Code::Blocks / Tickets / #851 Show length of selected text

Zdena K - 2019-08-25

Hello, I tried to create a patch for it (this is BlueHazzard's patch in reality). It's my first patch post (and I never worked with svn and so on) so please be tolerant if I did it wrong way.

Edit: Corrected one error in the patch.

Last edit: Zdena K 2019-08-25

SelectedTextLength.patch

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-08-28

As stated in the topic this patch is not acceptable. If we're going to add something here we should make this configurable (with formatting like strings)!

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bluehazzard - 2019-08-29

assigned_to: bluehazzard
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bluehazzard - 2019-08-29

Here is a patch with dynamic formatting and nice tags

fix_851_statusbar_dynamic_formatting.patch

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-08-30

Can you measure the performance impact of this change? This happens very often and shouldn't add delays to the editor!

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Ok, some performance on my pc: All numbers are returned from QueryPerformanceCounter https://docs.microsoft.com/en-us/windows/win32/api/profileapi/nf-profileapi-queryperformancecounter

This is a raw implementation in a new program:

Old implementation (without selection display)
Duration: 41.6641 , min: 35, max: 386, count: 10000
Duration: 43.0685 , min: 34, max: 478, count: 10000
Duration: 41.5154 , min: 35, max: 468, count: 10000
Duration: 40.7124 , min: 35, max: 359, count: 10000
Duration: 42.0436 , min: 35, max: 294, count: 10000
Duration: 40.2498 , min: 34, max: 465, count: 10000
Duration: 42.2045 , min: 35, max: 3513, count: 10000
Duration: 43.1462 , min: 35, max: 520, count: 10000
Duration: 39.3445 , min: 34, max: 261, count: 10000
Duration: 40.3568 , min: 34, max: 313, count: 10000
Duration: 42.0852 , min: 35, max: 418, count: 10000
Duration: 41.8663 , min: 35, max: 454, count: 10000
Duration: 41.0343 , min: 35, max: 458, count: 10000
Duration: 40.3655 , min: 35, max: 340, count: 10000
Duration: 41.511 , min: 35, max: 342, count: 10000
Duration: 42.37 , min: 35, max: 341, count: 10000
Duration: 40.8679 , min: 35, max: 444, count: 10000
Duration: 41.9529 , min: 34, max: 523, count: 10000
Duration: 40.8303 , min: 34, max: 364, count: 10000

New Implementation (with selection display)
Duration: 80.4447 , min: 66, max: 766, count: 10000
Duration: 81.6317 , min: 66, max: 638, count: 10000
Duration: 79.4575 , min: 67, max: 1109, count: 10000
Duration: 77.9361 , min: 67, max: 383, count: 10000
Duration: 81.8575 , min: 67, max: 612, count: 10000
Duration: 97.8307 , min: 67, max: 3672, count: 10000
Duration: 97.6453 , min: 67, max: 598, count: 10000
Duration: 97.7329 , min: 70, max: 539, count: 10000
Duration: 92.5459 , min: 72, max: 357, count: 10000
Duration: 77.3541 , min: 67, max: 745, count: 10000
Duration: 79.6439 , min: 66, max: 674, count: 10000
Duration: 79.6707 , min: 66, max: 518, count: 10000
Duration: 84.6628 , min: 66, max: 643, count: 10000
Duration: 77.4203 , min: 66, max: 987, count: 10000
Duration: 79.6052 , min: 66, max: 594, count: 10000
Duration: 83.6269 , min: 69, max: 3713, count: 10000
Duration: 80.4709 , min: 66, max: 503, count: 10000
Duration: 80.1143 , min: 66, max: 471, count: 10000
Duration: 76.4735 , min: 66, max: 791, count: 10000
Duration: 80.4669 , min: 67, max: 5612, count: 10000

So in the test setup the new implementation is half as fast...

This implementation is in codeblocks under the real code:

Old implementation
Duration: 239.804 , min: 174, max: 486, count: 51
Duration: 232.471 , min: 170, max: 628, count: 51
Duration: 219.647 , min: 115, max: 389, count: 51
Duration: 301.98 , min: 143, max: 1737, count: 51
Duration: 230.392 , min: 137, max: 548, count: 51
Duration: 227.706 , min: 171, max: 555, count: 51
Duration: 221.314 , min: 153, max: 376, count: 51
Duration: 235.039 , min: 173, max: 776, count: 51
Duration: 216.235 , min: 145, max: 368, count: 51

new implementation
Duration: 227.843 , min: 158, max: 425, count: 51
Duration: 230.431 , min: 169, max: 1096, count: 51
Duration: 270.373 , min: 170, max: 1740, count: 51
Duration: 250.902 , min: 144, max: 1320, count: 51
Duration: 253.765 , min: 144, max: 1891, count: 51
Duration: 251.765 , min: 145, max: 1650, count: 51
Duration: 263.922 , min: 165, max: 1140, count: 51
Duration: 327.824 , min: 158, max: 1962, count: 51
Duration: 249.706 , min: 174, max: 1057, count: 51
Duration: 261.392 , min: 172, max: 1575, count: 51

so the real impact is low... Good measurement is hard, because my pc is not on idle...
i am not sure why though

Teodor Petrov - 2019-08-31

What are the units of time? Seconds, miliseconds, microseconds?
And how do you measure inside codeblocks? Why do you have count=51?

If I'm you I'd measure the impact on the whole statusbar handling. If it is not visibly slower I'll be fine. Also make sure you test with all fields added. :) The whole status bar handling must not be above 1ms or ideally not above 0.1ms.

p.s. If you want to do microbenchmarks, just use a benchmark library like google benchmark.
p.p.s. What is your cpu, just for reference?

Last edit: Teodor Petrov 2019-08-31

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

So new benchmark:
Setup:
I use the same timer function as above from windows, Windows 7 wx302, AMD Phenon X2
I measure the time needed to create the status bar in codeblocks. For this i have to move the mouse cursor and select text. I sample 100 times and print statistical data from them:

Old printf
mean: 3.988ms | min: 2.855ms | max:  8.160ms | var: 0.818ms | count: 100
mean: 3.890ms | min: 2.792ms | max:  7.666ms | var: 0.715ms | count: 100
mean: 4.018ms | min: 2.845ms | max: 16.652ms | var: 1.521ms | count: 100
mean: 3.885ms | min: 2.860ms | max:  6.170ms | var: 0.713ms | count: 100
mean: 3.761ms | min: 2.972ms | max:  6.463ms | var: 0.510ms | count: 100
mean: 3.999ms | min: 2.933ms | max:  8.477ms | var: 0.922ms | count: 100
mean: 3.639ms | min: 2.801ms | max:  4.864ms | var: 0.448ms | count: 100
mean: 3.959ms | min: 0.523ms | max:  8.609ms | var: 0.966ms | count: 100
mean: 3.827ms | min: 2.899ms | max:  6.238ms | var: 0.558ms | count: 100
mean: 3.769ms | min: 2.904ms | max:  7.201ms | var: 0.666ms | count: 100
mean: 3.736ms | min: 0.611ms | max:  6.263ms | var: 0.713ms | count: 100
mean: 3.822ms | min: 2.840ms | max:  6.199ms | var: 0.655ms | count: 100
mean: 3.668ms | min: 2.870ms | max:  6.688ms | var: 0.587ms | count: 100
mean: 3.683ms | min: 2.865ms | max:  5.749ms | var: 0.510ms | count: 100

New implementation: print string: Line [line], Col [col], Pos [pos]
mean: 4.036ms | min: 0.601ms | max: 19.521ms | var: 1.913ms | count: 100
mean: 4.393ms | min: 3.207ms | max: 15.268ms | var: 1.289ms | count: 100
mean: 4.381ms | min: 3.012ms | max:  8.194ms | var: 0.904ms | count: 100
mean: 4.242ms | min: 3.139ms | max:  6.717ms | var: 0.715ms | count: 100
mean: 4.253ms | min: 3.144ms | max:  9.284ms | var: 0.987ms | count: 100
mean: 4.116ms | min: 3.021ms | max:  9.079ms | var: 0.875ms | count: 100
mean: 4.423ms | min: 2.870ms | max:  9.264ms | var: 1.019ms | count: 100
mean: 4.441ms | min: 2.958ms | max:  9.939ms | var: 1.236ms | count: 100
mean: 4.279ms | min: 2.889ms | max: 11.684ms | var: 1.180ms | count: 100
mean: 4.352ms | min: 2.919ms | max: 12.398ms | var: 1.183ms | count: 100
mean: 4.167ms | min: 2.919ms | max:  9.758ms | var: 0.882ms | count: 100
mean: 4.469ms | min: 3.036ms | max: 10.902ms | var: 0.967ms | count: 100
mean: 4.226ms | min: 2.958ms | max:  7.548ms | var: 0.749ms | count: 100
mean: 4.267ms | min: 3.276ms | max: 12.237ms | var: 0.981ms | count: 100

New implementation: print string Line [line], Col [col], Pos [pos], Sel[sel] (aka this patch)
mean: 3.778ms | min: 0.670ms | max:  9.167ms | var: 1.097ms | count: 100        
mean: 3.827ms | min: 2.889ms | max: 10.516ms | var: 0.876ms | count: 100        
mean: 3.711ms | min: 2.767ms | max:  5.348ms | var: 0.526ms | count: 100        
mean: 3.863ms | min: 2.806ms | max:  9.680ms | var: 0.884ms | count: 100        
mean: 3.909ms | min: 2.699ms | max:  7.226ms | var: 0.822ms | count: 100        
mean: 3.889ms | min: 2.752ms | max:  6.835ms | var: 0.808ms | count: 100        
mean: 3.750ms | min: 2.767ms | max:  5.759ms | var: 0.499ms | count: 100        
mean: 3.806ms | min: 2.845ms | max:  8.428ms | var: 0.868ms | count: 100        
mean: 3.899ms | min: 2.660ms | max: 12.667ms | var: 1.096ms | count: 100        
mean: 3.897ms | min: 2.660ms | max: 10.624ms | var: 1.011ms | count: 100        
mean: 4.034ms | min: 2.645ms | max:  7.993ms | var: 0.825ms | count: 100        
mean: 3.693ms | min: 2.772ms | max:  7.128ms | var: 0.737ms | count: 100        
mean: 3.596ms | min: 2.723ms | max:  5.500ms | var: 0.457ms | count: 100        
mean: 3.693ms | min: 2.855ms | max:  6.805ms | var: 0.577ms | count: 100        

Old printf with position arguments without selection (use "Line %1$d, Col %2$d, Pos %3$d"
mean: 3.968ms | min: 0.396ms | max: 16.089ms | var: 1.440ms | count: 100
mean: 3.922ms | min: 2.899ms | max:  8.663ms | var: 0.776ms | count: 100
mean: 3.771ms | min: 2.938ms | max:  6.395ms | var: 0.505ms | count: 100
mean: 3.810ms | min: 2.992ms | max:  8.653ms | var: 0.670ms | count: 100
mean: 3.890ms | min: 0.660ms | max:  9.245ms | var: 0.878ms | count: 100
mean: 3.813ms | min: 2.914ms | max:  6.517ms | var: 0.571ms | count: 100
mean: 3.867ms | min: 2.889ms | max:  5.769ms | var: 0.572ms | count: 100
mean: 4.039ms | min: 2.870ms | max:  8.311ms | var: 0.986ms | count: 100
mean: 4.094ms | min: 2.928ms | max:  7.460ms | var: 0.828ms | count: 100
mean: 3.787ms | min: 2.796ms | max:  6.409ms | var: 0.610ms | count: 100
mean: 3.940ms | min: 2.919ms | max:  6.605ms | var: 0.723ms | count: 100
mean: 4.036ms | min: 2.948ms | max:  6.605ms | var: 0.677ms | count: 100
mean: 4.002ms | min: 2.933ms | max:  9.832ms | var: 0.976ms | count: 100
mean: 4.189ms | min: 2.924ms | max: 10.027ms | var: 1.141ms | count: 100

Old printf with position arguments without selection
mean: 4.735ms | min: 0.631ms | max: 16.089ms | var: 1.738ms | count: 100
mean: 4.949ms | min: 3.852ms | max:  9.255ms | var: 0.868ms | count: 100
mean: 5.291ms | min: 2.924ms | max: 23.129ms | var: 2.254ms | count: 100
mean: 5.321ms | min: 3.261ms | max: 18.167ms | var: 1.854ms | count: 100
mean: 5.168ms | min: 3.623ms | max: 22.269ms | var: 2.023ms | count: 100
mean: 4.740ms | min: 3.711ms | max:  7.402ms | var: 0.766ms | count: 100
mean: 4.884ms | min: 3.046ms | max: 14.549ms | var: 1.441ms | count: 100
mean: 4.657ms | min: 0.733ms | max:  8.448ms | var: 1.066ms | count: 100
mean: 5.092ms | min: 3.755ms | max: 14.334ms | var: 1.672ms | count: 100
mean: 4.968ms | min: 3.735ms | max: 11.719ms | var: 1.255ms | count: 100
mean: 5.005ms | min: 3.725ms | max: 15.136ms | var: 1.436ms | count: 100
mean: 4.829ms | min: 3.750ms | max: 29.099ms | var: 2.565ms | count: 100
mean: 4.915ms | min: 3.662ms | max: 15.121ms | var: 1.447ms | count: 100
mean: 5.129ms | min: 3.354ms | max: 10.653ms | var: 1.236ms | count: 100

And because this is not nice to see i made a plot and attached it

status_bar_build_times.svg

bluehazzard - 2019-09-01

Conclusion, this patch does not make things slower in avarage. The build time is ca 4ms of the whole status bar

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-09-02

Can I see your measuring code? 4ms for statusbar refresh is massive amount of time. :(

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

calculation of values is not optimized, because it is outside of the time measurement

static std::vector<int64_t> summ;

void MainFrame::DoUpdateStatusBar()
{
    if (!GetStatusBar())
        return;
    if (Manager::IsAppShuttingDown())
        return;

    LARGE_INTEGER freq, time_begin, time_end;

    cbEditor* ed = Manager::Get()->GetEditorManager()->GetBuiltinActiveEditor();
    wxString personality(Manager::Get()->GetPersonalityManager()->GetPersonality());
    if (ed)
    {
        QueryPerformanceCounter(&time_begin);
        cbStyledTextCtrl * const control = ed->GetControl();

        int panel = 0;
        int pos = control->GetCurrentPos();
        wxString msg;
        SetStatusText(ed->GetFilename(), panel++);

        // Highlightbutton
        if (m_pHighlightButton)
        {
            EditorColourSet* colour_set = Manager::Get()->GetEditorManager()->GetColourSet();
            if (colour_set)
                changeButtonLabel(*m_pHighlightButton, colour_set->GetLanguageName(ed->GetLanguage()));
            else
                changeButtonLabel(*m_pHighlightButton, wxEmptyString);
        }
        // EOL mode
        panel++;
        switch (control->GetEOLMode())
        {
            case wxSCI_EOL_CRLF: msg = _T("Windows (CR+LF)"); break;
            case wxSCI_EOL_CR:   msg = _T("Mac (CR)");        break;
            case wxSCI_EOL_LF:   msg = _T("Unix (LF)");       break;
            default:                                          break;
        }
        SetStatusText(msg, panel++);
        SetStatusText(ed->GetEncodingName(), panel++);

        wxString statusFormat(m_StatusBarFormatString);
        statusFormat.Replace(_T("[line]"), wxString::Format(_T("%d"), control->GetCurrentLine() + 1));
        statusFormat.Replace(_T("[col]"), wxString::Format(_T("%d"), control->GetColumn(pos) + 1));
        statusFormat.Replace(_T("[pos]"), wxString::Format(_T("%d"), pos));
        statusFormat.Replace(_T("[sel]"), wxString::Format(_T("%d"), control->GetSelectionEnd() - control->GetSelectionStart()));
        SetStatusText(statusFormat, panel++);
        //msg.Printf(_("Line %3$d, Col %2$d, Pos %1$d, Sel %4$d"),pos, control->GetCurrentLine() + 1, control->GetColumn(pos) + 1, control->GetSelectionEnd() - control->GetSelectionStart());
        //SetStatusText(msg, panel++);

        SetStatusText(control->GetOvertype() ? _("Overwrite") : _("Insert"), panel++);
#if wxCHECK_VERSION(3, 0, 0)
        SetStatusText(ed->GetModified() ? _("Modified") : _T(""), panel++);
#else
        SetStatusText(ed->GetModified() ? _("Modified") : wxEmptyString, panel++);
#endif
        SetStatusText(control->GetReadOnly() ? _("Read only") : _("Read/Write"), panel++);
        SetStatusText(personality, panel++);

        QueryPerformanceCounter(&time_end);

        int64_t su = time_end.QuadPart - time_begin.QuadPart;

        summ.push_back(su);

        if(summ.size() == 100)
        {
            QueryPerformanceFrequency(&freq);

            int64_t summe = 0;
            int64_t smax = std::numeric_limits<int64_t>::min();
            int64_t smin = std::numeric_limits<int64_t>::max();
            for(auto item : summ)
            {
                smax = std::max(item, smax);
                smin = std::min(item, smin);
                summe += item;
            }

            const double mean = (double) summe / summ.size();

            double deviation = 0.0;

            for(auto item : summ)
                deviation += (item - mean) * (item - mean);

            deviation = deviation / summ.size();

            deviation = (sqrt(deviation) * 10E3) / (double) freq.QuadPart;
            const double tmax = (double) smax * 10E3 / (double) freq.QuadPart;
            const double tmin = (double) smin * 10E3 / (double) freq.QuadPart;
            const double tmean = (double) mean * 10E3 / (double) freq.QuadPart;

            wxString out = wxString::Format("mean: %3.3lfms | min: %03.3lfms | max: %03.3lfms | var: %03.3lfms | count: %d", tmean, tmin, tmax, deviation, (int) summ.size());
            Manager::Get()->GetLogManager()->Log(out);
            summ.clear();
        }
    }
    else
    {
        int panel = 0;
        EditorBase *eb = Manager::Get()->GetEditorManager()->GetActiveEditor();
        if ( eb )
            SetStatusText(eb->GetFilename(), panel++);
        else
            SetStatusText(_("Welcome to ") + appglobals::AppName + _T("!"), panel++);

        // Highlightbutton
        if (m_pHighlightButton)
            changeButtonLabel(*m_pHighlightButton, wxEmptyString);
        panel++;

        SetStatusText(wxEmptyString, panel++);
        SetStatusText(wxEmptyString, panel++);
        SetStatusText(wxEmptyString, panel++);
        SetStatusText(wxEmptyString, panel++);
        SetStatusText(wxEmptyString, panel++);
        SetStatusText(wxEmptyString, panel++);
        SetStatusText(personality, panel++);
    }
}

[edit:] fixed code... had some errors in it

Last edit: bluehazzard 2019-09-03

bluehazzard - 2019-09-03

i forgot to mention: i do compile with -g and no optimization...
[Edit:] With optimization 03 it is not better... avg 3.6ms with patch from top

Last edit: bluehazzard 2019-09-03

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-09-03

Why don't you measure the whole function?

Why do you do batches of 100, just calc the stats on the whole sample set.

Are you sure there is no frequency scaling?

Is everything else stopped (browser, chat, something else)?

Still 3.6ms is pretty expensive for this function. What is the file you're editing.

I guess you can push this change and we could optimize it later.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bluehazzard - 2019-09-03

Why don't you measure the whole function?
Does it really make a difference? The workload that needs the time (and we want to optimize) is in the time measurement part

Why do you do batches of 100, just calc the stats on the whole sample set.
I do want to mitigate CPU spikes by using batches of the avarage and find and ignore lokal maxima. Also i want to move the mouse only a few times until i see a result...

Are you sure there is no frequency scaling?
Of what? CPU? Can be...

Is everything else stopped (browser, chat, something else)?
No my normal programs are open (Browser, that even takes some CPU in the 5% (thank you fu java script and bloated webpages))

Still 3.6ms is pretty expensive for this function. What is the file you're editing.
Some normal 300 line code... I too think 3.6ms is too much, but i do not notice any slowdown, flashing or other fuzzyness. And my working PC is 10 Years old...

I guess you can push this change and we could optimize it later.
I will experiment a bit and if i do not find any speedup i will push it...

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Zdena K - 2019-09-06

If I can say anything to it, it works for me. And I don't see any influence on performance. No measurements, only feeling. Well done, bluehazzard.

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bluehazzard - 2019-09-18

Here is a other idea... No cleanup, and there is still a bug when you change the status bar setting, i just wanted to know if this idea pleases more.

(sadly c++ still does not have fast, typesafe flags...)

Clean up formatting is needed, also probably more encapsulation, but it is to late for this now.... But this concept would add nearly no overhead over the old system...

Last edit: bluehazzard 2019-09-18

status_bar_with_flags.patch

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-09-20

Does it imporve performance? If it doesn't there is no need for such additional complexity.

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bluehazzard - 2019-09-25

Ok this is my final proposal.
The performance impact for the default string replacing is minimal. Adding the selection part does slow down a bit, but there is scintilla the slow part. With the implemented flags this does not occure if the selection is not put in the status bar.

i choosed the struct way over the previous type safe flag way for readability... I checked the performace of assambler code and it si as good as it can get when -O3 is enabled...

statusBar_final_1.patch

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Teodor Petrov - 2019-09-27

What are the numbers?

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Show length of selected text

A free C, C++ and Fortran IDE

Milestone

Searches

Help

#851 Show length of selected text

Discussion