[container.requirements] Fix malformed Result specifications #6824
Conversation
|
Looking at [container.requirements] as a whole, there are a number of other places where this pattern appears, and those may also be worth fixing. |
|
It is clearly an issue that the semi-colon was not intended to be in code font. I think the rewording might bring more clarity too, but suggest retaining the (non-code-font) semicolons rather than switching to commas. |
|
I think those semi-colons should not have been in code font. But it would be clearer if subtle changes in font didn't affect the meaning. |
|
@jwakely so specifically, what do you recommend for this PR? I think I agree with @AlisdairM and the commas should remain semicolons, but the rest of the changes are solid. |
|
I've now applied respective changes in other portions of However, as Jonathan has said, it would be clearer if changes in font didn't affect the meaning. The change applies the following scheme everywhere: -A; B if x is constant.
+B if x is constant; A otherwise. |
|
I don't think "is constant" is a defined term. Does it mean a constant expression? Constant-initialized? What we mean is that its type is const. The existing "for constant |
Eisenwave
changed the title
|
I assume the commits history will have to be squashed or something at the end. Also, I was unable to fix the overfull hboxes here with anything but |
|
Could you please rebase, and squash the changes to prevent overful hboxes to wherever they are needed: each commit should individually pass the checks. |
[container.reqmts] Prevent semicolon font from altering semantics [container.rev.reqmts] Prevent semicolon font from altering semantics [associative.reqmts.general] Prevent semicolon font from altering semantics [unord.req.general] Prevent semicolon font from altering semantics [container.requirements] Replace 'is constant' with 'is const' [container.requirements] Replace 'is const' with 'is of type const X' [container.requirements] Fix overfull hboxes
|
@tkoeppe done. I presume that it's fine for the long commit message to contain the info about which commits it's been squashed from. Otherwise I can still clean that up. |
| \tcode{pair<const_iterator, const_iterator>} for constant \tcode{b}. | ||
| \tcode{pair<const_iterator, const_iterator>} | ||
| if \tcode{b} is of type \tcode{const X}; \\ | ||
| \tcode{pair<iterator, iterator>}~otherwise. |
There was a problem hiding this comment.
Please drop the ~ line-break prevention.
Instead of \\, can you add a hyphenation hint \- in a suitable position?
The Result specifications for
a.front(),a.back(),a[n], anda.at(n)are formatted as:In other words, the result of
a.front()is the C++ typereference; const_referencein the event thatais constant. Otherwise, the standard imposes no requirements on the result type of the expression.Seeing that
reference; const_referencecannot be a type and that it would be surprising if no requirements were imposed on non-consta, this is presumably an editorial issue and should be rephrased.