Elapsed time larger than 6 digits ignores alignment width

[ivan236634452]

Tested in release 1.4.2 and today's v1.x branch. All alignments are affected.

Code to reproduce (right alignment):

spdlog::set_pattern("|%12u|");
spdlog::info("");
std::this_thread::sleep_for(std::chrono::milliseconds(1500));
spdlog::info("");
spdlog::info("");

spdlog::set_pattern("|%12i|");
spdlog::info("");
std::this_thread::sleep_for(std::chrono::milliseconds(1500));
spdlog::info("");
spdlog::info("");

Expected output:

|      001100|
|  1514900200|
|      173200|
|      000004|
|     1515305|
|      000196|

Actual output:

|      001100|
|      1514900200|
|      173200|
|      000004|
|      1515305|
|      000196|

[gabime]

Thanks for reporting.

[ivan236634452]

The culprit is "format" method in elapsed_formatter class in "pattern_formatter-inl.h":

void format(const details::log_msg &msg, const std::tm &, memory_buf_t &dest) override
{
    auto delta = (std::max)(msg.time - last_message_time_, log_clock::duration::zero());
    auto delta_units = std::chrono::duration_cast<DurationUnits>(delta);
    last_message_time_ = msg.time;
    ScopedPadder p(6, padinfo_, dest);
    fmt_helper::pad6(static_cast<size_t>(delta_units.count()), dest);
}

I would've actually preferred to get rid of leading zeroes altogether. For me this:

|         2 |
|     14189 |
|        30 |
|         0 |
|       178 |
|        10 |
|         1 |
|    345212 |

reads easier than this:

|    000002 |
|    014189 |
|    000030 |
|    000000 |
|    000178 |
|    000010 |
|    000001 |
|    345212 |

My current fix is this:

void format(const details::log_msg &msg, const std::tm &, memory_buf_t &dest) override
{
    auto delta = (std::max)(msg.time - last_message_time_, log_clock::duration::zero());
    auto delta_units = std::chrono::duration_cast<DurationUnits>(delta).count();
    last_message_time_ = msg.time;
    const auto field_size = fmt_helper::count_digits(delta_units);
    ScopedPadder p(field_size, padinfo_, dest);
    fmt_helper::append_int(delta_units, dest);
}

But it is slower (probably), and breaks backward compatibility (to a significantly greater extent). But I'm just a scrub user, not a library designer, and can afford to not care about the latter.

[gabime]

@ivan236634452 Fixed. I agree it is more readable, so I changed according to your proposal. It is also few nanos faster!