Multi-cursor editing

Alan Kilborn

So I tried the following with no BetterMultiselection plugin, nor any Pythonscript influence. I’m interested in what you think about it.

I started with this text:

12 56498><717 5123 131
21 78483><000 0456 235
16 89410><575 5156 464
97 89411><056 6045 454
01 25822><644 8970 544

123
456789

12 56498><717 5123 131
21 78483><000 0456 235
16 89410><575 5156 464
97 89411><056 6045 454
01 25822><644 8970 544

and created “a 5 one-line column selection” like you did on the first five lines. Thus the most-recent (i.e., the “main” selection caret) was on the 5th line.

My goal in this testing was to use the down arrow (and only that key) to move my “selection carets” so that they were in the same spot in the lower grouping of five lines as they were in the top set of five lines.

I found I could not do it.

What I obtained was four carets in column 1 and one caret (the original “main” selection one) in between the >< on the last line of the file.

To be fair to N++, I did the same test in the Scintilla demo editor, with the same result.

Maybe my expectation in this scenario is unrealistic?

Ekopalypse

@astrosofista said in Multi-cursor editing:

… almost all the time

Ooopppss - did not happen to me yet.
Is it possible that a very, very large amount of data had to be selected?
I have done a small performance/stress test using this script to generate the test data.

editor.appendText('This is a teest to see if it works\r\n'*100000)

It took about 1 minute, on my PC, until all 100_000 instances were selected.
For a test with 1000 instances, which in my opinion is still not
a real world test - I mean, who has 1000 instances of the same word
in their source code? - it actually selected it immediately.

@guy038 , @Alan-Kilborn

I ASSUME one of it, guess which one, is for the plugin admin,
maybe a naming thing?? But of course, if both are the same then the
non-plugin-admin version could have been deleted.

Python27 should be loadable from within plugin directory.
There was a change in npp source which allows plugins now to find
their needed dlls within its directory.

guy038

Hello, @czmaniak, @alan-kilborn, @peterjones, @ekopalypse, @astrosofista,

Regarding the Python problem, I deleted my Python install and re-installed the plugin with, unfortunately, the same result. My python27.dll library has version 2.7.17150.1013 and contains 2,649,600 bytes. But, never minds : it should be related to my old Win-XP laptop !!

Now, Alan, assuming the text,below :

123
456789

12 56498><717 5123 131
21 78483><000 0456 235
16 89410><575 5156 464
97 89411><056 6045 454
01 25822><644 8970 544

123
456789

12 56498><717 5123 131
21 78483><000 0456 235
16 89410><575 5156 464
97 89411><056 6045 454
01 25822><644 8970 544

123
456789

I did your test, without any plugin, involved :

I placed a caret, between the > and < chars inside the first block, so 5 ones, using the Ctrl key
Then, hitting 9 times on the Down arrow, only, I obtained exactly what you got !
Now, cancel any null selection
Create a normal column selection, between the > and < chars of the first block
Then, hitting 9 times on the Down arrow, only, anunique cursor was present on the 5th line of the second block

Now, enable the BetterMultiSelection plugin
Place a caret, between the > and < chars inside the first block, so 5 ones, using the Ctrl key
If you hit 9 times on the Down arrow, only, this time, the 5 carets are correctly located, between the > and < chars of the second block of five lines :-))
Cancel any null selection
Create a normal column selection, between the > and < chars of the first block
If you hit 9 times on the Down arrow, only, again, the 5 carets are correctly located, between the > and < chars of the second block of five lines :-))

So, the @dail’s BetterMultiSelection plugin is definitively a must and should be installed and enabled every time ;-))

IMPORTANT : To cancel any complicated number of selections, resulting of the use of the BetterMultiSelection plugin, simply hit the ESC key !

Cheers,

guy038

Ekopalypse

@guy038

Ahh, yes - this was one of the reason to drop XP support as this
API is not supported by XP.

astrosofista

@Ekopalypse said in Multi-cursor editing:

Ooopppss - did not happen to me yet.
Is it possible that a very, very large amount of data had to be selected?

No, it wasn’t. It was late and I’m sorry but can’t exactly remember what was the selection to be made, surely the target were partial words. But don’t worry, maybe the issue had other causes. If I can reproduce it will let you know.

Best Regards

astrosofista

@guy038, @alan-kilborn, All

I tried to replicate your tests as long as I understood them, with just Notepad++ basic multiselection feature, that is, disabling the BMS plugin and without running Ekopalypse PS:

Giphy

In the first test I created a null selection column by pressing Shift + Alt + Down Arrow 5 times. As soon as I moved the caret down with an arrow key, the multiselection got lost. End of test.

Second test. Created a null selection column pressing Cntrl + Click 5 times. Moved it down and arrived to line 14 with 5 cursors but misalignated. To resolve this unwanted result hit End and all the 5 cursors aligned at the end lines —or hit Home to get a similar output—. Then went up and got the same as before.

Conclusion: It matters how the multiselection is created. I’m really puzzled.

Best Regards

Ekopalypse

@astrosofista

scintilla knows and uses different selection modes.

By using SHIFT+ALT+ARROWDOWN/UP you are creating a rectangular selection. By using CTRL+mouse_click you are creating the stream mode.

astrosofista

@Ekopalypse

-Thank you. I was unaware of these different modes of selection, considered them as an unique mode, without any difference.

So the selection mode that is created by means of your Python plugin, for example, like in the gif animation you uploaded to this thread two days ago, seems to be a stream mode. Am I right?

Alan Kilborn

@astrosofista said in Multi-cursor editing:

So the selection mode that is created by means of (PythonScript) plugin…seems to be a stream mode.

It’s not explicit in the Python code presented, but most of the P.S. functions that deal with creating selections assume you want stream mode. There are only a few functions that allow you to create column-block selections:

editor.setRectangularSelectionAnchor()
editor.setRectangularSelectionAnchorVirtualSpace()
editor.setRectangularSelectionCaret()
editor.setRectangularSelectionCaretVirtualSpace()

Ekopalypse

@astrosofista said in Multi-cursor editing:

Am I right?

Yes. :-)

astrosofista

@Alan-Kilborn

Good to know, thank you :)

astrosofista

@Ekopalypse

Thank you :)

astrosofista

@astrosofista said in Multi-cursor editing:

@Ekopalypse said in Multi-cursor editing:

Ooopppss - did not happen to me yet.
Is it possible that a very, very large amount of data had to be selected?

No, it wasn’t. It was late and I’m sorry but can’t exactly remember what was the selection to be made, surely the target were partial words. But don’t worry, maybe the issue had other causes. If I can reproduce it will let you know.

Hi @Ekopalypse, All

I think I found the mentioned issue. When a character as a comma (,) is selected —shown at the top of the posted imagen— the Python script works as expected, but things go weird when you select a period or a dot (.) and run the PS —at the bottom of the imagen—, as the multiselection now encompass all the characters of the document:

Hope this helps.

Alan Kilborn

@astrosofista

Apparently the script is treating the . as the regular expression for “any single character”. Thus, when the script is run with only a . character selected, you get a blinking (if so configured) caret at every position in the doc. This script modification corrects it:

from Npp import editor
first_line = editor.getFirstVisibleLine()
editor.setSearchFlags(0)          # <---------- the new line
editor.setTarget(0, editor.getTextLength())
editor.multipleSelectAddEach()
editor.rotateSelection()
editor.setFirstVisibleLine(first_line)

I’m not sure of the complete set of possible arguments to .setSearchFlags() – it is not documented so well in the P.S. docs–but I found these in the Notepad++ source code:

#define SCFIND_NONE 0x0
#define SCFIND_WHOLEWORD 0x2
#define SCFIND_MATCHCASE 0x4
#define SCFIND_WORDSTART 0x00100000
#define SCFIND_REGEXP 0x00200000
#define SCFIND_POSIX 0x00400000
#define SCFIND_CXX11REGEX 0x00800000

My modification to the script is using SCFIND_NONE.

astrosofista

@Alan-Kilborn

Definitively regex, or at least some of its functions are invoked in the original PS. I typed an \s at the end of the first paragraph —I mean the GNU License sample text—, then selected the backslash, run the script, then the s and the script… PS ignored single characters and only take into consideration whole words, so when I selected the first T, PS did nothing, but This was selected and PS correctly matched the other two instances of the word.

Anyway, the script modification you provided solved the issue. A big thanks for your assistance and fast reply.

Best Regards.

Ekopalypse

@astrosofista @Alan-Kilborn

Thanks for testing and improving.
I haven’t noticed it yet but it makes sense.
The documentation for searchFlags is described here.

Alan Kilborn

@Ekopalypse said in Multi-cursor editing:

The documentation for searchFlags is described here.

Yes but that documentation falls flat when you need an actual number to pass to a function!

Is there some Python magic one can do to find these identifiers? What I mean is, if I do dir(MENUCOMMAND) in the P.S. console, I see:

>>> dir(MENUCOMMAND)
['CLEAN_RECENT_FILE_LIST', 'EDIT_AUTOCOMPLETE', 'EDIT_AUTOCOMPLETE_CURRENTFILE', 'EDIT_AUTOCOMPLETE_PATH', 'EDIT_BEGINENDSELECT', ...

Some similar way of finding these “SCFIND_” things?

Ekopalypse

@Alan-Kilborn

for readability you normally use FINDOPTION.NONE instead of the 0.
In order to get all values from such enum you can call its values method.
For example: FINDOPTION.values() would return

{0: Npp.FINDOPTION.NONE, 1048576: Npp.FINDOPTION.WORDSTART, 2: Npp.FINDOPTION.WHOLEWORD, 4: Npp.FINDOPTION.MATCHCASE, 2097152: Npp.FINDOPTION.REGEXP, 8388608: Npp.FINDOPTION.CXX11REGEX, 4194304: Npp.FINDOPTION.POSIX}

Alan Kilborn

@Ekopalypse said in Multi-cursor editing:

for readability you normally use FINDOPTION.NONE instead of the 0.

Yes, but you have to know that EXISTS first! :-)

To your point about “readability”…that’s why I was asking.
I would much rather NOT use the magic number of zero.

I didn’t see anything that logically links FINDOPTION to the SCFIND nomenclature. Maybe I missed the connection.

Alan Kilborn

@Ekopalypse said in Multi-cursor editing:

For example: FINDOPTION.values() would return

Just a slight note for any future readers:

FINDOPTION.values() at the Pythonscript console yields:

>> FINDOPTION.values()
Traceback (most recent call last):
  File "<console>", line 1, in <module>
TypeError: 'dict' object is not callable

The proper thing to do is FINDOPTION.values (no parentheses).