[lambda2] Andrzej's review

From: Andrzej Krzemienski (akrzemi1_at_[hidden])
Date: 2021-03-31 11:00:13

Hi Everyone,
I recommend to REJECT Boost.Lambda2 library in its current form. My opinion
is not very strong, and it is based on the perceived trade-off between
usefulness and technical problems.

I acknowledge that there are practical use cases that the library can
address, that are not provided in documentation. (I recommend that examples
with pointers, and maybe a complex one with column-oriented design be
added).

I consider the ADL issue that Barry pointed out a serious one. This will be
an additional place where users would suddenly be caught by a magical
problem that they will spend hours understanding. I think that the main
problem here is that Boost.Lambda2 defines the operators that should belong
to namespace std::placeholders. I also note that if this was added to the
Standard Library, the problem would be nonexistent, because they would
reside in namespace std.

Could I propose an alternative design. Add placeholders _1, _2, ... into
namespace `boost::lambda2::placeholders` specialize std::is_placeholder`
for them, so that `std::bind` can recognize them. Add operators in the same
namespace. Then, recommend the following usage of the library:

```
using namespace boost::lambda2::placeholders;

void fun()
{
  auto pred = std::bind(_1, &Employee::name) == std::bind(_2,
&Employee::name);
}
```

This single using directive would make both `std::bind` and Boost.Lambda2
operators work. I would find such a library acceptable. And in fact I could
make it a condition for a conditional acceptance.

Now, the standard questions:

>    - What is your evaluation of the design?
>
I object to the choice to add operators for placeholders from one namespace
(std::placeholders) into another namespace (boost::lambda2). Save for that,
I find anything else good.

   - What is your evaluation of the implementation?
>
Very sweet and clean. I only expect other operators to also work. But Peter
already explained that this is his plan.

   - What is your evaluation of the documentation?
>

This library doesn't require much documentation of the interface. But I
fill it is missing the Motivation section. I would expect to see practical
examples: not long ones, but ones that I will find in real programs, like
dealing with pointers.

   - What is your evaluation of the potential usefulness of the library?
>

Others have convinced me that it will find sufficient application in real
programs.

   - Did you try to use the library? With which compiler(s)? Did you
>      have any problems?
>

I tried to reimplement it, and played with some toy examples in Wandbox.
(newest GCC). It worked fine.

   - How much effort did you put into your evaluation? A glance? A quick
>      reading? In-depth study?
>

I was able to reimplement it, read the entire implementation (which was not
difficult: it is like seventy lines). I spent like three hours in total
studying it.

   - Are you knowledgeable about the problem domain?
>

I used Boost.Lambda and Boost.Bind in the past.

Finally, I would like to thank Peter for writing and sharing this library.

Regards,
&rzej;