Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?

Alan Kilborn

I do only Python3 however - is N++ PythonScript still at Python2?

I use ANSI / Unix file endings as default. I suppose I could try using UTF8 files and see if that introduces any encoding issues with the scripts and environments I’m running. PS3 would be essential though for me.

Curious about this.
First, line-endings don’t matter (in this).
Second, definitely try UTF8; the sooner the world can forget about “ANSI” the better off things will be. :-)

Lastly, for “pythonscripting” I don’t see a big difference between Python2 and Python3, syntactically. Sure, if you are doing some “deeper” things, then maybe, but for the vast majority of scripts I’ve done, when I finally move to PS3 (haven’t yet I guess due to laziness) I don’t expect to have to change anything.

So I guess my point is, if you want to stick with ANSI, it seems reasonable to work with the Python2 version of PS, as such an “investment” isn’t “lost” later.

@Ekopalypse may have some further thoughts on this; am I off-base in my thinking, here?

Michael Vincent

@Alan-Kilborn said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

Lastly, for “pythonscripting” I don’t see a big difference between Python2 and Python3, syntactically

Agree, it’s not that hard to port our Py2 to Py3 stuff. What I wasn’t sure about was using PythonScript2 to do language parsing and analysis on Python3 scripts I’ll be writing for work stuff - not to be run as Notepad++ Python scripts. That may not make a difference, but wanted to clarify.

Cheers.

dail

@Michael-Vincent said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

@Ekopalypse
@dail

This old discussion almost going on 2 years - any progress? What were the challenges with NppLSP - technical or just time? Is the “python script based LSP client” still in existence - any improvements / testing?

Cheers.

I obviously haven’t touched the code in quite a while and don’t have any immediate plans to pick it back up any time soon.

The struggles I had at the time were severe lack of documentation. Yes there was documentation, but they covered the basics and individual pieces. Also it came down to just having enough time as well to fight through the documentation.

Specifically what was lacking was the higher level usage-case. What gets sent before other messages, what parameter effects later behavior, when to send these commands, recommended best practices, etc.

On top of this was the problem of transitioning to a newer specification…so some LSP servers supported one, and not the other. Other LSP servers partially supported one of them with certain features not supported, or had broken implementations.

My hope is that over the past 2 years the environment has improved.

Michael Vincent

@dail

Thank you for the feedback!

Ekopalypse

I guess the most critical part of changing from 2 to 3 is how python handles strings and this is the
reason that PS3 is still in alpha because, as far as I understand, this isn’t currently handled in boost::python.

@Michael-Vincent
not sure I understand correctly what you are trying to achieve.
If you want to get auto completion and such stuff to work with python3 scripts in notepad++,
you normally have to use PS3 plugin.
Using PS2 plugin would mean you have to start a python3 process, run this in combination with a
a jedi.py script which then parses your notepad++ .py file and outputs it findings to stdout which then is
parsed back to PS2. Doesn’t sound desirable.

Michael Vincent

@Ekopalypse said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

If you want to get auto completion and such stuff to work with python3 scripts in notepad++,
you normally have to use PS3 plugin.

Yes, that’s my desire. I’m not interested in writing Notepad++ PythonScript automation, I’m interested in using Notepad++ to write scripts in version 3 of the Python language to be used for my work projects and while writing those scripts, I’d like to have some “Language Server” features (i.e., autocomplete, etc) above and beyond the standard Notepad++ “function name completion” through the use of a Language Server “plugin” (or in this case, PS3 adapted script capability you speak of).

I hope that makes sense now in exhaustive detail. I realize the confusion around “PythonScript” and “python scripts” :-)

Cheers.

Ekopalypse

@Michael-Vincent
ok, this is my “lsp” script (called python_ide.py)

''' Basic functionality which one would expect from a full-fledged IDE '''
import sys
from Npp import (
    editor, editor1, editor2, notepad,
    SCINTILLANOTIFICATION, NOTIFICATION, LANGTYPE
    )

from pyflakes import api
from pyflakes import reporter as rep
import jedi
jedi.preload_module(['os',])


class IDE(rep.Reporter):

    def __init__(self):
        rep.Reporter.__init__(self, sys.stdout, sys.stderr)
        self.auto_indent = True
        self.check_types = False
        self.linting = True
        self.reporter = None
        self.is_python = False
        self.STYLE_WARNING = 55
        self.excluded_styles = [1, 3, 4, 6, 7, 12, 16, 17, 18, 19]
        self.prev_completion_list = []


    def initialize(self):
        editor1.styleSetFont(self.STYLE_WARNING, 'Consolas')
        editor1.styleSetItalic(self.STYLE_WARNING,True)
        editor1.styleSetFore(self.STYLE_WARNING, (200,160,12))
        editor1.styleSetBack(self.STYLE_WARNING, notepad.getEditorDefaultBackgroundColor())
        editor2.styleSetFont(self.STYLE_WARNING, 'Consolas')
        editor2.styleSetItalic(self.STYLE_WARNING,True)
        editor2.styleSetFore(self.STYLE_WARNING, (200,160,12))
        editor2.styleSetBack(self.STYLE_WARNING, notepad.getEditorDefaultBackgroundColor())
        editor.callbackSync(self.on_charadded, [SCINTILLANOTIFICATION.CHARADDED])
        editor.callbackSync(self.on_modified, [SCINTILLANOTIFICATION.MODIFIED])
        notepad.callback(self.on_buffer_activated, [NOTIFICATION.BUFFERACTIVATED])
        notepad.callback(self.on_file_saved, [NOTIFICATION.FILESAVED])
        self.is_python = notepad.getCurrentLang() == LANGTYPE.PYTHON


    def on_modified(self, args):
        if args['modificationType'] & 0x100000 == 0x100000 and args['text'] in ['\r','\n', '\r\n']:
            self._indent(args['position'], args['text'])


    def on_buffer_activated(self, args):
        self.is_python = notepad.getCurrentLang() == LANGTYPE.PYTHON
        self.path = notepad.getCurrentFilename()


    def on_file_saved(self, args):
        if self.is_python:
            editor.annotationClearAll()
            self.lint()


    def on_charadded(self, args):
        if self.is_python:
            c = chr(args['ch'])
            if c in '\r\n:-+*/#=)':
                return

            pos = editor.getCurrentPos()
            if editor.getStyleAt(pos) in self.excluded_styles:
                return

            source = editor.getText() + c
            line = editor.lineFromPosition(pos) + 1
            column = editor.getColumn(pos)

            if c == '(':
                script = jedi.Script(code=source).get_signatures(line, column)
                self.show_calltip(script, pos)
            else:
                if c == '.' and editor.callTipActive():
                    editor.callTipCancel()
                script = jedi.Script(code=source).complete(line, column)
                self.autocomplete(script, pos)


    def unexpectedError(self, filename, msg):
        # HACK: to avoid encoding declaration in Unicode string error
        if msg != 'problem decoding source':
            print(f'File "{filename}", line {0}, unexpected error:{msg}')


    def syntaxError(self, filename, msg, lineno, offset, text):
        line = text.splitlines()[-1]
        if offset is not None:
            offset -= (len(text) - len(line)) + 1
        else:
            offset=''

        print(f'File "{filename}", line {lineno},{offset} Error:{msg}')
        print(f'    => {line}')
        self.show_error(filename,lineno, msg)


    def flake(self, message):
        msg = message.message % message.message_args
        for line_ignore_msg in ['# ignore_undefined_name', '# ignore_unused_import',]:
            if line_ignore_msg in editor.getLine(message.lineno-1 if message.lineno>0 else message.lineno):
                return
        self.show_error(message.filename, message.lineno, msg)


    def show_error(self, filename, line_no, msg):
        line_no = line_no-1 if line_no>0 else line_no
        editor.annotationSetText(line_no, f'{msg}')
        editor.annotationSetVisible(2)
        editor.annotationSetStyle(line_no, self.STYLE_WARNING)
        editor.ensureVisible(line_no)
        editor.gotoLine(line_no)
        editor.verticalCentreCaret()


    def lint(self):
        if self.linting:
            filename = notepad.getCurrentFilename()
            if filename.endswith('.py'):
                api.check(editor.getText(), filename, self)


    def autocomplete(self, completions, pos):
        if completions:
            completion_list = sorted([x.name for x in completions if not x.name.startswith('_')])
            if completion_list:
                if self.prev_completion_list != completion_list:
                    self.prev_completion_list = completion_list
                    word_start = editor.wordStartPosition(pos, True)
                    already_typed = pos - word_start
                    already_typed = 0 if already_typed < 0 else already_typed
                    editor.autoCShow(already_typed, ' '.join(completion_list))


    def show_calltip(self, call_tips, pos):
        if call_tips:
            tips = ', '.join([x.name for x in call_tips[0].params])
            tips += '\n'
            if tips.strip():
                editor.callTipShow(pos, f' {tips[:-1]} ')


    def _indent(self, position, text):
        if self.is_python and self.auto_indent:
            indent = editor.getLineIndentation(editor.lineFromPosition(position))
            if (editor.getCharAt(position-1) == 58 and  # 58 == :
                editor.getStyleAt(position-1) not in self.excluded_styles):
                tabwidth = editor.getTabWidth()
                text += ' '*(indent//tabwidth+1)*tabwidth
            else:
                text += ' '*indent
            editor.changeInsertion(text)

and this is what I have in user startup.py

from python_ide import IDE
__ide = IDE()
__ide.initialize()

You need to pip-install pyflakes and jedi and use the PS3 alpha version.
A word of warning, although I don’t think it will affect you.
I had crashes when using this together with Npp objects.
As long as you only use standard Python scripts/modules, as you mentioned,
you should not have this problem.

Michael Vincent

@Ekopalypse said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

You need to pip-install pyflakes and jedi and use the PS3 alpha version.

Sorry for the following probably basic Python questions, but I’m getting this error when running my startup.py:

Traceback (most recent call last):
  File "C:\usr\bin\npp64\plugins\PythonScript\scripts\startup.py", line 31, in <module>
    from python_ide import IDE
  File "C:\usr\bin\npp64\plugins\PythonScript\scripts\python_ide.py", line 8, in <module>
    from pyflakes import api
ModuleNotFoundError: No module named 'pyflakes'

I’m assuming this is because PyFlakes is not installed in my PythonScript3 distro. I did a:

python3 -mpip install pyflakes jedi

and it worked to install to my OS python:

PS VinsWorldcom C:\usr\bin\npp64\plugins\PythonScript > python3 -V
Python 3.6.3
PS VinsWorldcom C:\usr\bin\npp64\plugins\PythonScript > python3
Python 3.6.3 (v3.6.3:2c5fed8, Oct  3 2017, 18:11:49) [MSC v.1900 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import jedi
>>> import pyflakes

So I’m guessing I need to have PythonScript3 find my local Python3 libraries or install PyFlakes and Jedi in the PythonScript3 lib/ directory.

Any tips?

Cheers.

Ekopalypse

@Michael-Vincent

which PS3 plugin did you install? Depending on it you have to use the correct python version and copy
the pythonXX.dll to the PS plugin directory.
Like I have the latest PS, hence I use python 3.8

Ekopalypse

@Michael-Vincent

Now I’m thinking that, if I remember correctly, you have to use python 3.8. PS3 started with this version.

Michael Vincent

@Ekopalypse said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

which PS3 plugin did you install? Depending on it you have to use the correct python version and copy
the pythonXX.dll to the PS plugin directory.

I installed v3.0.4 and have the DLL for python 3.8 (that was bundled with the above ZIP) already in place:

PS VinsWorldcom C:\usr\bin\npp64\plugins\PythonScript > ls -al
total 5680
drwxr-x---  1 VinsWorldcom                0 Nov  5 15:00 .\
drwxr-x---  1 VinsWorldcom                0 Nov  5 15:00 ..\
drwxr-x---  1 VinsWorldcom                0 Sep  7 07:33 doc\
drwxr-x---  1 VinsWorldcom                0 Nov  5 14:27 lib\
-rw-r----a  1 VinsWorldcom          4209224 Jul 20 16:06 python38.dll
-rw-r----a  1 VinsWorldcom          1606144 Sep  7 07:31 PythonScript.dll
drwxr-x---  1 VinsWorldcom                0 Nov  5 14:34 scripts\

Am I to understand that I need to upgrade my OS Python3.6 version to match the 3.8 for PythonScript?

Cheers.

Ekopalypse

@Michael-Vincent

Yes, and copy the pythonXX.dll from the installation to C:\usr\bin\npp64\plugins\PythonScript
then PS sees all your pip installed modules.
In addition, check the prefer installed python libraries from PS configuration dialog.

Michael Vincent

@Ekopalypse

Thanks - I actually got it working by copying PyFlakes, Jedi (and Parso which seemed to be a Jedi dependency) install directories from my system Python3 to the PythonScript/lib folder.

Autocompletion now works for “objects”:

import socket

addrs = socket.getaddrinfo("localhost", 12345, socket.AF_UNSPEC, socket.SOCK_DGRAM)
sock = None

for addr in addrs:
    af, socktype, proto, cname, sa = addr

    try:
        sock = socket.socket(af, socktype, proto)
[...]

and then typing:

sock. brings up a list of all methods I can call on the socket object - but it’s slow - noticeably slow in populating that autocomplete list.

Pretty cool though! I’ll do some more playing with this and thanks for the script and help setting it up.

Cheers.

Ekopalypse

@Michael-Vincent

it should be slow only for the first time it discovers an object. If you use certain objects more often
then others, you could try to preload them jedi.preload_module(['os','sys', ...])

Michael Vincent

@Ekopalypse said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

preload them jedi.preload_module([‘os’,‘sys’, …])

Thanks for the tip! Python is obviously not my strength; hence, the want for some “IDE-lite” help from Notepad++.

Cheers.

Ekopalypse

@Michael-Vincent

:-D - the same would happen to me if I try to write perl :-D

TroshinDV

@pidgeon777 said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

Microsoft Language Server Protocol (LSP) defines the protocol used between an editor or IDE and a language server that provides language features like autocomplete, go to definition, find all references etc.:

У меня есть немного для php, js, vbs, 1C, MySql.
https://github.com/trdm/jn-npp-scripts

Michael Vincent

@Ekopalypse said in Microsoft Language Server Protocol (LSP): A possible revolution for Notepad++?:

ok, this is my “lsp” script (called python_ide.py)

I wonder if this pylspclient can be used? It’s supposedly an LSP client written in Python so theoretically PythonScript could run it.

I’ve been playing with it from the command line with my system Python3 and not having much luck with the examples/python-language-server.py script included and having the python-language-server installed.

Cheers.

Ekopalypse

@Michael-Vincent

let me have a look and play with it a bit.

Ekopalypse

@Michael-Vincent

It looks like the main problem might be that this lspclient is written for Python2.
Let me see if I can make a working version. Probably in 6-7 hours.