fish-shell/doc_src/cmds/time.rst
Johannes Altmanninger 414d9a1eb1 Reference more non-fish shell builtins that have relevant differences
When writing scripts for other shells, it can be confusing and annoying
that our `man` function shadows other manual pages, for example `exec(1p)`
from [Linux man-pages]. I almost never want to see the fish variant for such
contended cases (which obviuosly don't include fish-specific commands like
`string`, only widely-known shell builtins).

For the contented cases like `exec`, the POSIX documentation is more
substantial and useful, since it describes a (sub)set of languages widely
used for scripting.

Because of this I think we should stop overriding the system's man pages.
Nowadays we offer `exec -h` as intuitive way to show the documentation for
the fish-specific command (note that `help` is not a good replacement because
it uses a web browser).

Looking through the contended commands, it seems like for most of them,
the fish version is not substantially different from the system version.
A notable exception is `read` but I don't think it's a very important one.

So I think we should can sacrifice a bit of the native fish-scripting
experience in exchange for playing nicer with other shells. I think the
latter is more important because scripting is not our focus, the way I see it.
So maybe put our manpath at the end.

In lieu of that, let's at least have `exec.rst` reference the system variant.

[Linux man-pages]: https://www.kernel.org/doc/man-pages/

Closes #10376
2024-04-20 13:34:08 +02:00

81 lines
3.5 KiB
ReStructuredText

.. _cmd-time:
time - measure how long a command or block takes
================================================
Synopsis
--------
.. synopsis::
time COMMAND
Description
-----------
.. only:: builder_man
NOTE: This page documents the fish keyword ``time``.
To see the documentation on any non-fish versions, use ``command man time``.
``time`` causes fish to measure how long a command takes and print the results afterwards. The command can be a simple fish command or a block. The results can not currently be redirected.
For checking timing after a command has completed, check :ref:`$CMD_DURATION <variables-special>`.
Your system most likely also has a ``time`` command. To use that use something like ``command time``, as in ``command time sleep 10``. Because it's not inside fish, it won't have access to fish functions and won't be able to time blocks and such.
How to interpret the output
---------------------------
Time outputs a few different values. Let's look at an example::
> time string repeat -n 10000000 y\n | command grep y >/dev/null
________________________________________________________
Executed in 805.98 millis fish external
usr time 798.88 millis 763.88 millis 34.99 millis
sys time 141.22 millis 40.20 millis 101.02 millis
The time after "Executed in" is what is known as the "wall-clock time". It is simply a measure of how long it took from the start of the command until it finished. Typically it is reasonably close to :envvar:`CMD_DURATION`, except for a slight skew because the two are taken at slightly different times.
The other times are all measures of CPU time. That means they measure how long the CPU was used in this part, and they count multiple cores separately. So a program with four threads using all CPU for a second will have a time of 4 seconds.
The "usr" time is how much CPU time was spent inside the program itself, the "sys" time is how long was spent in the kernel on behalf of that program.
The "fish" time is how much CPU was spent in fish, the "external" time how much was spent in external commands.
So in this example, since ``string`` is a builtin, everything that ``string repeat`` did is accounted to fish. Any time it spends doing syscalls like ``write()`` is accounted for in the fish/sys time.
And ``grep`` here is explicitly invoked as an external command, so its times will be counted in the "external" column.
Note that, as in this example, the CPU times can add up to more than the execution time. This is because things can be done in parallel - ``grep`` can match while ``string repeat`` writes.
Example
-------
(for obvious reasons exact results will vary on your system)
::
>_ time sleep 1s
________________________________________________________
Executed in 1,01 secs fish external
usr time 2,32 millis 0,00 micros 2,32 millis
sys time 0,88 millis 877,00 micros 0,00 millis
>_ time for i in 1 2 3; sleep 1s; end
________________________________________________________
Executed in 3,01 secs fish external
usr time 9,16 millis 2,94 millis 6,23 millis
sys time 0,23 millis 0,00 millis 0,23 millis
Inline variable assignments need to follow the ``time`` keyword::
>_ time a_moment=1.5m sleep $a_moment
________________________________________________________
Executed in 90.00 secs fish external
usr time 4.62 millis 4.62 millis 0.00 millis
sys time 2.35 millis 0.41 millis 1.95 millis