Holding down a key still generates a lot of duplicated logs.

[tlckfeks]

#15 issue has not been fixed yet.
Hold down a key still generates a lot of duplicated logs.

[suurjaak]

What happens if you close the program, set KeyboardJoinInterval to a higher value in the configuration file, e.g. to 0.5, and re-launch the program?

[tlckfeks]

I tried changing "KeyboardJoinInterval" to some values.
Unfortunately I don't need KeyboardJoinInterval at all.
I often hold down one key for a long time.
Most keylog software recognizes holding a key down as a single press.
And whether I press a key 5 or 10 times per second, They accurately recognize and record it.
The log of InputScope is still confuses me. :(

[suurjaak]

I tried changing "KeyboardJoinInterval" to some values.

Unfortunately I don't need KeyboardJoinInterval at all.

I am rather confused about how these two sentences add up.

[tlckfeks]

It means that I tried modifying the value of "KeyboardJoinInterval" to 1, 0.5, etc.
And that didn't help the Inputscope to accurately count the physical keystrokes.

[suurjaak]

Are you saying that modifying KeyboardJoinInterval had no effect whatsoever?

Are you sure modified the right configuration file, and that you do not have more than instance of InputScope running?

[tlckfeks]

Let's take an example.
Suppose I set the value of KeyboardJoinInterval to '0.5' and I pressed R key 5 times for 1 second at a constant rate.
In this case, There are 1 log of the R key in Inputscope.db.
Did I press R key once? No, I pressed the R key 5 times.

And Suppose I set the value of KeyboardJoinInterval to '0.05' and I hold down R key for 5 seconds.
In this case, there are more than 10 logs of R key in Inputscope.db.
Did I press R key more than 10 times? No, I just pressed the R key once.

To be precise, Resetting the value of KeyboardJoinInterval had an effect on this issue.
But I think It just doesn't help the Inputscope to accurately count the physical keystrokes.

[suurjaak]

Okay, so the parameter works, just that the default value of 0.05 is not a good fit for your setup.

If you set it to 0.1, does it still count a long keypress as multiple events?

[tlckfeks]

I set it to 0.1 and checked that it counts a long key press as two events or sometimes more than 3 events.
And if I quickly press a key repeatedly for some reason, Inputscope can ignore some events.
I've been testing keystrokes with InputScope, Mousotron and Keyplus.
I checked that InputScope might ignore some events.
I don't know much about programming, but I think InputScope is very inaccurate for counting keystrokes.
In contrast, KeyPlus and Mousotron accurately count keystrokes.

[suurjaak]

This is mostly a limitation of the underlying input listener library that InputScope uses. It does not provide information whether a key is physically depressed, it only provides keyboard events like typical programs receive. So the only way it can tell that a series of key events is a long keypress, is looking at the time intervals and using the KeyboardJoinInterval as a threshold.

A long keypress registering as two events is expected. When you press and hold a key like S, there is a delay between the first and second s produced, a delay that InputScope needs to count as two events - because it is indistinguishable from somebody typing really fast.

Out of interest - what are your keyboard settings for "repeat delay" and "repeat rate" options in Windows settings?

[tlckfeks]

delay : short
rate : fast

Inputscope also counts keystrokes inaccurately by CPU usage.
I tested this on an older computer. The value of KeyboardJoinInterval is '0.1'
Depending on the CPU utilization, the result of counting keystrokes may always be different.
If CPU usage is high, Inputscope will count keystrokes more inaccurately.
In contrast, KeyPlus and Mousotron accurately count keystrokes.
https://player.vimeo.com/video/731268286

[suurjaak]

The short delay setting explains why 0.05 is too low for your setup.

CPU usage should not be a factor. Bear in mind that there is a delay of up to 5 seconds between an input event occurring and the input event being visible in statistics. InputScope does database writes in bursts, for lesser overhead.

[tlckfeks]

What is the difference in the way KeyPlus, Whatpulse, Mousotron and Inputscope count keystrokes?
Unlike Inputscope, They do not require any configuration modifications by users.
Besides, They count physical keystrokes accurately.
As a result, I am using other software for counting keystrokes.
Inputscope seems very unstable at counting keystrokes.

[suurjaak]

Looking into things further, I see that I misspoke earlier - it is actually very possible for InputScope to detect whether a key is being held down physically.

If memory serves, I took the current approach because I wanted to count all the keys produced, not only whether a key had been depressed. I had quite forgotten about this.

I made a few changes, can try this version: https://erki.lap.ee/stuff/inputscope_1.7.dev1_x64.exe

It should be able to count long keypresses as a single event now. There is also an option in the tray menu for toggling this sticky mode off or on.

[tlckfeks]

I have tried the dev version.
In fact, it seemed to have taken a considerable amount of time,
so I decided not to take any further interest in this issue.
But finally this issue has been fixed. I am glad that this issue has been resolved.

[suurjaak]

Released version 1.7 with sticky long keypress as the default setting.

Thank you for the thorough reporting.