Hacker News Reader

Comment by kazinator | Hacker News Reader

> Now onto the new instruction:

Adding a special VM instruction for such operations as inheriting from a class is a strikingly bad design decision.

You want to minimize the proliferation of VM instructions as much as possible.

A rule of thumb is: is it unreasonable to compile into a function call? Or else, is it likely to be heavily used in inner loops, requiring the fastest possible dispatch? If no, don't add an instruction for it. Compile it into an ordinary call of a run-time support function.

You're not going to inherit from a class hundreds of millions of times in some hot loop where the application spends 80% of its time; and if someone contrives such a thing, they don't necessarily deserve the language level support for cutting their run-time down.

View on HN

This is a good point. One of the things I have to balance with the book is teaching the optimal way to do something without dragging through reader through a large amount of verbose, grungy code. So sometimes (but not too often) I take a simpler approach even if it's not the best one.

With this VM, we have plenty of opcode space left, so it's natural to just add another instruction for inheritance, even if it does mean that the bytecode dispatch loop doesn't fit in the instruction cache quite as well.

I'll think about this some more. I'm very hesitant to make sweeping changes to the chapter (I want nothing more in life than for the book to be done), but this would be a good place to teach readers this fundamental technique of compiling language constructs to runtime function calls.

Probably sufficient to add it as a chapter footnote/challenge like you did with flexible array members for the string implementation. Maybe with a pointer to an example if you have time to find one.

This. Compromises are fine: real-world solutions make them all the time. What’s valuable to readers is displaying the reasoning behind them.

Hi Bob! I just want to tell you that I've just started reading your book and I'm enjoying it very much.

Maybe you could add a design note in the chapter with the mentioned content?

Keep it up!

> Maybe you could add a design note in the chapter with the mentioned content?

Great idea!

Made a note of it: https://github.com/munificent/craftinginterpreters/issues/62...

> technique of compiling language constructs to runtime function calls.

That can be as basic as generating a function call AST node directly from a construct, or doing a node-to-node transformation.

The compiler doesn't see anything but a function call node (though any source code tracking info and whatnot still references the original construct).

If you're doing things with the AST other than just generating code, you may want to have a node for that original construct. (For instance, feeding cross-referencing info to a language server or whatever.)

The implementation is actually a single-pass compiler so it goes from token stream straight to bytecode generation with no intermediate AST or other IR in between. Even so, it would be straightforward to emit the bytecode instruction sequence for a call to a special runtime function.

Well, the thing about this approach, I think, is that your code itself essentially becomes something like an image of your syntax tree (a la recursive descent parsing). That's fine if you want a sort of "toy" interpreter or interpreter that doesn't care about speed much.

But you're also using bytecode, which involves more caring about speed, so you've got a bit of a mismatch - as your code tries to match the multiple layers you're skipping, your code will get more complicated or your VM gets more arbitrary, so of the situation you have here.

> or interpreter that doesn't care about speed much.

This is the same approach taken by Lua and many of the original Pascal compilers, which had to run on very slow hardware, so there's pretty good precedent that you can get adequate performance.

> But you're also using bytecode, which involves more caring about speed, so you've got a bit of a mismatch

I wouldn't describe it as a mismatch as much as it is a trade-off. Because the compiler only has a peephole view into the source when it needs to generate code, many optimizations are off the table. However, because we have full control over the instruction set, we can sometimes tweak the bytecode format in order to more naturally align with the compiler.

In return for not needing an intermediate representation or AST, we get a much simpler compiler, especially in C where you have to worry about memory management.

It is reasonable to say this is a tradeoff. However, one of the costs is having the VM structure depend on the language structure, which kind of a software engineering no-no. But yes, tradeoffs always happen.

> which kind of a software engineering no-no.

Optimization along any axis (code size, simplicity, compile-time performance, runtime performance, memory size, etc.) always involves some level of violating software engineering norms.

One way to think of software engineering is that it's the practice of optimizing long-term developer velocity. Any practice that optimizes other factors generally does so at the expense of that. That's OK. Meta-software engineering is about knowing when long-term maintainability is the right factor to optimize for. :)

That is really only a political no-no. The issue is that if your VM takes off in popularity, some people get upset that it's a poor fit for their favorite language.

Technically, it's perfectly fine for a VM to be designed so that the semantic gap is small between it and the intended language family.

Languages evolve, and changes need to be made.

Many languages didn't support "arrow functions" AKA lambdas until fairly recently. Seeing how one would go about making large changes to the language is a good lesson to learn. One could argue that we should have known superclasses were going to be supported from the beginning, and I'd agree, hindsight is always 20:20.

Off-point. Lambdas are different, in the sense that with such functions you 1) need to think about the closure of the function 2) might need to reconsider what part of memory will be executable (you certainly don't want an executable heap). The change here is in the other direction - you remove a feature that is easily swapped with a builtin, which is implemented in terms of other already existing language features.

Sharp edged words. High performance code only exists when you've developed experience from building low performance code.

I think a few explanatory notes and discussion of suitable alternative approaches is warranted but not a complete rewrite because its somehow wrong.

Each VM has its own quirks which produce their own opportunities. Why would this one be any different? Especially as a teaching instrument where it has to be simpler / clearer than a production codebase?

Completely agree. Instruction sets are very inflexible. If a language has a new keyword or instruction, developers will spend a lot of time avoiding it at all costs and hiding it behind methods.

The new keyword requires a concrete class to instantiate. Abstractions such as interfaces cannot be used with new. The resulting instruction will directly reference that concrete class which makes it part of the software's binary interface. Its behavior can't be overridden: a constructor will always return an instance of the constructed class, never a subtype.

The factory design pattern was invented to solve these problems. Replacing new with normal method calls pretty much takes care of everything. The author himself has written about this on his blog:

http://journal.stuffwithstuff.com/2010/09/18/futureproofing-...

http://journal.stuffwithstuff.com/2010/12/14/the-trouble-wit...

It depends on the language, but in the Smalltalk vm I'm writing, super class methods get called a lot it seems. More than enough to have their own bytecode.

> You want to minimize the proliferation of VM instructions as much as possible.

I guess that seems like half the trade off to me, or else VMs would all be OISCs. What you really want to do is to approximate Huffman encoding in your ISA, balancing that with ease of parsing (so bytes generally rather than bits).

But this OP_INHERIT instruction isn't for calling methods; it's just for setting up the inheritance relationship.

The opcode OP_INHERIT takes two classes as parameter on the stack so this is equivalent to do a function/method calls with the class and the super class and let the function do the magic of filling the vtable, etc.

> Adding a special VM instruction for such operations as inheriting from a class is a strikingly bad design decision

I am struggling how this can be seen as a runtime instruction anyway - surely inheritance is a compile time semantic?

For statically-typed languages, yes, but many dynamically-typed languages allow class definition and inheritance at run-time. Read the classes and instances chapter to see that this interpreter indeed does.

Reply on news.ycombinator.com