Refactor numeric_cast #1
Conversation
This allows to extend numeric_cast to other argument types than mp_integer.
This returns a value instead of an optional but an invariant can fail.
numeric_cast can now be used instead.
This is to demonstrate the use of numeric_cast
std::numeric_limits methods are not declared constexpr in old versions of VS so we need to replace them by preconditions
|
|
||
| optionalt<T> operator()(const mp_integer &mpi) const | ||
| { | ||
| constexpr const auto precondition = |
There was a problem hiding this comment.
Surely this will be a static error on VS2013 which doesn't have constexprs in the standard library
There was a problem hiding this comment.
Hmm, potentially yeah. You could put an ifdef around the constexpr I suppose, or we might have a macro that conditionally expands to constexpr in util somewhere.
| static optionalt<T> numeric_cast(const mp_integer &mpi) | ||
| template <typename U = T, | ||
| typename std::enable_if<std::is_signed<U>::value, int>::type = 0> | ||
| static auto get_val(const mp_integer &mpi) -> decltype(mpi.to_long()) |
romainbrenguier
left a comment
romainbrenguier
left a comment
There was a problem hiding this comment.
Thanks for the proposal. What I don't find nice is that we pull a lot of things from arith_tools to mp_arith which I think we want to keep seperate. Could we instead define these numeric_cast things in a new file?
|
|
||
| const mp_integer mp_zero=string2integer("0"); | ||
|
|
||
| class exprt; |
|
The problem is that I think all the numeric cast overloads need to be defined in one go. So as long as all the numeric_cast definitions are moved, that will work, but moving just a few of them won't |
romainbrenguier
force-pushed
the
refactor/expr_cast_to_util
branch
2 times, most recently
from
d69a2dd to
7fbfeec
Compare
|
I'm not sure what to do with this branch. I don't plan to work on it any more, as all the suggested changes are very minor. I'd suggest that you cherry-pick any changes you like (if any) and close it once you've extracted anything useful. |
Yes no problem, I already integrated some of the changes in my branch. Just leave this open in case there are other changes I could use. |
romainbrenguier
force-pushed
the
refactor/expr_cast_to_util
branch
6 times, most recently
from
70598c4 to
9a811d8
Compare
We should keep the struct tag type, not follow it and get the struct type
3 participants
No description provided.