GSL - Microsoft's Guideline Support Library

[George P]

Has anyone tried the GSL?

I was reading this:
https://www.codingame.com/playgrounds/7920/how-not_null-can-improve-your-code

and was intrigued by the not_null<> construct.

[Niccolo]

Anyone who walked through Stroustrup's "C++ Principles and Practice" became familiar with "the original" from Stroustrup and friends at isoccp.com

[Cubbi]

We use it at work, here and there.

[George P]

So one "no," and one "yes, conditional."

*waiting for others' opinions that likely won't be expressed*

isocpp.com is now http://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines

That ^^^^ is an document I have saved the link to, and occasionally refer to when I want to fry some brain cells.

From the VERY brief glance I made at the GSL github page it appears to be an attempt by MS to take the written guidelines and turn them into a library.

What may be easily apparent to a C++ professional isn't to this self-learner. M'ok?

I am trying to self-teach the new C++11/14/17 core features, a 3rd party library is "on the list."

[George P]

Ooooh, reading this on the GSL ( https://visualstudiomagazine.com/articles/2016/06/01/using-the-not_null-template.aspx ) is encouraging. Header only for the not_null<> template construct as well as the other .

*downloading the GSL*

[Niccolo]

So one "no," and one "yes, conditional."

I didn't mean to convey a "no" on that question. I'm rather Microsoft averse, myself. The link to the Microsoft download is on the isocpp guidelines introduction, but my gravitational pull is back toward the guideline documentation as a point of focus.

...and I was wrong about my citation, it comes from Stroustrup's "Tour of C++", 2nd edition - and he references the GSL and it's namespace somewhat liberally in that text.

Microsoft's "ownership" of that GSL library is technically true, but rather "non-Microsoft" behavior relative to their past. They've "turned a new leaf", as it were, on open source software, and are participating with the C++ standards committees and on non Windows development with their tools and languages more than I would ever have expected.

I still approach Microsoft according to advise I read years ago about doing business with them. To paraphrase, if you shake hands with Microsoft, make sure you get all your fingers back.

I would argue that Microsoft is curating the GSL library you see, but I don't think they are the true origin of the work at large. It is open source. I accept it mainly because Stroustrup has his hand on it, references it, suggests it, and....it does make sense.

You may want to review the 2nd edition of "A Tour of C++" if you haven't already.

While I regard myself as a professional, having done this work since the late (late) 70's, I submit that all programmers and software engineers are autodidacts beyond college, lest they remain skilled only as frozen in the time they graduated. C++ didn't even exist in my early years, and I had incorporated my consulting firm, landed my first fortune 500 contract, fathered to girls, then approached divorce before C++ was generally regarded as even existing in the industry (late 80's).

It will happen again, repeatedly, without end in sight, that new languages and language features emerge regularly, and so we study, without end.

I submit the key differentiation of "professional" developers is commitment to the engineering practices, and, of course, the nature of being paid. I know many "professionals", or so self declared, who have no idea what professionalism is, and may well be regarded as stuck in a previous era.

However, there is an opposing problem I see in discussions everywhere you would do well to avoid. Way too many are over-focused on the latest version release of the language(s). At this point we should all have a firm grip on C++17. The compilers are compliant, and 17 was a small, incremental advance over 14, which itself wasn't a huge jump beyond 11. C++11 was the huge jump, and it took several years for all of the compilers to become fully compliant. It is, now, aged (well, too).

C++20 is a big jump. There are no compliant compilers. It makes no sense to over-focus on it until compiler compliance is possible. Of course, the people working on those compilers had better be focused as their primary objective, but that's not most of us.

All of these advancements were intended and designed for gradual, incremental incorporation into existing practice. If you're a student just now becoming competent, you probably began with C++11 and maybe C++14 in full view. Many of us come from a time decades before then. We know full well C++20 features will be incorporated in new code starting in mid to late 2020, but that's the earliest.

Even then, many will just at that point finally master C++17's minor additions, and lots of code will be in use (and relied upon) from much earlier epochs.

GSL is good stuff, and it actually follows what Stroustrup introduces in the "Principles" book - some convenience features worthy of your consumption.

Do you know boost?

A lot of what we know regard as the C++ 11 standard library components migrated from boost first.

There are way too many libraries in boost to consider all of it for your use. It is great to have around, though, so keep an updated build of it handy (a good portion of it is header only, but some must be built).

[Albatross]

I believe the isocpp site indicates that the Microsoft-provided implementation of the GSL is listed on their site primarily because it's one of the first such implementations, and that other implementations of the GSL are encouraged: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#faq51-is-githubcommicrosoftgsl-the-gsl

-Albatross

[George P]

@niccolo, point me towards another implementation of the GSL that isn't MS then.

Hard to use an idea as a header/source unit for a project.

Boost has too much a lag time for immediate implementation of temporary types.

https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#faq53-why-werent-the-gsl-types-proposed-through-boost

As I see it the GSL is an adjunct to the Core Guidelines, just as Boost is adjunct to the ISO standard. The GSL simply isn't as formalized.

[JLBorges]

> point me towards another implementation of the GSL that isn't MS then.

That implementation is distributed under an MIT licence; it is as free as it gets.

[George P]

@JLBorges, as free as it gets, and as far as I can find the only library that is available to support directly the C++ Core Guidelines.

Just because it is MS doesn't mean it is bad, or good, especially when the GSL is designed to be very lightweight and used to make the guidelines easier (less language restrictive) to implement.

https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#gsl-guidelines-support-library

As I originally asked, "Has anyone tried the GSL?" and so could share their experience(s). Good, bad or indifferent.

The library being header-only looks easy to use.

[MikeyBoy]

We use the GSL here at work, although I haven't looked too deeply at it myself - I need to get more familar with it. I've found gsl::span useful before.

I haven't used the not_null feature myself - and hadn't been aware of it until this thread - but a quick git grep shows me that it's been used in some parts of our codebase. It looks useful, so I'm going to start using it where appropriate.

One thing that article doesn't make clear enough, is that it's important to consider whether it's valid for the pointer in question to have a value of 0 at runtime. For some pointers, a value of 0 is something that should never happen, and is indicative of a bug; these cases are ideal for using not_null. For others, it is a valid condition, and should be checked for and handled explicitly, rather than using not_null. (Yeah, I know that's obvious, but it's something that could easily be missed by an overzealous convert :) )

[George P]

@MikeyBoy, thanks for the reply. :)

I do try to have a bit of thought about code design before deciding to jump in and use some new (to me) construct. not_null is not something I can throw around with abandon thinking it will solve all the world's problems. ;)