1. Proposal
C++20 added formatting of chrono types with
but left unspecified
what happens during localized formatting when the locale and literal encodings
do not match ([LWG3565]).
Consider the following example:
wherestd :: locale :: global ( std :: locale ( "Russian.1251" )); auto s = std :: format ( "День недели: {}" , std :: chrono :: Monday );
"День недели"
means "Day of week"
in Russian.
(Note that
should be replaced with
if [P2372] is adopted but
that’s non-essential.)
If the literal encoding is UTF-8 and the "Russian.1251" locale exists we have a mismatch between encodings. As far as we can see the standard doesn’t specify what happens in this case.
One possible and undesirable result (mojibake) is
where"День недели: \xcf\xed "
" \xcf\xed "
is "Пн"
("Mon"
in Russian) in CP1251 and is not valid
UTF-8.
Another possible and desirable result is
where everything is in one encoding (UTF-8)."День недели: Пн"
We propose clarifying the specification to prevent mojibake when possible by allowing implementation do transcoding or substituting the locale so that the result is in a consistent encoding.
This issue is not resolved by [LWG3547] / [P2372], the latter only
reduces the scope of the problem to format strings with the
specifier only.
The resolution proposed here is compatible with [P2372].
SG16 Unicode reviewed [LWG3547] and there was a strong support for the direction of this paper. SG16 poll results:
Require implementations to make
substitutions with
as if transcoded to UTF-8 when the literal encoding E associated with the format
string is UTF-8, for an implementation-defined set of locales.
SF | F | N | A | SA |
---|---|---|---|---|
1 | 6 | 2 | 0 | 0 |
Consensus: Consensus in favour.
Permit such substitutions when the encoding E is any Unicode encoding form.
SF | F | N | A | SA |
---|---|---|---|---|
0 | 7 | 2 | 0 | 0 |
Consensus: Consensus in favour.
Prohibit such substitutions otherwise.
SF | F | N | A | SA |
---|---|---|---|---|
1 | 3 | 3 | 1 | 1 |
Consensus: No consensus.
SA reason: Over-constrains implementations. May be sensible for implementations to perform all conversions uniformly.
2. Implementation experience
The proposal has been implemented in the open-source {fmt} library ([FMT]) which includes chrono formatting facilities and tested on a variety of platforms.
3. Wording
All wording is relative to the C++ working draft [N4892].
Update the value of the feature-testing macro
to the date of
adoption in [version.syn]:
Change in [time.format]:
Each conversion specifier conversion-spec is replaced by appropriate
characters as described in Table [tab:time.format.spec]; the formats specified
in ISO 8601:2004 shall be used where so described. Some of the conversion
specifiers depend on the locale that is passed to the formatting function if the
latter takes one, or the global locale otherwise.
If the string literal
encoding is a Unicode encoding form and the locale is among an
implementation-defined set of locales, each replacement that depends on
the locale is performed as if the replacement character sequence is transcoded
to the string literal encoding.
If the formatted object does not contain the information the conversion
specifier refers to, an exception of type
is thrown.
4. Acknowledgement
Thanks Hubert Tong for bringing up this issue during the discussion of [P2093].