Pattern matching in C++ and D
Francesco Mecca
me at francescomecca.eu
Tue Jan 7 14:00:01 UTC 2020
I found a paper regarding a proposal for pattern matching in C++
on HN a few days ago:
http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1371r1.pdf
Given the time I spent with OCaml and the OCaml compiler recently
I find that pattern matching is an essential feature for a modern
language and that there are cases in which the pattern matching
compiler outputs better code than nested handwritten if-else-if's.
In D we have the Sumtype library that is an excellent library for
sumtypes and coincidentally provides pattern matching based on
the type of the value being matched.
As outlined from the paper there are many other kinds of pattern
that could be very useful in current D code. I won't make
examples here because the paper is full of examples that are
pretty easy to mentally translate from C++ to D.
In the "Design Decision" chapter of the paper the authors discuss
about not restricting side effects. In OCaml there are the same
problems with guards that could have side effects. Example:
```
let function_with_side_effects x = ...
let match x = match x with
| p1 -> e1
| p2 when function_with_side_effects x -> e2
| _ -> e3
```
If x is modified by functions with side effects the pattern
matching could be not exhaustive and in the worst case the result
undefined. The same applies when a guard uses boolean comparison
that has been overridden by the developer.
In D we can do better than that by forcing guard expressions to
be pure.
I also believe that ranges provides a better interface for a more
expressive pattern matching but I have to think more about that
(the paper shortly discuss that in section 10.2).
The main shortcoming with D is that we don't have a builtin tuple
type and destructoring assignments.
Timon Gehr was working on that
(https://github.com/tgehr/DIPs/blob/tuple-syntax/DIPs/DIP1xxx-tg.md) but the last commit is 2 years old.
More information about the Digitalmars-d
mailing list