Show HN: Dlg – Zero-cost printf-style debugging for Go

In short: You're right - Go will still evaluate argument expressions. In hindsight, I should've made that more obvious, not calling this out initially was an expert blind spot on my part, not an intent to mislead.

If we consider this example:

  func risky() error {
    return fmt.Errorf("unexpected error")
  }

  func risky2() error{
    return fmt.Errorf("also an error")
  }

  func main() {
    fmt.Printf("something failed: %v", risky())
    dlg.Printf("something failed: %v", risky2())
  }

  0x10009f354  d503201f  NOOP                ; Dead code of dlg.Printf
  0x10009f358  b0000000  ADRP 4096(PC), R0   ; Load "also an error" string address
  0x10009f35c  9124ec00  ADD $2363, R0, R0   ; Calculate exact string address
  0x10009f360  d28001a1  MOVD $13, R1        ; Set string length=13 (R1)
  0x10009f364  aa1f03e2  MOVD ZR, R2         ; Clear R2
  0x10009f368  aa1f03e3  MOVD ZR, R3         ; Clear R3
  0x10009f36c  aa0303e4  MOVD R3, R4         ; Clear R4
  0x10009f370  97ffd890  CALL fmt.Errorf(SB) ; Call fmt.Errorf

I'll add a section to the README that explains this. Thanks for pointing it out.

It's easy to show that you're right and that code isn't dead so it can't be eliminated:

    package main
    
    import (
     "fmt"
    
     "github.com/vvvvv/dlg"
    )
    
    func risky() error {
     _, err := fmt.Printf("unexpected error\n")
     return err
    }
    
    func main() {
     dlg.Printf("something failed: %s", risky())
     risky()
    }

I've added a section to the README that goes into the subtleties of compiletime elimination and what to watch out for to ensure that calls to dlg.Printf can actually be eliminated: https://github.com/vvvvv/dlg?tab=readme-ov-file#true-zero-co...

I'd appreciate any feedback on whether these changes make the behavior more transparent and less misleading.

This is a thing I had a little "are you sure?" itch in my brain about, but you've expressed so clearly in four lines of comment, and this is me showing gratitude for the comment and the untangling of my brain.

ETA: Presumably if there was at least an option for passing a closure that would provide a way to address this? Since the closure will only uselessly compile, and not actually uselessly run?

I’ve been doing this (manually) for about 10+ years.

I rarely use branches, my local work area is about 50-100 commits on top of master.

I pick/reorder/edit with “git rebase -i”

I’ll prefix the commit summary with a word that helps me keep straight debug changes from ones that will go toward a MR.

I usually leave them as uncommitted changes, then git checkout to undo them. And using something like `git add -p` to skip Printfs if I need to commit other changes.

Thank you!

You're absolutely right, it annoys me as well.

It's a bit finicky to do so, but I'm currently in an exploratory phase of providing an even more fine grained stack tracing feature and filtering out dlg itself is definitly on my todo list.

> it doesn't appear to truly be zero cost if you log variables that can't be eliminated.

I'd say it is fair to call it zero cost, if the costs you are seeing are due to the way you are using it. If the values being logged are values you are already computing and storing, constants, or some mix of the two (concatenated via its printf function), by my understanding (caveat: I've never actually used Go myself) all the logging code should be stripped out as dead code in the link stage.

Obviously if you are performing extra computations with intermediate values to produce the log messages, your code there might produce something that is not reliably detected as eliminable dead code so there would be cost there.

> The only way (I believe) to implement zero cost is some type of macro system which go does not support.

That looks to be effectively what it is doing, just at the link stage instead of in a preprocessor. Where C & friends would drop the code inside #ifdef when the relevant value is not set (so it won't be compiled at all) this should throw it out later in the process if DLG_STACKTRACE isn't set at compile time.

So there will always be a compile-time cost (unlike with a preprocessor macro, something will be produced then thrown away and the analysis of what to throw away will take longer with more code present), but not a run-time cost if the logging is not enabled, assuming you don't have any logic in the trace lines beside the calls to this module.

Yeah I want a log package where the log statements that don't log also don't have their arguments evaluated. Half of my gc pressure is little strings created to call the log package that don't get logged.

See the standard library "slog" package: https://pkg.go.dev/log/slog

Which is a relatively recent addition to the Go standard library, based on the collective experience with logging over the years. Compare with the original log standard library: https://pkg.go.dev/log and you can see the changes in even just the time since Go was developed in logging best practices.

Slog, once you understand the interfaces involved in stuff, has performed most every trick I could desire, up to and including things like subloggers with their own priorities for outputs that override the global settings.

I'd like to see some slog integration in this package, though I say that without a design in mind. As it stands, this is exactly what it says on the tin; debugging-only output that you fully intend to just disappear out of the production binary.

The trick it uses is a fairly straightforward use of Go's tag-based compilation features, which I do not mean as a criticism, but to point out that any Go programmer who likes the looks of this but wants to adapt it to their own task will find it easy to simply take the technique directly and adapt it in not very many lines of code.

I do write tests but I'm not a fan of aiming for 100% coverage because it's incredibly tedious to achieve, and the gains in my experience are marginal.

That being said even with 100% coverage your code may still contain bugs. If you've never had to debug code with extensive tests then hats off to you.