si_trace_print

A rust library to print messages indented to stack depth optionally preceded by the function name.

Useful for trace printing function flows.

• Use
  • Trace-printing example
  • Manually setting the indentation
• Shortcomings
  • Slow
  • Release builds

Use

Add si_trace_print entry to the project Cargo.toml section [dependencies].

Trace-printing example

The most common use will likely be debug-only printing to stderr with a precding function name.

use si_trace_print::{
    den, deo, dex, defn, defx,
};

fn main() {
    den!("hello from main");
    deo!("main will be doing stuff...");
    func1(3);
    deo!("main is done doing stuff...");
    dex!("goodbye from main");
}

fn func1(_var: usize) {
    defn!("({:?})", _var);
    defo!("doing even more stuff...");
    defx!("({:?})", _var);
}

This printed

$ cargo run
→hello from main
 main will be doing stuff...
    →func1: (3)
     func1: doing even more stuff...
    ←func1: (3)
 main is done doing stuff...
←goodbye from main

If built with --release then the statements are not compiled and nothing would be printed.

An example using a variety of the available eprintln macros.

extern crate si_trace_print;
use si_trace_print::{pf1n, pf2n, pfn, pn, po, px};

fn main() {
    pn!("hello from main");
    pfn!("hello again from main");
    pf1n!("hello again from main!");
    pf2n!("HELLO AGAIN FROM MAIN!!!");
    po!("main will be doing stuff...");
    mod1::mod2::func1(3);
    po!("main is done doing stuff...");
    px!("goodbye from main");
}

mod mod1 {
    pub mod mod2 {
        use si_trace_print::{
            pf1n, pf1o, pf1x, pf1ñ, pf2n, pf2o, pf2x, pf2ñ, pfn, pfo, pfx, pfñ, pñ,
        };
        pub fn func1(var: usize) {
            pf1n!("({:?})", var);
            pf1o!("func1 calling func2...");
            func2(var + 1);
            pf1x!("({:?})", var);
        }
        fn func2(var: usize) {
            pf2n!("({:?})", var);
            pf2o!("calling func3...");
            func3();
            pf2x!("({:?})", var);
        }
        fn func3() {
            pfn!();
            func4();
            pfo!("almost complete...");
            pfx!();
        }
        fn func4() {
            pñ!("func4 is a short function.");
            pfñ!("func4 is a short function.");
            pf1ñ!("func4 is a short function.");
            pf2ñ!("func4 is a short function.");
        }
    }
}

should print

→hello from main
→main: hello again from main
→main: hello again from main!
→main: HELLO AGAIN FROM MAIN!!!
 main will be doing stuff...
    →mod2::func1: (3)
     mod2::func1: func1 calling func2...
        →mod1::mod2::func2: (4)
         mod1::mod2::func2: calling func3...
            →func3:
                ↔func4 is a short function.
                ↔func4: func4 is a short function.
                ↔mod2::func4: func4 is a short function.
                ↔mod1::mod2::func4: func4 is a short function.
             func3: almost complete...
            ←func3:
        ←mod1::mod2::func2: (4)
    ←mod2::func1: (3)
 main is done doing stuff...
←goodbye from main

Manually setting the indentation

The first use of a library macro will set the "original" stack depth. This is later used to calculate indentation offsets. If the first use of this library is several functions into a program then later printing may be lose indentation.

use si_trace_print::{
    pfo, pfn, pfx, pfñ,
};

fn main() {
    func1(3);
    pfx!("goodbye from main (this is not indented!)");
}

fn func1(var: usize) {
    func2(var);
    pfñ!("({:?}) (this is not indented!)", var);
}

fn func2(var: usize) {
    // this is the first call to a si_trace_print function
    // the "original" stack offset will be set from here
    pfn!("({:?})", var);
    pfo!("stack_depth {:?}, stack_offset {:?}", stack_depth(), stack_offset());
    pfx!("({:?})", var);
}

prints poorly indented output

→func2: (3)
 func2: stack_depth 15, stack_offset 0
←func2: (3)
↔func1: (3) (this is not indented!)
←main: goodbye from main (this is not indented!)

Explictly call stack_offset_set near the beginning of the thread.

use si_trace_print::{
    pfo, pfn, pfx, pfñ,
};

fn main() {
    // the "original" stack offset will be set from here
    stack_offset_set(None);
    func1(3);
    pfx!("goodbye from main");
}

fn func1(var: usize) {
    func2(var);
    pfñ!("stack_depth {:?}, stack_offset {:?}", stack_depth(), stack_offset());
}

fn func2(var: usize) {
    pfn!("({:?})", var);
    pfo!("stack_depth {:?}, stack_offset {:?}", stack_depth(), stack_offset());
    pfx!("({:?})", var);
}

this printed

            →func2: (3)
             func2: stack_depth 15, stack_offset 2
            ←func2: (3)
        ↔func1: stack_depth 14, stack_offset 1
    ←main: goodbye from main

The indentation is improved but is too far indented. The indentation amount to pass to stack_offset_set can be somewhat unpredictable. It depends on build settings and which thread is running, among other things. In this case, experimentation revealed value -1 to be best:

// ...
fn main() {
    stack_offset_set(Some(-1));
// ...

this printed

        →func2: (3)
         func2: stack_depth 15, stack_offset 1
        ←func2: (3)
    ↔func1: stack_depth 14, stack_offset 0
←main: goodbye from main

Shortcomings

Slow

This trace function may significantly slow a program. It is recommended to use the debug version of provided macros.

Release builds

The calculation of function depth depends on stack frames counted by backtrace::trace. In --release builds or under other optimization profiles, some functions may be optimized inline. The count of stack frames may not change among function calls. This means the printed indentation will not reflect function call depth. This can be forcibly avoided by adding attribute #[inline(never)] to such functions.