Re: [PATCH v2 1/1] git-gui: Auto-rescan on activate

[Stefan Haller <stefan@haller-berlin.de>]

On 17.11.20 13:05, Pratyush Yadav wrote:
> On 17/11/20 12:13PM, Stefan Haller wrote:
>> On 17.11.20 8:36, Pratyush Yadav wrote:
>>> Hi Stefan,
>>>
>>> On 14/11/20 08:14PM, Stefan Haller wrote:
>>>> On 03.11.20 17:16, Stefan Haller wrote:
>>>>> Do an automatic rescan whenever the git-gui window receives focus. Most other
>>>>> GUI tools do this, and it's very convenient; no more pressing F5 manually.
>>>>>
>>>>> People who don't like this behavior can turn it off in the Options dialog.
>>>>
>>>> Ping - any thoughts? I have been running with this patch for a few weeks
>>>> now, and I already don't want to miss it any more.
>>>
>>> I have been staring at your patch for the last few days with indecision. 
>>> I have finally made up my mind. I do not think it is a good idea to hurt 
>>> the experience of a significant population of our users without at least 
>>> telling them how they can fix it.
>>>
>>> The config option is not very visible at all. Experience has told me 
>>> that people don't often go looking around in the options menu to find a 
>>> fix for their problem. So we need to do a better job of educating them 
>>> why they might be experiencing slowdowns and how they can avoid them.
>>>
>>> Let's take inspiration from git status. When the local branch diverges 
>>> from the upstream branch, git status shows you how many commits your 
>>> branch is ahead and how many commits behind upstream. This can be an 
>>> expensive operation if the divergence point is far behind. In those 
>>> cases, git status prints something like:
>>>
>>>   It took 30.00 seconds to calculate the branch ahead/behind values.
>>>   You can use '--no-ahead-behind' to avoid this.
>>>
>>> This made me aware this option existed and that I can use it to avoid 
>>> slowdowns.
>>>
>>> We should do something similar for the auto rescan. Measure how long it 
>>> takes to finish the rescan and if it takes longer than X seconds then 
>>> tell the user that they can use this option to disable this. If they 
>>> don't mind the delay they can keep on using it.
>>>
>>> I am working on a patch to add this. Will send it in a day or two.
>>
>> Sounds good to me. While I personally don't think such a check is
>> necessary in this case, I also have nothing against it if you find it
>> important.
>>
>> It just needs to be possible to disable that check itself, too. I
>> certainly wouldn't want to be bothered by it, even if the rescan should
>> take longer than whatever threshold you decide on. So if you put up a
>> dialog to inform the user, the dialog should ideally have a "Don't show
>> again" option.
> 
> That's the plan. It will be a yes/no/cancel prompt. Saying yes or no 
> sets auto rescan to on/off and the message won't pop up again. Saying 
> cancel does nothing and you will see the popup again on the next long 
> rescan.

Interesting. After thinking about this for a while, I'm not convinced
that a Yes/No/Cancel dialog is the best user experience for this, for
the following reasons:

- It isn't obvious whether clicking No will turn auto-rescan off only
  for this repo, or globally (unless you provide two different buttons
  for this).

- It doesn't teach users how they can turn it back on if they clicked No
  too hastily (e.g. because they didn't immediately understand the
  difference between No and Cancel).

- It isn't really intuitively obvious what "Cancel" means. What is the
  operation that is being cancelled here? The rescan itself has already
  happened. (Yes, I know that the operation being cancelled is the
  process of deciding whether the option should be turned off, but as I
  said, I don't find this intuitive.)

I think it might be a better user experience to have a dialog like this:

      The automatic rescan on activating the application has
      taken more than X seconds. If this bothers you, you can
      turn it off in the Preferences dialog.

            [x] Don't show again
                                             [OK]


The only downside is that it's more work to implement, as you can't use
tk_messageBox.

All of this is just my humble opinion; if you decide to stick to your
plan, that's ok with me too.


>>>> Cc:-ing a few people who were involved in the discussion on Pratyush's
>>>> similar patch last summer. [0]
>>>>
>>>>
>>>> [0] <https://lore.kernel.org/git/20190728151726.9188-1-
>>>>      me@yadavpratyush.com/>
>>>>
>>>>
>>>>>
>>>>> Signed-off-by: Stefan Haller <stefan@haller-berlin.de>
>>>>> ---
>>>>>  git-gui.sh     | 5 +++++
>>>>>  lib/option.tcl | 1 +
>>>>>  2 files changed, 6 insertions(+)
>>>>>
>>>>> diff --git a/git-gui.sh b/git-gui.sh
>>>>> index 867b8ce..14735a3 100755
>>>>> --- a/git-gui.sh
>>>>> +++ b/git-gui.sh
>>>>> @@ -906,6 +906,7 @@ set font_descs {
>>>>>  }
>>>>>  set default_config(gui.stageuntracked) ask
>>>>>  set default_config(gui.displayuntracked) true
>>>>> +set default_config(gui.autorescan) true
>>>>>
>>>>>  ######################################################################
>>>>>  ##
>>>>> @@ -4007,6 +4008,10 @@ bind .   <Alt-Key-2> {focus_widget $::ui_index}
>>>>>  bind .   <Alt-Key-3> {focus $::ui_diff}
>>>>>  bind .   <Alt-Key-4> {focus $::ui_comm}
>>>>>
>>>>> +if {[is_config_true gui.autorescan]} {
>>>>> +	bind .   <FocusIn>  { if {"%W" eq "."} do_rescan }
>>>>> +}
>>>>> +
>>>>>  set file_lists_last_clicked($ui_index) {}
>>>>>  set file_lists_last_clicked($ui_workdir) {}
>>>>>
>>>>> diff --git a/lib/option.tcl b/lib/option.tcl
>>>>> index e43971b..9e83db7 100644
>>>>> --- a/lib/option.tcl
>>>>> +++ b/lib/option.tcl
>>>>> @@ -145,6 +145,7 @@ proc do_options {} {
>>>>>  		{b merge.diffstat {mc "Show Diffstat After Merge"}}
>>>>>  		{t merge.tool {mc "Use Merge Tool"}}
>>>>>
>>>>> +		{b gui.autorescan  {mc "Auto-Rescan On Activate"}}
>>>>>  		{b gui.trustmtime  {mc "Trust File Modification Timestamps"}}
>>>>>  		{b gui.pruneduringfetch {mc "Prune Tracking Branches During Fetch"}}
>>>>>  		{b gui.matchtrackingbranch {mc "Match Tracking Branches"}}
>>>>> --
>>>>> 2.29.2
>>>>>
>>>
>