Consider dropping the decimal places for KiB/s 52.00 KiB/s

Consider dropping the decimal places for KiB/s 52.00 KiB/s

[Jonny Grant]

Hello

May I suggest taking off the .00 KiB/s suffix, has that been considered? As the decimal places don't appear to change, they're stuck on .00.

I noticed the KiB/s is neatly rounded to the kibibyte.

Apologies I'm not using latest release. 
git version 2.40.1


$ git clone git://gcc.gnu.org/git/gcc.git git_1
Cloning into 'git_1'...
remote: Enumerating objects: 2949348, done.
remote: Counting objects: 100% (209238/209238), done.
remote: Compressing objects: 100% (14579/14579), done.
Receiving objects:   7% (210878/2949348), 76.33 MiB | 52.00 KiB/s


Meanwhile I'll keep waiting for this slow server.

Kind regards, Jonny

Re: Consider dropping the decimal places for KiB/s 52.00 KiB/s

[Taylor Blau]

On Thu, Nov 30, 2023 at 06:11:57PM +0000, Jonny Grant wrote:
> Hello
>
> May I suggest taking off the .00 KiB/s suffix, has that been
> considered? As the decimal places don't appear to change, they're
> stuck on .00.

I wonder if you have a throttled connection that is locked to 52KiB/s
exactly. The progress code that generates the throughput is in
progress.c::display_throughput(), which computes the rate. It's computed
in bytes/misec, and then passed to throughput_string() (really,
`strbuf_humanise_rate()`), which formats it appropriately.

If you're in the KiB range, it will print the decimal component, which
is:

    ((bytes & ((1<<10)-1)) * 100) >> 10

> $ git clone git://gcc.gnu.org/git/gcc.git git_1
> Cloning into 'git_1'...
> remote: Enumerating objects: 2949348, done.
> remote: Counting objects: 100% (209238/209238), done.
> remote: Compressing objects: 100% (14579/14579), done.
> Receiving objects:   7% (210878/2949348), 76.33 MiB | 52.00 KiB/s

On my machine:

    $ git.compile clone git://gcc.gnu.org/git/gcc.git
    [...]
    Receiving objects:  11% (342176/2949348), 108.09 MiB | 24.01 MiB/s

I suppose we could consider dropping the decimal component if it's a
round number, but I think that it may produce awkward flickering if the
rate oscillates between a round number and a non-round number.

Thanks,
Taylor

Re: Consider dropping the decimal places for KiB/s 52.00 KiB/s

[Dragan Simic]

On 2023-11-30 20:07, Taylor Blau wrote:
> On Thu, Nov 30, 2023 at 06:11:57PM +0000, Jonny Grant wrote:
>> $ git clone git://gcc.gnu.org/git/gcc.git git_1
>> Cloning into 'git_1'...
>> remote: Enumerating objects: 2949348, done.
>> remote: Counting objects: 100% (209238/209238), done.
>> remote: Compressing objects: 100% (14579/14579), done.
>> Receiving objects:   7% (210878/2949348), 76.33 MiB | 52.00 KiB/s
> 
> On my machine:
> 
>     $ git.compile clone git://gcc.gnu.org/git/gcc.git
>     [...]
>     Receiving objects:  11% (342176/2949348), 108.09 MiB | 24.01 MiB/s
> 
> I suppose we could consider dropping the decimal component if it's a
> round number, but I think that it may produce awkward flickering if the
> rate oscillates between a round number and a non-round number.

You're right, the resulting flickering would look really annoying.  In 
fact, I already modified the reported download speed in another project 
to avoid pretty much the same kind of flickering, and it looked much 
better without it.

Re: Consider dropping the decimal places for KiB/s 52.00 KiB/s

[Jonny Grant]

On 30/11/2023 19:07, Taylor Blau wrote:
> On Thu, Nov 30, 2023 at 06:11:57PM +0000, Jonny Grant wrote:
>> Hello
>>
>> May I suggest taking off the .00 KiB/s suffix, has that been
>> considered? As the decimal places don't appear to change, they're
>> stuck on .00.
> 
> I wonder if you have a throttled connection that is locked to 52KiB/s
> exactly.

You're right - I changed to https and it went to 6 MiB/s. Seems throttled on git://
I should have checked that first.

>  The progress code that generates the throughput is in
> progress.c::display_throughput(), which computes the rate. It's computed
> in bytes/misec, and then passed to throughput_string() (really,
> `strbuf_humanise_rate()`), which formats it appropriately.
> 
> If you're in the KiB range, it will print the decimal component, which
> is:
> 
>     ((bytes & ((1<<10)-1)) * 100) >> 10
> 
>> $ git clone git://gcc.gnu.org/git/gcc.git git_1
>> Cloning into 'git_1'...
>> remote: Enumerating objects: 2949348, done.
>> remote: Counting objects: 100% (209238/209238), done.
>> remote: Compressing objects: 100% (14579/14579), done.
>> Receiving objects:   7% (210878/2949348), 76.33 MiB | 52.00 KiB/s
> 
> On my machine:
> 
>     $ git.compile clone git://gcc.gnu.org/git/gcc.git
>     [...]
>     Receiving objects:  11% (342176/2949348), 108.09 MiB | 24.01 MiB/s
> 
> I suppose we could consider dropping the decimal component if it's a
> round number, but I think that it may produce awkward flickering if the
> rate oscillates between a round number and a non-round number.

Now I can see the advantage of it as it is. wget is similar.
Kind regards, Jonny