More human readable elapsed time

[xiaozhuai]

std::cout << format_time_userfriendly(std::chrono::milliseconds(3600l * 1000l + 123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(3600l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(24l * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(25l * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds((1l + 24l * 7l) * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";

It will prints

1h 0.123s
1h 5.123s
1d 8m 5.123s
1d 1h 8m 5.123s
7d 1h 8m 5.123s

The delimiter currently is default to ' ', it can be '' and prints

1h0.123s
1h5.123s
1d8m5.123s
1d1h8m5.123s
7d1h8m5.123s

current log

Elapsed time to handle qt5-base:x64-osx: 8.504 min

8.504 min is confusing, that will be

Elapsed time to handle qt5-base:x64-osx: 8m 30.240s

[xiaozhuai]

The CI fails, someone please help

[autoantwort]

Yeah the ci is broken since #773

[BillyONeal]

CI fixed in 17bb1b8, please merge

[xiaozhuai]

CI fixed in 17bb1b8, please merge

I've merge to master and it fails with different problem

[BillyONeal]

CI fixed in 17bb1b8, please merge

I've merge to master and it fails with different problem

That looks like a legitimate problem created by this PR, not a baseline issue. (You can download the diff and git apply it)

[xiaozhuai]

That looks like a legitimate problem created by this PR, not a baseline issue. (You can download the diff and git apply it)

But OSX and Linux build passing. So odd.

[vicroms]

The issue is the code formatting, you need to apply the diff generated by CI to format it correctly. Alternatively, if you have clang format installed you can run the format CMake target to format all the sources.

[xiaozhuai]

The issue is the code formatting, you need to apply the diff generated by CI to format it correctly. Alternatively, if you have clang format installed you can run the format CMake target to format all the sources.

Got it, thx!

[dg0yt]

IMO the proposed output is still somewhat controversial with regard to readability. I wonder if there are any external examples of good practices.
There are also too many digits of little significance.
My wish list is probably like:

59.9 s
1:01 min
2:59 h
1d 2:49 h

[xiaozhuai]

std::cout << format_time_userfriendly(std::chrono::milliseconds(3600l * 1000l + 123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(3600l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(24l * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds(25l * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";
std::cout << format_time_userfriendly(std::chrono::milliseconds((1l + 24l * 7l) * 3600l * 1000l + 60l * 8l * 1000l + 5123l)) << "\n";

For these cases, what format do you prefer?

[dg0yt]

I would cut the line for my examples at unit boundaries. For you initial examples, my personal proposal results in:

1:00 h
1:00 h
1:08 h
1d 1:08 h
7d 1:08 h

More "precision" is just noise.

But others should have a chance to give feedback, too.

[xiaozhuai]

Let's hear what others have to say.

[xiaozhuai]

What about this

01:00:00.123
01:00:05.123
1d 00:08:05.123
1d 01:08:05.123
7d 01:08:05.123

[dg0yt]

What about this

No. I think this perfectly machine readable. But the PR title promises something else.

[autoantwort]

Personally I find 8.504 min not confusing and faster to read than 8m 30.240s 😅 But it could be shortened to 8.5 min

[xiaozhuai]

Probably this problem is difficult to agree on, so I will close this pr.