("@model IndexModel")) End Sub Sub TestForLoop() Dim lp =

@i

Dim y = lp.ParseZml() Dim z = "@for (var i = 0; i < 10 + 1; i++) {

@i

}" lp =

@i

y = lp.ParseZml() z = "@for (var i = 0; i < Model.Count; i++) {

@i

}" Assert.AreEqual(y, z) lp = let="i++">

@i

y = lp.ParseZml() z = "@for (var i = 0; i < Model.Count; i++) {

@i

}" Assert.AreEqual(y, z) lp = Model.Count" %> let="i -= 2">

@i

y = lp.ParseZml() z = "@for (var i = 0; i > Model.Count; i -= 2) {

@i

}" Assert.AreEqual(y, z) End Sub Sub TestForLoopSteps() Dim lp =

@i

Dim y = lp.ParseZml() Dim z = "@for (var i = 0; i <10 + 1; i += 2) {

@i

}" lp =

@i

y = lp.ParseZml() z = "@for (int i = Model.Count - 1; i > -1; i--) {

@i

}" Assert.AreEqual(y, z) lp =

@i

y = lp.ParseZml() z = "@for (byte i = Model.Count - 1; i > -1; i -= 2) {

@i

}" Assert.AreEqual(y, z) End Sub Sub TestForEachLoop() Dim lp =

@i

Dim y = lp.ParseZml() Dim z = $"@foreach (var i in {Qt}abcd{Qt}) {{

@i

}}" Assert.AreEqual(y, z) lp =

@i

y = lp.ParseZml() z = $"@foreach (int i in {Qt}abcd{Qt}) {{

@i

}}" Assert.AreEqual(y, z) lp =

@i

y = lp.ParseZml() z = $"@foreach (var i in {Qt}abcd{Qt}) {{

@i

}}" Assert.AreEqual(y, z) End Sub Sub TestNestedForEachLoops() Dim lp =

Country: @country

City: @city

Dim y = lp.ParseZml() Dim z = "@foreach (var country in Model.Countries) {

Country: @country

@foreach (var city in country.Cities) {

City: @city

} }" Assert.AreEqual(y, z) End Sub Sub TestIf() Dim x = 3 and y<5" %>>

a = 4

Dim y = x.ParseZml().ToString() Dim z = "@if (a>3 & y<5) {

a = 4

}" Assert.AreEqual(y, z) End Sub Sub TestIfElse() Dim x = 3 andalso b == 5" %>>

test 1

test 2

Dim y = x.ParseZml().ToString() Dim z = "@if (a != 3 && b == 5) {

test 1

} else {

test 2

}" Assert.AreEqual(y, z) End Sub Sub TestElseIfs() Dim x = >

Very weak

>

Weak 2

>

Accepted

>

Good

>

Very Good

Excellent

Dim y = x.ParseZml().ToString() Dim z = "@if (grade < 30) {

Very weak

} else if (grade < 50) {

Weak 2

} else if (grade < 65) {

Accepted

} else if (grade < 75) {

Good

} else if (grade < 85) {

Very Good

} else {

Excellent

}" Assert.AreEqual(y, z) End Sub Sub TestNestedIfs() Dim x =

Test

Not Test

Show Items

item.Name

item.Name

Done

Dim y = x.ParseZml().ToString() Dim z = "@if (Model.Count == 0) { @if (Model.Test) {

Test

} else {

Not Test

} } else {

Show Items

@foreach (var item in Model) { @if (item.Id % 2 == 0) {

item.Name

} else {

item.Name

} }

Done

}" Assert.AreEqual(y, z) End Sub Sub TestAttr() Dim x = Dim y = x.ParseZml.ToString() Dim z = $" " Assert.AreEqual(y, z) End Sub Sub TestLayout() Dim x = _Layout Dim y = x.ParseZml.ToString() Dim z = "@{ Layout = '_Layout'; }".Replace(SnglQt, Qt) Assert.AreEqual(y, z) x = y = x.ParseZml.ToString() Assert.AreEqual(y, z) End Sub Sub TestImports() Dim x = Microsoft.eShopWeb.Web Dim y = x.ParseZml().ToString() Dim z = "@using Microsoft.eShopWeb.Web" Assert.AreEqual(y, z) x = y = x.ParseZml().ToString() z = "@using Microsoft.eShopWeb.Web.ViewModels.Manage" Assert.AreEqual(y, z) x = y = x.ParseZml().ToString() z = "@using Microsoft.eShopWeb.Web.Pages @using Microsoft.AspNetCore.Identity @using Microsoft.eShopWeb.Infrastructure.Identity" Assert.AreEqual(y, z) End Sub Sub TestNamespace() Dim x = Microsoft.eShopWeb.Web Dim y = x.ParseZml().ToString() Dim z = $"@namespace Microsoft.eShopWeb.Web" Assert.AreEqual(y, z) x = y = x.ParseZml().ToString() z = $"@namespace Microsoft.eShopWeb.Web.ViewModels.Manage" Assert.AreEqual(y, z) x = y = x.ParseZml().ToString() z = "@namespace Microsoft.eShopWeb.Web.Pages @namespace Microsoft.AspNetCore.Identity @namespace Microsoft.eShopWeb.Infrastructure.Identity" Assert.AreEqual(y, z) End Sub Sub TestAllImports() Dim x = $"Microsoft.eShopWeb.Web Microsoft.eShopWeb.Web.ViewModels.Account Microsoft.eShopWeb.Web.Pages " Dim y = x.ToXml.ParseZml().ToString() Dim z = "@using Microsoft.eShopWeb.Web @using Microsoft.eShopWeb.Web.ViewModels @using Microsoft.eShopWeb.Web.ViewModels.Account @using Microsoft.eShopWeb.Web.ViewModels.Manage @using Microsoft.eShopWeb.Web.Pages @using Microsoft.AspNetCore.Identity @using Microsoft.eShopWeb.Infrastructure.Identity @namespace Microsoft.eShopWeb.Web.Pages @addTagHelper *, Microsoft.AspNetCore.Mvc.TagHelpers" Assert.AreEqual(y, z) End Sub Sub TestHelperImports() Dim x = Dim y = x.ParseZml.ToString() Dim z = "@addTagHelper *, Microsoft.AspNetCore.Mvc.TagHelpers" Assert.AreEqual(y, z) x = y = x.ParseZml.ToString() Assert.AreEqual(y, z) x = Microsoft.AspNetCore.Mvc.TagHelpers y = x.ParseZml.ToString() Assert.AreEqual(y, z) x = Microsoft.AspNetCore.Mvc.TagHelpers y = x.ParseZml.ToString() Assert.AreEqual(y, z) x = y = x.ParseZml.ToString() z = "@addTagHelper *, Microsoft.AspNetCore.Mvc.TagHelpers @addTagHelper *, MyHelpers" Assert.AreEqual(y, z) End Sub Sub TestInvokes() Dim x = 3 'a' Ali Dim y = x.ParseZml.ToString() Dim z = $"@Foo(3, 'a', {Qt}Ali{Qt}, m => m.Name, (int n) => n + 1, (int x, int y) => x + y, (double a, float b) => a + b)" Assert.AreEqual(y, z) x = Scripts false y = x.ParseZml.ToString() z = "@RenderSection('Scripts', required: false)".Replace(SnglQt, Qt) Assert.AreEqual(y, z) x = 3 'a' Ali y = x.ParseZml.ToString() z = "@{ await " & $"Foo(3, 'a', {Qt}Ali{Qt}, m => m.Name, (int n) => n + 1, (int x, int y) => x + y, (double a, float b) => a + b);" & " }" Assert.AreEqual(y, z) End Sub Sub TestNestedInvokes() Dim x = 3 Scripts false Ali @x @y false Ali Dim y = x.ParseZml.ToString() Dim z = "@Foo( 3, RenderSection('Scripts', required: false), 'Ali', m => m.Name, (int x, int y) => Test(x, y, (double a, float b) => a + b), await Foo2(false, 'Ali'))".Replace((SnglQt, Qt), (vbCrLf, "")) Assert.AreEqual(y, z) End Sub Sub TestSections() Dim x = Dim y = x.ParseZml.ToString() Dim z = "@section Scripts { }".Replace(SnglQt, Qt) Assert.AreEqual(y, z) End Sub Sub TestDots() Dim x = Dim y = x.ParseZml.ToString() Dim z = "@{ hasExternalLogins = (await SignInManager.GetExternalAuthenticationSchemesAsync()).Any(); }" Assert.AreEqual(y, z) End Sub End Class End Namespace

[VBAndCs]

Note: The problem may happened after I added this:

        <TestMethod>
        Sub TestInjects()
            Dim x = <zml xmlns:z="zml">
                        <z:inject A.type="T(Of integer)"
                            B.type="string"
                        />
                    </zml>

            Dim y = x.ParseZml.ToString()
            Dim z =
"@inject T<int> A
@inject string B"

            Assert.AreEqual(y, z)

        End Sub

using the literal 'integer' may be a common case with the previous issue!

[CyrusNajmabadi]

@VBAndCs please put your enormous codeblocks either behind a <details> tag, or put them in a gist somewhere or on github. You've made it nearly impossible to use this github page now.

[CyrusNajmabadi]

I added a new test method2 in the class, and ran the tests, but the test method 2 failed. When I checked it, I found that this part of the string literal has changed

I am unable to repro anything here. What are you explicit repro steps for how you ran into a problem?

[VBAndCs]

I put the link put it is a few hours older than the file of the issue.. Try to write some new method in the middle of this file, using the same shape of other test methods (define xml literal and some string literals that represents the result.. Use some vb keywords as values of attrs in xml literals. This is all what I did but I can't tell what exactly triggered this action. If I know it wouldn't be so bad, because I would avoid the reason. Now I divided this large file onto 6 smaller files to minimize the damage, and add new testes at the end of the file or in a new one. But the most important to me now is having a VS option to disable the implicit multi line string literals in vb.net, so the editor expects the pair of quotes in the same line. It is easier to me to paste & vbCrLf & at the end of each line than reviewing all these damaged literals every time the issue occurs. This option is a good compromise for vb.net, Roslyn and developers who work in such parsing projects. This may not be a regular issue in other types of apps. This option should be per solution or per project, not for VS at general (it can not be a code fix, because the problem will still exist as long as the editor searching for the matching quote all over the file). And it will be a relief when this option to auto convert the multi line literals like this:

dim x = "line 1
line 2"

to the old school one line concatenated lateral:

dim x = "line 1" & vbCrLf &
              "line 2"

Thanks.

[CyrusNajmabadi]

I put the link put it is a few hours older than the file of the issue.. Try to write some new method in the middle of this file, using the same shape of other test methods (define xml literal and some string literals that represents the result..

I can't repro this.

[CyrusNajmabadi]

But the most important to me now is having a VS option to disable the implicit multi line string literals in vb.net, so the editor expects the pair of quotes in the same line.

Roslyn has no options to disable language features from scanning/parsing. Even if you set the language version to a point prior to when these were supported, the scanner/parser will stil understand them, you will just get an error that the feature isn't allowed.

[CyrusNajmabadi]

This option is a good compromise for vb.net, Roslyn and developers who work in such parsing projects.

Roslyn hasn't experienced this issue ourselves. And i don't believe we've heard this from other devs working in parsing projects. Perhaps it's something about your particular setup. However, without a repro, there is no way to tell.

[CyrusNajmabadi]

And it will be a relief when this option to auto convert the multi line literals like this: ... to the old school one line concatenated lateral:

Maybe i'm misunderstanding, but that seems much worse. It's much clearer to have:

dim x = "line 1
line 2"

instead of

dim x = "line 1" & vbCrLf &
              "line 2"

That's why VB added multi-line strings in the first place.

[VBAndCs]

Roslyn hasn't experienced this issue ourselves. And i don't believe we've heard this from other devs working in parsing projects.

Maybe there are only few folks who are writing such a parser in VB.net code! I write code in this app daily, and I ran into the issue twice in a few days. I can't ask you to do so, but you should try to imagine what can cause this in worst case scenarios.

That's why VB added multi-line strings in the first place.

A worst appearance is better than wasting time in fixing these random changes once and twice and maybe more. I am searching for a remedy here not a luxury. Moving all strings to a resource file will consume time, and is not suitable for testing purpose.

[paul1956]

Maybe a simple refactoring to convert multiline literals to some other format (single line, XML) and back is the solution. I would find that useful I have run into different people using different methods and having 1 method in a project would be clean.

[CyrusNajmabadi]

Maybe there are only few folks who are writing such a parser in VB.net code!

I've certainly written lots of code like that. I haven't run into what you're running into. Can you provide repro steps to show where the problem is?

A worst appearance is better than wasting time in fixing these random changes once and twice and maybe more. I am searching for a remedy here not a luxury. Moving all strings to a resource file will consume time, and is not suitable for testing purpose.

1. It would take just a few minutes to write a tool to do this using roslyn.
2. i have no idea why this wouldn't work for testing purposes.

[CyrusNajmabadi]

Maybe a simple refactoring to convert multiline literals to some other format (single line, XML) and back is the solution.

@paul1956 Sure. If you or @VBAndCs want to provide such a refactoring, that might be nice to have. I'm happy to review it if you do this.

[VBAndCs]

i have no idea why this wouldn't work for testing purposes.

In testing I want to be able to see the text in place, and quickly change it as the code changes. Maybe after I finish developing the project, I can move all literals away.

[CyrusNajmabadi]

In testing I want to be able to see the text in place, and quickly change it as the code changes. Maybe after I finish developing the project, I can move all literals away.

You can easily do that in a resource file. They're just xml. So have the file open, and edit in place. I do this all the time with resources in the roslyn codebase.

[CyrusNajmabadi]

At this point, without a repro, i'm not sure what can be done with this issue. @VBAndCs can you close it out? If you want to contribute any features here i'm happy to take a look. However, i haven't been able to see any problems in the editor in this space, and you haven't been able to show how you ran into it. Given that you can't even force hte issue to happen to you, it doesn't sound like something that actually commonly happens. Perhaps you accidentally did this yourself without realizing it?

[VBAndCs]

@paul1956 Changing string format will not solve the problem as long as the editor is looking for the matching qoute across all the file, so that writing any starting quote in any part of the file can mess up all next code lines, regardless of how string literals are formated.

[VBAndCs]

@CyrusNajmabadi This issue is about a repeatedly random error, so, I can't tell you how to reproduce it. You have the code that have the issue, and you know how the editor works, so you can try to figure out how this issue happens. I don't think it is a good thing to close this issue. I expect I will see more of this soon, and I may notice the trigger. Also, other can have this issue, while moving into the newly porn VS.NET 2019. Closing is not the best action to do right now. Thanks.

[CyrusNajmabadi]

o that writing any starting quote in any part of the file can mess up all next code lines, regardless of how string literals are formated.

@VBAndCs I cannot repro this. what is your repro that shows that you can "mess up all next lines" by doing this? For all i know, this is due to some plugin or something else you're doing.

[CyrusNajmabadi]

This issue is about a repeatedly random error, so, I can't tell you how to reproduce it.

So, how do you know it's roslyn that's doing this?

You have the code that have the issue, and you know how the editor works, so you can try to figure out how this issue happens.

I have tried. I can't figure anything out that would cause this. Everything i try locally (including deleting random quotes, or adding random quotes, or change code and editing around things), doens't repro any problems.

Also, other can have this issue, while moving into the newly porn VS.NET 2019

Nothing changed here. VB added this back in VS2015.

I don't think it is a good thing to close this issue.

1. you haven't identified that roslyn is actually the problem here.
2. you can't reproduce this issue yourself.
3. you haven't provided a way for others to repro this.

What do you think can be done with the issue in this state?

[VBAndCs]

@CyrusNajmabadi The last issue is not about quotes only, as there is an additional literal: the xml literal. Adding or removing spaces around <' and '\> in string literals indicates the editor tries to format it as xml literals! My code has two kinds of literals, and each of them can be embedded in the other for the sake of parsing, code generation, and testing! Add to this they both can contain VB keywords, which I defined as xml tags and which are common with c# like if and double which appear in string literal! I don't think VB parser/Editor was tested for this unusual mix before! So, you can try many different combination of these components. I advice you try adding a new sub in the middle of the file, and make a copy on of the code of others sub by hand as if you writing a new code. the problem can happing while writing any of these chars: ", >, <, /> </ writing in both xml literals and string literal. and try to encode <if> .. </if> and "@if () {}" . This is all what I am doing. I will continue working and observing. I am not pushing you to do any thing, so don't feel obligated to provide s solution now. And sorry if I am not aware of which component can cause this. But if you inform other teams to investigate if you think it belongs to them. I also referred to this in one issue of VB.NET repo. If this happened a third time I will report if from VS. Thank.

[CyrusNajmabadi]

So, you can try many different combination of these components.

I have tried this. I cannot repro what you are saying.

I advice you try adding a new sub in the middle of the file, and make a copy on of the code of others sub by hand as if you writing a new code. the problem can happing while writing any of these chars: ", >, <, /> </ writing in both xml literals and string literal. and try to encode .. and "@if () {}" . This is all what I am doing.

I have tried this. I cannot repro what you are seeing.

I don't think VB parser/Editor was tested for this unusual mix before!

It definitely was. We've written tons of tests for both C# and VB using the VB editor. This includes lots of combnations of strings/xml-literals/etc. inside pieces of VB code. Many many thousands of tests have been written this way.

[VBAndCs]

I caught one:

1. Paste this sub:

   Sub TestElseIfs2()
       Dim x =
               <zml xmlns:z="zml">
                   <p>test</p>
               </zml>
   
       Dim y = x.ParseZml().ToString()
       Dim z =
   "@if (grade < 30)
   {
   <p>Very weak</p>
   }
   else
   {
   <p>Excellent</p>
   }"
   
       Assert.AreEqual(y, z)
   
   End Sub

2. change </zml> to <p></p> Note that the editor with show an eeror in the string literal under < 30.

3. Now, any action will reformat the string literal. Put the caret between the <p></p> and press enter. This will reformat the string literals not only in the current sub, but until the end of the file!

[CyrusNajmabadi]

Now, any action will reformat the string literal.

This is not true. I was able to replace </zml> with

and then complete the` thus ending the xml snippet.

If you don't actually end the snippet then of course VB will think the remainder is part of the XML. That's how xml works after all :)

Put the caret between the

and press enter. This will reformat the string literals not only in the current sub, but until the end of the file!

I would recommend finishing your incomplete constructs.

Note: if you don't want automatic pretty-listing of code, i already showed how you could disable that here: https://github.com/dotnet/roslyn/issues/35028#issuecomment-483909549

It seems you did not listen to that advice. If you have pretty listing enabled, then the LS will pretty list your code. If you don't want this happen automatically, then disable the feature. Then you can hit enter between <p></p> and nothing else will pretty list.

[VBAndCs]

This is a bad design for xml literals. This issue can always happen when editing an xml literal such as pasting a part of a literal then starting to complete it, or mistakenly selecting the end tag with a previous part and deleting... etc. Disabling the editor formatting is not a solution or you can say it is a zero solution for equations as in mathematics. As I discussed above about the innocent string literals, if the editor found a broken xml literal it must stop formatting it until it is fixed. The editor can have a sign that what it thinks as an xml literal is actually containing a full vb.net lines of code, at least one method declaration statement. And as I said before, the editor can even copy the last correct syntax tree as a reference so a single edit can not compromise a whole code file In fact, I encountered many issues with xml literals in last month, and reported them in vb repo. On top of that, the disability to use some & symbol folod with copy and some other famous shortcuts, the dysfunctional xsd schemas, and not preserving white spaces in xml literals that contains text nodes, which can be solved by passing an optional param to the XElement.Parse method when vb/Roslyn parses the literal. Xml literals need more inhancements for sure, and I suggested more of them: https://github.com/dotnet/vblang/issues/397

[CyrusNajmabadi]

Disabling the editor formatting is not a solution or you can say it is a zero solution for equations as in mathematics.

Why is it not a solution?

And as I said before, the editor can even copy the last correct syntax tree as a reference so a single edit can not compromise a whole code file

You can def contribute a fix here if you want :-)