Outcome 2.1 library

Niall Douglas

Distributed under the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)

Table of Contents

Introduction

Outcome is a set of tools for reporting and handling function failures in contexts where directly using C++ exception handling is unsuitable. Such contexts include:

there are programs, or parts thereof, that are compiled with exceptions disabled;
there are parts of program that have a lot of branches depending on types of failures, where if-statements are cleaner than try-catch blocks;
there is a hard requirement that the failure path of execution should not cost more than the successful path of execution;
there are situations, like in the filesystem library, where whether the failure should be handled remotely (using a C++ exception throw), or locally cannot be made inside the function and needs to be decided by the caller, and in the latter case throwing a C++ exception is not desirable for the aforementioned reasons;
there are parts of the program/framework that themselves implement exception handling and prefer to not use exceptions to propagate failure reports across threads, tasks, fibers, etc;
one needs to propagate exceptions through layers that do not implement exception throw safety;
there is an external requirement (such as a company-wide policy) that failure handling paths are explicitly indicated in the code.
where interoperation with C code, without having to resort to C++ exception wrapper shims, is important.

Outcome addresses failure handling through returning a special type from functions, which is able to store either a successfully computed value (or void), or the information about failure. Outcome also comes with a set of idioms for dealing with such types.

Particular care has been taken to ensure that Outcome has the lowest possible impact on build times, thus making it suitable for use in the global headers of really large codebases. Storage layout is guaranteed and is C-compatible for result<T, E>¹, thus making Outcome based code long term ABI-stable.

note

Ben Craig’s work on P1886 Error speed benchmarking has led to a better_optimisation branch intended to be merged end of 2020 as Outcome v2.2.0, after twelve months of testing. This branch has a number of major changes to Outcome v2:

A new trait is_move_bitcopying<T> is added, which opts types into a library-based emulation of P1029 move = bitcopies. Experimental std::error is opted in by default. If this trait is true for your T or E type, Outcome will track moved-from status for your type, and will only call your type’s destructor if it was not moved from. If your compiler’s optimiser is sufficiently able to fold code, this improves codegen quality for Experimental Outcome very considerably, approaching the same gains as P1029 types would have. Note that the empirical performance difference will likely be nil, but the codegen does look much more elegant.
If for basic_result<T, E> both T and E are trivially copyable, union-based rather than struct-based storage will be used. This significantly improves performance in synthetic benchmarks which do nothing in deep call stacks of function calls except create and return result<T, E>, and makes Outcome return competitive results to alternative error handling choices, improving comparative optics. It is not expected that the performance difference will be detectable empirically in real world code. It is expected that the build time impact of union storage won’t be noticeable, as union storage for trivially copyable types is much easier than for non-TC types.
The compile time requirement for E types to have a default constructor is removed.

Sample usage

The main workhorse in the Outcome library is result<T>: it represents either a successfully computed value of type T, or a std::error_code/boost::system::error_code² representing the reason for failure. You use it in the function’s return type:

outcome::result<string> data_from_file(string_view path) noexcept;
View this code on Github

It is possible to inspect the state manually:

if (outcome::result<string> rslt = data_from_file("config.cfg"))
  use_string(rslt.value());                   // returns string
else
  throw LibError{rslt.error(), "config.cfg"}; // returns error_code
View this code on Github

Or, if this function is called in another function that also returns result<T>, you can use a dedicated control statement:

outcome::result<int> process(const string& content) noexcept;

outcome::result<int> int_from_file(string_view path) noexcept
{
  BOOST_OUTCOME_TRY(str, data_from_file(path));
  // if control gets here data_from_file() has succeeded
  return process(str);  // decltype(str) == string
}
View this code on Github

BOOST_OUTCOME_TRY is a control statement. If the returned result<T> object contains an error information, the enclosing function is immediately returned with result<U> containing the same failure information; otherwise an automatic object of type T is available in scope.

note

This library joined the Boost C++ libraries in the 1.70 release (Spring 2019). It can be grafted into much older Boost releases if desired.

If you choose a C-compatible T and E type. ^[return]
result<T> defaults to std::error_code for Standalone Outcome, and to boost::system::error_code for Boost.Outcome. You can mandate a choice using std_result<T> or boost_result<T>. ^[return]