Example is bad
qknight opened this issue · comments
I try to understand how to add proper error handling by looking at:
status-code/example/file_io_error.cpp
Lines 172 to 190 in b70d070
But that example only handles the error case, can you extend the code so it also does something useful?
Expected behavior: the open_resource should return a file handle so I can read/write something. Instead it has a LINE and FILE, what is that used for?
You're right that that example isn't much use. It was intended for a WG21 audience only.
I'll delete it shortly, is the example in the Readme.md okay for you?
For some time I was under the impression that the error would contain the actual return value of the function but could not see how this would be implemented.
I guess this was never the plan, right? The concept is always what is shown in the Readme.md:
system_error2::system_code sc; // default constructs to empty
native_handle_type h = open_file(path, sc);
// Is the code a failure?
if(sc.failure())
{
// Do semantic comparison to test if this was a file not found failure
// This will match any system-specific error codes meaning a file not found
if(sc != system_error2::errc::no_such_file_or_directory)
{
std::cerr << "FATAL: " << sc.message().c_str() << std::endl;
std::terminate();
}
}
Where the sc is passed in as an additional value and processed before using the return value.
The WG21 example was for exposition, showing how lightweight exception throws of std::error would work.
You're right that for current C++, either you pass in a lvalue ref as a parameter, or use Experimental.Outcome to return an outcome::experimental::result<T> whose alternate E value is an error. So the Result would be T if successful, E if an error.
Thanks for your insights. I guess that using outcome::experimental::result<T> is not straight forward so we bail and use the former method you mentioned.
Thanks for your insights. I guess that using outcome::experimental::result is not straight forward so we bail and use the former method you mentioned.
Would you be able to clarify how result<T> is not straightforward? It would be useful to learn what the obstacles would be.
I'll reopen this if it's okay, to remind me to go delete that bad example file.
I tried to put together a minimal example and failed since I don't understand your outcome implementation, especially the experimental part of it. Also I failed to get the template magic together.
Here is some pseudo-code I want to illustrate the concept I was trying:
// probably include the generated header from here i suppose...
// https://github.com/ned14/outcome
#include "outcome.hpp"
outcome::experimental::result<T> moveMountain(bool input) {
if (input == true) {
return "worked";
} else {
// system_error2::system_code sc; // default constructs to empty
return system_error2::errc::no_such_file_or_directory
};
};
outcome::experimental::result<T> ret = moveMountain(true);
... check returned type and print the result ...
... probably with logic from here: https://stackoverflow.com/questions/13636540/how-to-check-for-the-type-of-a-template-parameter ...
if (ret is of type E)
std::cerr << "error happened" << ret..message().c_str() << std::endl;
else
std::cout << ret;
outcome::experimental::result<T> ret = moveMountain(false);
if (ret is of type E)
std::cerr << "error happened" << ret..message().c_str() << std::endl;
else
std::cout << ret;
Can we create a minimal example from that? Thanks for your help, btw!
It would also help if the two versions of the documentation, both links below, would actually server the same content:
- https://ned14.github.io/status-code/index.html
- https://github.com/ned14/status-code#quick-synthesis-of-a-custom-status-code-domain-for-any-arbitrary-enumeration-type
And if the github pages webpage (second link) could container a link to the source code (first link).
How about this https://godbolt.org/z/Tbr5h7:
#include "outcome-experimental.hpp"
#include <iostream>
#include <string>
namespace outcome_e = OUTCOME_V2_NAMESPACE::experimental;
template<class T> using result = outcome_e::status_result<T>;
result<std::string> moveMountain(bool input) {
if (input == true) {
return "worked";
} else {
// system_error2::system_code sc; // default constructs to empty
return outcome_e::errc::no_such_file_or_directory;
};
};
void test1()
{
if(auto r = moveMountain(true))
{
std::cout << r.value() << std::endl;;
}
else
{
// BUG: message() is supposed to print without c_str()!
std::cerr << "error happened " << r.error().message().c_str() << std::endl;
}
}NOTE TO SELF: message() isn't serialising to ostream, and I could have sworn I fixed that six months ago!
Re: those two pages getting out of sync, apologies, the website is manually regenerated using tools on my laptop which I keep forgetting to do. I'll leave this issue open to remind me.
Wow, that works! Thanks!
Can you please add this as a comment to the source code as well:
// Use https://github.com/ned14/outcome#usage-as-a-single-header-file and go to the download page, there is the file outcome-experimental.hpp also.
#include "outcome-experimental.hpp"
Ok, can you check what I just pushed in that commit there to see if it addresses all your issues?
The only remaining issue is the docs which need to be regenerated. For that I need my personal laptop, so I'll keep this issue open.
Regened the docs, so closing.