C#•4d ago

Object passed by reference unexpectedly

Why is the context of a modified when context is modified?

c#
        TokenContext context = new TokenContext(0,1,0,0,src); // start = 0
        Token a = new Token(new SortedSet<int>{0,1}, new List<Token>{}, context);
        context.start = 1;
        context.end = 2;
        Token b = new Token(new SortedSet<int>{1,2}, new List<Token>{}, context);
        Console.WriteLine(a.context.start); // 1
        Console.WriteLine(b.context.start); // 1

c#
        TokenContext context = new TokenContext(0,1,0,0,src); // start = 0
        Token a = new Token(new SortedSet<int>{0,1}, new List<Token>{}, context);
        context.start = 1;
        context.end = 2;
        Token b = new Token(new SortedSet<int>{1,2}, new List<Token>{}, context);
        Console.WriteLine(a.context.start); // 1
        Console.WriteLine(b.context.start); // 1

21 Replies

333fred•4d ago

Presumably, TokenContext is a class, yes?

Big ChungusOP•4d ago

yes

c#
public class TokenContext {
    public Source source;
    public int start;
    public int end;
    public int line;
    public int col;
    public TokenContext(int start, int end, int line, int col, Source source) {
        this.source = source;
        this.start = start;
        this.end = end;
        this.line = line;
        this.col = col;
    }
}

c#
public class TokenContext {
    public Source source;
    public int start;
    public int end;
    public int line;
    public int col;
    public TokenContext(int start, int end, int line, int col, Source source) {
        this.source = source;
        this.start = start;
        this.end = end;
        this.line = line;
        this.col = col;
    }
}

c#
public class Token {
    public SortedSet<int> types;
    public List<Token> children;
    public string text;
    public TokenContext context;

    public Token(SortedSet<int> types, List<Token> children, TokenContext ctx) {
        this.types = types;
        this.children = children;
        this.text = ctx.source.GetText().Slice(ctx.start, ctx.end - ctx.start).ToString();
        this.context = ctx;
    }
...
}

c#
public class Token {
    public SortedSet<int> types;
    public List<Token> children;
    public string text;
    public TokenContext context;

    public Token(SortedSet<int> types, List<Token> children, TokenContext ctx) {
        this.types = types;
        this.children = children;
        this.text = ctx.source.GetText().Slice(ctx.start, ctx.end - ctx.start).ToString();
        this.context = ctx;
    }
...
}

333fred•4d ago

Well, that's how classes work Classes are also known as reference types Do you know C?

Big ChungusOP•4d ago

yes, I'm used to c++

333fred•4d ago

Ok, so TokenContext context is actually TokenContext* context In C#, that's the difference between a struct and a class A class is a reference type, and always allocated on the heap A variable referring to one is actually a pointer to heap data

Big ChungusOP•4d ago

right, so I should almost never use classes then I suppose

333fred•4d ago

No, that is exactly the wrong approach C# is a GC language. Automatic memory management is the whole point

Big ChungusOP•4d ago

so then what should I do when I want a copy of an object?

333fred•4d ago

Depends on the context. In your case here, you'd probably take one of a few possible approaches: 1. Use a record for TokenContext, it probably should be one anyways. Then you wouldn't reassign individual properties like you're doing there, you'd with to do non-destructive mutation 2. Use a builder for TokenContexts. That's effectively what you're doing here, you're treating TokenContext as if it was a builder. But combining these two things into one isn't great; you've opened yourself up to bugs because TokenContext is mutable. 3. You could certainly make TokenContext a struct here. It's not the worst idea, given that there's only 5 pieces of data and they can presumably be enregistered by the JIT. In any of these options, though, I would absolutely make TokenContext immutable. After you create a Token, there's almost certainly no reason to permit modification of TokenContext

Big ChungusOP•4d ago

welp, I'm well out my depth here. I'll probably just use struct since it makes the most sense to me, even if it isn't the most performant I'm interested to know what I should be using for Token tho

333fred•4d ago

Nearly certainly class But I would say you're coding very much like this is C++, not C# We don't do public fields in C#, we use properties, so that we abstract whether things are actually fields

Big ChungusOP•4d ago

not sure what that means

333fred•4d ago

Do you know what properties are?

Big ChungusOP•4d ago

nope

333fred•4d ago

Alright. So, my suggestion is to do a few intro to C# things Clearly, you know about programming, so you don't need the hello world, I've never seen a line of code in my life But you should really learn about the basic building blocks of C#, like "what is a class", "what is a property", etc

333fred•4d ago

Take a look through some of the articles here: https://learn.microsoft.com/en-us/dotnet/csharp/tour-of-csharp/overview

Overview - A tour of C#

New to C#? Learn the basics of the language. Start with this overview.

Big ChungusOP•4d ago

yup, this is OOP hell

333fred•4d ago

Being perfectly honest: if that's your first reaction, why are you using C#? C# is a multiparadigm language, but types are at its core.

Big ChungusOP•4d ago

because I need cross platform support and while my c++ code would probably work on mostly any system, compiling it would be a pain