rhai: Frame before call to rust a function missing
HEAVY EDIT after I understood a crucial detail.
The frame before calling to a rust function is missing. I can inject the frame in the rust function called, but there are two problems:
- Source is not set in
NativeCallContext - When I have a method of a struct it is difficult to get
NativeCallContext
(I was able to get a NativeCallContext for methods using a wrapper written in rhai.)
My wish: To have the call to my_err automatically in the source chain, so I could avoid both problems. However, I suspect I am misunderstanding something.
use rhai::{Engine, EvalAltResult, FuncRegistration, NativeCallContext, Position};
fn my_err(ctx: NativeCallContext) -> Result<(), Box<EvalAltResult>> {
// Name is set
println!("fn_name: {:?}", ctx.fn_name());
// Source is missing
println!("source: {:?}", ctx.source());
// Position set
println!("postiion: {:?}", ctx.position());
// I'd like to emulate the rhai error pattern for the rust function, so users can send me traces,
// of course here we did not call rust_function
let src = EvalAltResult::ErrorInFunctionCall(
"rust_function".into(),
file!().into(),
"a rust error happened".into(),
Position::new(line!() as u16, 0),
);
EvalAltResult::ErrorInFunctionCall(
ctx.fn_name().into(),
ctx.source().unwrap_or("").into(),
src.into(),
ctx.position(),
)
.into()
}
fn main() {
let mut engine = Engine::new();
engine.set_optimization_level(rhai::OptimizationLevel::None);
FuncRegistration::new("my_err")
.in_global_namespace()
.register_into_engine(&mut engine, my_err);
let mut ast = engine
.compile(
r#"
fn test_inner() {
my_err();
}
fn test_other() {
test_inner();
}
test_other();
"#,
)
.unwrap();
ast.set_source("inline");
let err = engine.run_ast(&ast);
println!("err: {}", err.err().unwrap());
}
Output
fn_name: "my_err"
source: None
postiion: 3:5
err: Runtime error: a rust error happened
in call to function 'rust_function' @ 'src/main.rs' (line 17, position 0)
in call to function 'my_err' (line 3, position 5)
in call to function 'test_inner' @ 'inline' (line 7, position 5)
in call to function 'test_other' @ 'inline' (line 10, position 1)
About this issue
- Original URL
- State: closed
- Created 2 years ago
- Comments: 19
The error message is clear and it adds even more helpful information than I wished for.
About the API change. I think for developers it is enough to have both
sourcefields and clear documentation what the fields contain. So the breaking change is not needed. Of course in the future when you make a breaking release:call_position,call_source,fn_name,fn_sourcewould be nice.fn_sourcejust because thefn_namefield already exists.I am very sorry for the chaos, I stop polluting that other issue. I only wanted to make that other person aware that there is a discussion already.
We have three separate issues and you have to think about all of them separately. I discovered them all at once. Only one issue is really problematic, the other issues are semantic and could be seen different.
Issue one:
sourceandpositionnot matching inNativeCallContextandErrorInFunctionCall(problematic)Traces produced by rhai or if one inspects
NativeCallContextorErrorInFunctionCalldo not match source/position: If you go tosourceat thatposition, you are at the wrong place, becausesourceis wrong,positionis correct. You need to have multiple files for this to be a problem. With a single file the problem is not visible. It has nothing to do with rust functions. You just need multiple modules/files.Issue two: The
ErrorInFunctionCallchain does not include calls to a rust functionBy default in a trace you do not get a:
in call to function 'rust_fucntion' (line ?, position ?)line by default. I solved that by inserting a
ErrorInFunctionCallin error-handling in the rust function. This is the issue that you tend to refer, but I learned that is not a big problem, when I understoodissue one. I even think, it might be perfectly reasonable that the rust function has to return a correct:ErrorInFunctionCall. This might be a documentation issue. The only problem: since currentlysourceis not correct, I cannot do “the workaround” without theinternal-feature.Issue three: The initial source is empty
Since
sourceis always one step behind, of course the initial source has to be empty. This issue would be fixed, by either fixingsourceor addingcall_source.About all issues
Depending on how
issue oneis fixed, some of the issues might go away. They are separate but influence each other. That is why I wanted to make you aware of all these issues, while onlyissue oneis problematic.