Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add precision modifier for seconds in chrono format #3148

Merged
merged 7 commits into from
Nov 2, 2022
Merged

Add precision modifier for seconds in chrono format #3148

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
85a4cef
Reuse precision modifier in seconds specification
Sep 28, 2022
21a828c
Apply clang-format
Oct 20, 2022
6ecde30
Make write_fractional_seconds more readable
Oct 21, 2022
e74bc99
Add alias for typename Duration::rep
Oct 26, 2022
54fa95f
Fixes
Nov 2, 2022
e30efa7
Fix build: make n unsigned properly
Nov 2, 2022
5408dd1
Merge write_fractional_seconds call
SappyJoy Nov 2, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
90 changes: 80 additions & 10 deletions include/fmt/chrono.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1056,6 +1056,44 @@ void write_fractional_seconds(OutputIt& out, Duration d) {
}
}

template <typename Char, typename OutputIt, typename Duration>
void write_fractional_seconds(OutputIt& out, Duration d, int precision) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's merge this overload with the one above to reduce code duplication.

constexpr auto num_fractional_digits =
count_fractional_digits<Duration::period::num,
Duration::period::den>::value;

using subsecond_precision = std::chrono::duration<
typename std::common_type<typename Duration::rep,
std::chrono::seconds::rep>::type,
std::ratio<1, detail::pow10(num_fractional_digits)>>;
if (precision > 0) {
*out++ = '.';
auto fractional =
detail::abs(d) - std::chrono::duration_cast<std::chrono::seconds>(d);
auto subseconds =
std::chrono::treat_as_floating_point<
typename subsecond_precision::rep>::value
? fractional.count()
: std::chrono::duration_cast<subsecond_precision>(fractional)
.count();
uint32_or_64_or_128_t<long long> n =
to_unsigned(to_nonnegative_int(subseconds, max_value<long long>()));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can subseconds be negative?

Copy link
Contributor Author

@SappyJoy SappyJoy Nov 2, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's nonnegative because fractional is nonnegative:

const auto fractional = detail::abs(d) - std::chrono::duration_cast<std::chrono::seconds>(d);

But subseconds can be signed or unsigned and to_unsigned raises error if argument is unsigned, so i'll make static_cast.

int num_digits = detail::count_digits(n);
int zeroes =
std::min(std::max(0, num_fractional_digits - num_digits), precision);
if (num_fractional_digits > num_digits) out = std::fill_n(out, zeroes, '0');
int remaining = precision - zeroes;
if (remaining < num_digits) {
n /= to_unsigned(detail::pow10(to_unsigned(num_digits - remaining)));
out = format_decimal<Char>(out, n, remaining).end;
return;
}
out = format_decimal<Char>(out, n, num_digits).end;
remaining -= num_digits;
out = std::fill_n(out, remaining, '0');
}
}

// Format subseconds which are given as a floating point type with an appropiate
// number of digits. We cannot pass the Duration here, as we explicitly need to
// pass the Rep value in the chrono_formatter.
Expand All @@ -1081,6 +1119,26 @@ void write_floating_seconds(memory_buffer& buf, Duration duration) {
num_fractional_digits);
}

template <typename Duration>
void write_floating_seconds(memory_buffer& buf, Duration duration,
int precision) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Similarly, let's merge the two write_floating_seconds overloads.

if (precision < 0) {
write_floating_seconds(buf, duration);
return;
}

FMT_ASSERT(std::is_floating_point<typename Duration::rep>::value, "");
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's add an alias for typename Duration::rep and use it here and in several places below.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

auto val = duration.count();

format_to(
std::back_inserter(buf), runtime("{:.{}f}"),
std::fmod(val *
static_cast<typename Duration::rep>(Duration::period::num) /
static_cast<typename Duration::rep>(Duration::period::den),
static_cast<typename Duration::rep>(60)),
precision);
}

template <typename OutputIt, typename Char,
typename Duration = std::chrono::seconds>
class tm_writer {
Expand Down Expand Up @@ -1518,6 +1576,9 @@ class tm_writer {
};

struct chrono_format_checker : null_chrono_spec_handler<chrono_format_checker> {
bool is_floating_point = false;
bool has_precision = false;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We only need one flag for the condition has_precision && !is_floating_point, something like has_precision_integral.


FMT_NORETURN void unsupported() { FMT_THROW(format_error("no date")); }

template <typename Char>
Expand All @@ -1530,7 +1591,11 @@ struct chrono_format_checker : null_chrono_spec_handler<chrono_format_checker> {
FMT_CONSTEXPR void on_24_hour_time() {}
FMT_CONSTEXPR void on_iso_time() {}
FMT_CONSTEXPR void on_am_pm() {}
FMT_CONSTEXPR void on_duration_value() {}
FMT_CONSTEXPR void on_duration_value() const {
if (has_precision && !is_floating_point) {
FMT_THROW(format_error("precision not allowed for this argument type"));
}
}
FMT_CONSTEXPR void on_duration_unit() {}
};

Expand Down Expand Up @@ -1831,14 +1896,20 @@ struct chrono_formatter {
if (ns == numeric_system::standard) {
if (std::is_floating_point<rep>::value) {
auto buf = memory_buffer();
write_floating_seconds(buf, std::chrono::duration<rep, Period>(val));
write_floating_seconds(buf, std::chrono::duration<rep, Period>(val),
precision);
if (negative) *out++ = '-';
if (buf.size() < 2 || buf[1] == '.') *out++ = '0';
out = std::copy(buf.begin(), buf.end(), out);
} else {
write(second(), 2);
write_fractional_seconds<char_type>(
out, std::chrono::duration<rep, Period>(val));
if (precision >= 0) {
write_fractional_seconds<char_type>(
out, std::chrono::duration<rep, Period>(val), precision);
} else {
write_fractional_seconds<char_type>(
out, std::chrono::duration<rep, Period>(val));
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Now that write_fractional_seconds overloads have been merged, the if statement is no longer needed. You can just call

write_fractional_seconds<char_type>(
              out, std::chrono::duration<rep, Period>(val), precision);

}
return;
}
Expand Down Expand Up @@ -1999,18 +2070,17 @@ struct formatter<std::chrono::duration<Rep, Period>, Char> {
if (begin == end) return {begin, begin};
begin = detail::parse_width(begin, end, handler);
if (begin == end) return {begin, begin};
auto checker = detail::chrono_format_checker();
checker.is_floating_point = std::is_floating_point<Rep>::value;
if (*begin == '.') {
if (std::is_floating_point<Rep>::value)
begin = detail::parse_precision(begin, end, handler);
else
handler.on_error("precision not allowed for this argument type");
checker.has_precision = true;
begin = detail::parse_precision(begin, end, handler);
}
if (begin != end && *begin == 'L') {
++begin;
localized = true;
}
end = detail::parse_chrono_format(begin, end,
detail::chrono_format_checker());
end = detail::parse_chrono_format(begin, end, checker);
return {begin, end};
}

Expand Down
19 changes: 18 additions & 1 deletion test/chrono-test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -466,7 +466,7 @@ TEST(chrono_test, format_default_fp) {

TEST(chrono_test, format_precision) {
EXPECT_THROW_MSG(
(void)fmt::format(runtime("{:.2}"), std::chrono::seconds(42)),
(void)fmt::format(runtime("{:.2%Q}"), std::chrono::seconds(42)),
fmt::format_error, "precision not allowed for this argument type");
EXPECT_EQ("1ms", fmt::format("{:.0}", dms(1.234)));
EXPECT_EQ("1.2ms", fmt::format("{:.1}", dms(1.234)));
Expand Down Expand Up @@ -611,12 +611,28 @@ TEST(chrono_test, cpp20_duration_subsecond_support) {
EXPECT_EQ(fmt::format("{:%S}", std::chrono::nanoseconds{-13420148734}),
"-13.420148734");
EXPECT_EQ(fmt::format("{:%S}", std::chrono::milliseconds{1234}), "01.234");
// Check subsecond presision modifier.
EXPECT_EQ(fmt::format("{:.6%S}", std::chrono::nanoseconds{1234}),
"00.000001");
EXPECT_EQ(fmt::format("{:.18%S}", std::chrono::nanoseconds{1234}),
"00.000001234000000000");
EXPECT_EQ(fmt::format("{:.{}%S}", std::chrono::nanoseconds{1234}, 6),
"00.000001");
EXPECT_EQ(fmt::format("{:.6%S}", std::chrono::milliseconds{1234}),
"01.234000");
EXPECT_EQ(fmt::format("{:.6%S}", std::chrono::milliseconds{-1234}),
"-01.234000");
EXPECT_EQ(fmt::format("{:.3%S}", std::chrono::seconds{1234}), "34.000");
EXPECT_EQ(fmt::format("{:.3%S}", std::chrono::hours{1234}), "00.000");
EXPECT_EQ(fmt::format("{:.5%S}", dms(1.234)), "00.00123");
EXPECT_EQ(fmt::format("{:.8%S}", dms(1.234)), "00.00123400");
{
// Check that {:%H:%M:%S} is equivalent to {:%T}.
auto dur = std::chrono::milliseconds{3601234};
auto formatted_dur = fmt::format("{:%T}", dur);
EXPECT_EQ(formatted_dur, "01:00:01.234");
EXPECT_EQ(fmt::format("{:%H:%M:%S}", dur), formatted_dur);
EXPECT_EQ(fmt::format("{:.6%H:%M:%S}", dur), "01:00:01.234000");
}
using nanoseconds_dbl = std::chrono::duration<double, std::nano>;
EXPECT_EQ(fmt::format("{:%S}", nanoseconds_dbl{-123456789}), "-00.123456789");
Expand All @@ -630,6 +646,7 @@ TEST(chrono_test, cpp20_duration_subsecond_support) {
auto formatted_dur = fmt::format("{:%T}", dur);
EXPECT_EQ(formatted_dur, "-00:01:39.123456789");
EXPECT_EQ(fmt::format("{:%H:%M:%S}", dur), formatted_dur);
EXPECT_EQ(fmt::format("{:.3%H:%M:%S}", dur), "-00:01:39.123");
}
// Check that durations with precision greater than std::chrono::seconds have
// fixed precision, and print zeros even if there is no fractional part.
Expand Down