Boost logo

Boost :

Subject: Re: [boost] [outcome] success-or-failure objects
From: Gavin Lambert (gavinl_at_[hidden])
Date: 2018-01-23 22:08:30


On 24/01/2018 08:00, Niall Douglas wrote:
> What I'm trying to do here is explain how these ValueOrError Concept
> matching objects are philosophically different in design to the proposed
> WG21 objects. We model success-vs-failure. They model value-vs-error.
> That has implications throughout the whole design of Outcome, which the
> tutorial hopefully covers.

I think Vinícius' point is that for people unfamiliar with the terms
used in WG21 papers, this is a meaningless statement as both terms seem
equivalent in general English.

Using ValueOrError rather than value-or-error would be some improvement
as it hints that you're talking about a specific concept; adding a link
to the WG21 paper would be better.

However like Vicinius I'm not sure that this is a point that even needs
to be made.

Additionally, I'm not even sure that it's accurate. We've just had a
big [system] discussion expressly about the fact that the presence of an
error_code does not necessarily imply failure -- there are some codes
that indicate warnings or otherwise qualified successes, and ultimately
the decision of what constitutes success is up to the interpretation of
the caller, not the callee or the code transport mechanism.

The whole thing about "programmable actions in response to no-value" is
a good feature but it doesn't sound like it should be described as
related to success or failure in those terms.


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk