Find and Add To Selection In 7.7

Alan Kilborn

See one of the sub-points of this thread…meaning that you can currently do it with LuaScript.

Pythonscript will need to be updated to include the functionality, and it just hasn’t been done yet.

Of course, Notepad++ itself could be updated as well to do it…

cipher-1024

Thanks Alan, I need to get over to plug-in development more often. Hopefully the Python maintainers are able to get something out soon. Or maybe @dail converted me to Lua!

dinkumoil

@dail , @cipher-1024 , @Alan-Kilborn and all

I’ve tried to get the benefit of SCI_MULTIPLESELECTADDNEXT and SCI_MULTIPLESELECTADDEACH following the advice of @dail in this already mentioned thread, but I wasn’t successful.

I’ve set up the little script of @dail as a LUA Script plugin startup script and restarted Notepad++. I can confirm that after that there was a new entry in the submenu of the LUA Script plugin showing the expected text and keyboard shortcut. I selected a word by double-clicking on it or pressing CTRL+Right.The word became the main selection and all other instances of it in the whole text became marked. Then I pressed the keyboard shortcut assigned to the new Scintilla features, but I got no multiple cursors, nothing happened.

Of course I’ve used Notepad++ v7.7 and the latest version of the LUA Script plugin, i.e. v0.9 from the provided download link.

I’ve tried something similar using the NppExec plugin with the bare numeric message codes for SCI_MULTIPLESELECTADDNEXT (2688) and SCI_MULTIPLESELECTADDEACH (2689) but it didn’t work either.

Has anybody some advice whether there are some Notepad++ settings that might interfere with SCI_MULTIPLESELECTADDNEXT and SCI_MULTIPLESELECTADDEACH to prevent them from working properly?

Ekopalypse

@dinkumoil
you need to set target start and end before calling one of the new functions.

dinkumoil

@Ekopalypse

Thank you, that helped! I’ve used the NppExecplugin and the sequence:

Select a word.
Send SCI_TARGETWHOLEDOCUMENT (message code 2690) to set Scintilla’s target.
Send SCI_MULTIPLESELECTADDEACH (message code 2689) to multi-edit all instances of the word.

Ekopalypse

@dinkumoil

if different findoption should be used you can add a step 2a. SCI_SETSEARCHFLAGS

dinkumoil

@Ekopalypse

Thanks for that additional hint, much appreciated!

Here is my complete NppExec script:

sci_sendmsg SCI_GETSELECTIONEMPTY

if $(MSG_RESULT) == 1 then
  sci_sendmsg SCI_WORDRIGHTEXTEND
endif

sci_sendmsg SCI_SETSEARCHFLAGS SCFIND_WHOLEWORD 

sci_sendmsg 2690
sci_sendmsg 2689

It automatically selects the word right from the cursor but only if there is no active selection.

Ekopalypse

@dinkumoil - nice one, thanks for sharing.
One question, as I don’t have NppExec currently installed,
is it really needed to do SCI_WORDRIGHTEXTEND?

From the docs it states If the current selection is empty then select word around caret

dinkumoil

@Ekopalypse said:

is it really needed to do SCI_WORDRIGHTEXTEND?

According to my tests - yes.

If the SCI_WORDRIGHTEXTEND part is removed from the script, executing it (without an active selection) selects the word right from the cursor and marks all other instances of the word in the whole text but there are no multiple cursors.

Ekopalypse

@dinkumoil - thanks for clarifying.

Alan Kilborn

It automatically selects the word right from the cursor but only if there is no active selection.

I tried the NppExec script and I found that if I run it without a selection but with the caret inside or at the start of a word, it will select from the caret to the right to just before the start of the next word, and then fail to give multiple carets; example with caret sitting on second s of highlighted sci_sendmsg:

Imgur

Starting with an entire word preselected (here the 2nd sci_sendmsg in the file) however works just fine:

Imgur

Perhaps I miss something about the case where there is no selected text first…

dinkumoil

@all

Improved version of my NppExec script:

sci_sendmsg SCI_GETSELECTIONEMPTY

if $(MSG_RESULT) == 1 then
  sci_sendmsg SCI_GETCURRENTPOS
  set local $(CurPos) ~ $(MSG_RESULT)

  sci_sendmsg SCI_WORDSTARTPOSITION $(CurPos) 1

  if $(MSG_RESULT) != $(CurPos) then
    sci_sendmsg SCI_WORDLEFT
  endif

  sci_sendmsg SCI_WORDRIGHTEXTEND
endif

sci_sendmsg SCI_SETSEARCHFLAGS SCFIND_WHOLEWORD 

sci_sendmsg 2690
sci_sendmsg 2689

If the cursor is placed in the middle of a word without an active selection when executing the script, it is moved to the next word boundary to its left before selecting the word to the right of its new position, i.e. the word in whose mid it was placed before.

@Alan-Kilborn
Oh, I’ve posted before I could even see your posting…

Alan Kilborn

@dinkumoil said:

Oh, I’ve posted before I could even see your posting…

I tried the NppExec script just above and had the same result as the earlier NppExec script when no selection before running…

dinkumoil

@Alan-Kilborn said:

I tried the NppExec script just above and had the same result as the earlier NppExec script when no selection before running…

It works at my site. Did you restart Notepad++ after updating the script?

Alan Kilborn

@dinkumoil

Did you restart Notepad++ after updating the script?

No, but I didn’t “save” it, I was just running it as a NppExec temporary script…

Alan Kilborn

Totally just for fun, I managed to replicate the functionality with Pythonscript in its current (1.4) form:

import ctypes
from ctypes.wintypes import HWND
from ctypes import byref, wintypes, Structure, sizeof

def get_focused_window():
    class GUITHREADINFO(Structure):
        _fields_ = [
            ("cbSize", wintypes.DWORD),
            ("flags", wintypes.DWORD),
            ("hwndActive", wintypes.HWND),
            ("hwndFocus", wintypes.HWND),  # <--- what we really want
            ("hwndCapture", wintypes.HWND),
            ("hwndMenuOwner", wintypes.HWND),
            ("hwndMoveSize", wintypes.HWND),
            ("hwndCaret", wintypes.HWND),
            ("rcCaret", wintypes.RECT)
            ]
    guiThreadInfo = GUITHREADINFO(cbSize=sizeof(GUITHREADINFO))
    ctypes.windll.user32.GetGUIThreadInfo(0, byref(guiThreadInfo))
    return guiThreadInfo.hwndFocus

current_view_hwnd = get_focused_window()

SciLexer = ctypes.WinDLL('SciLexer.dll', use_last_error = True)
Scintilla_DirectFunction = SciLexer.Scintilla_DirectFunction
direct_pointer = SendMessage(current_view_hwnd, 2185, 0, 0)

editor.setSearchFlags(FINDOPTION.WHOLEWORD | FINDOPTION.MATCHCASE)

if editor.getSelectionEmpty():
    start_of_word_pos = editor.wordStartPosition(editor.getCurrentPos(), True)
    end_of_word_pos = editor.wordEndPosition(start_of_word_pos, True)
    if start_of_word_pos != end_of_word_pos:
        editor.setSelection(end_of_word_pos, start_of_word_pos)

Scintilla_DirectFunction(direct_pointer, 2689, 1, 0)

dinkumoil

Hmm, I found a strange pitfall with my code.

I use for testing the same snippet of source code @dail used in his posting in the other thread:

static std::string getWordAt(GUI::ScintillaWindow *window, int pos) {
  int word_start = window->Call(SCI_WORDSTARTPOSITION, pos, true);
  int word_end = window->Call(SCI_WORDENDPOSITION, pos, true);
  return getRange(window, word_start, word_end);
}

If I place the cursor before the word pos near the end of line 1 and press my keyboard shortcut everything works fine. If I do the same with placeing it before the word int near the end of line 1, the script selects the word int and the following space character and nothing more happens. It looks like the sci_sendmsg SCI_WORDRIGHTEXTEND gives a wrong result.

I will try to figure out what’s going on…

EDIT: It’s because int is followed by a space character and pos by opening parenthesis. This would make this feature nearly useless. Gonna do further investigations…

Alan Kilborn

@dinkumoil

Yea, if you’ll take note, I had an intentional big amount of spaces to the right of the word I had my caret in for my testing. :)

dinkumoil

@Alan-Kilborn

Does your Python script work as intended or is it a problem with the algorithm used?

Alan Kilborn

@dinkumoil

Not sure if I understand the q… The PS doesn’t use the “word-right-extend” stuff, which seems to be where the trouble lies with the NppExec scripts presented to this point…

I didn’t do exhaustive testing on the PS, but yea, seems to work as intended.