Bug 360340

Summary:	autocomplete ignores special characters at beginning of string (problem for coding eg in PHP)
Product:	[Applications] kate	Reporter:	Steve V. <blauesnichts>
Component:	kwrite	Assignee:	KWrite Developers <kwrite-bugs-null>
Status:	RESOLVED INTENTIONAL
Severity:	normal	CC:	christoph, justin.zobel, kredba
Priority:	NOR
Version First Reported In:	20.08.3
Target Milestone:	---
Platform:	openSUSE
OS:	Linux
Latest Commit:	https://commits.kde.org/ktexteditor/8d0f6ff8014cdf32be6410efa184a968ce53525a	Version Fixed/Implemented In:
Sentry Crash Report:

Description Steve V. 2016-03-09 21:47:53 UTC

Autocomplete seems to ignore special charactes at the beginning of a string when making suggestion list. The list includes too many entries, all matching strings no matter if the same special characters are at the beginning or not.

When starting to type a PHP variable "$var" autocomplete suggests everything starting with "var" not taking into account the "$" at the start, same happens when typing "<?php", which is just annoying.

This makes using autocomplete for coding in PHP nearly unusable, because it does not only suggest the different PHP variables already in the file (e.g. $var_1, $var_2) but also any single string value that starts with "var".

Reproducible: Always

Steps to Reproduce:
Test with this in file:

$test_1 = "variable_in_string";

Then type

$var

and get autocompletion suggestion "variable_in_string" which will leave you with

$variable_in_string

although this variable never existed in your code.

Actual Results:  
Suggestion list is longer than it should be for using it in a PHP file.

Expected Results:  
Shorter suggestion list, only including entries that match the full string typed in, including any special characters at the beginning.

Happens for
KWrite
Version 4.14.3
Under KDE 4.14.9
openSuSe 13.2

Did not happen for earlier version of KWrite under openSuSe 12.1

Comment 1 Christoph Cullmann 2019-08-24 20:09:52 UTC

Good point, see

https://phabricator.kde.org/D23410

Comment 2 Christoph Cullmann 2019-08-25 09:52:00 UTC

Git commit 8d0f6ff8014cdf32be6410efa184a968ce53525a by Christoph Cullmann.
Committed on 25/08/2019 at 09:51.
Pushed by cullmann into branch 'master'.

improve word completion to use highlighting to detect word boundaries

Summary:
This allows better completion e.g. for Perl or PHP

Test Plan:
Open some Perl script.

Type:

lala_test = 1

$lala = 2

Now type

$lal

You shall now get only "lala" as proposal

Reviewers: dhaumann

Reviewed By: dhaumann

Subscribers: kwrite-devel, kde-frameworks-devel

Tags: #kate, #frameworks

Differential Revision: https://phabricator.kde.org/D23410

M  +7    -5    src/completion/katewordcompletion.cpp

https://commits.kde.org/ktexteditor/8d0f6ff8014cdf32be6410efa184a968ce53525a

Comment 3 Christoph Cullmann 2020-02-01 20:22:43 UTC

This did lead to other regressions :/

Reverted atm..

See bug 412502

Comment 4 Justin Zobel 2020-10-30 02:35:42 UTC

(In reply to Christoph Cullmann from comment #3)
> This did lead to other regressions :/
> 
> Reverted atm..
> 
> See bug 412502

Was a way ever found to do it without causing regressions?

Comment 5 Justin Zobel 2020-11-21 11:38:17 UTC

Confirmed still an issue in kwrite 20.08.3.

Comment 6 David Kredba 2021-01-16 10:04:58 UTC

Kate version 20.12.1.
It does not suggest words where any of the first characters is a letter with diacritics. For the Czech language words like "Žitná", "Vítězné", and "Dlážděná" are not suggested for autocomplete. If the diacritics it present later in the word it suggests it like in the word "Smrčinská".
Locale used is cs_CZ.UTF-8.
(Examples used are a street names in the CZ.)

Comment 7 Christoph Cullmann 2024-10-20 20:37:05 UTC

I don't see a way to have that without other regressions. I rather have a few more than too few alternatives.
For a more language aware completion we now support LSP.

(In reply to David Kredba from comment #6)
> Kate version 20.12.1.
> It does not suggest words where any of the first characters is a letter with
> diacritics. For the Czech language words like "Žitná", "Vítězné", and
> "Dlážděná" are not suggested for autocomplete. If the diacritics it present
> later in the word it suggests it like in the word "Smrčinská".
> Locale used is cs_CZ.UTF-8.
> (Examples used are a street names in the CZ.)

That works in the current release.