Off-by-one error when selecting text

[hakonrk]

After double-clicking a word to start a selection, and growing the selection by moving the mouse cursor beyond the end of the initially selected word, the selection grows to one character beyond the mouse cursor. Example:

foo[bar]=1

Let's say you want to select only foo[bar]. You start by double-clicking on foo, thereby highlighting only that word (using the default select_by_word_characters setting). Then you drag the mouse to the closing bracket to also include [bar] in the selection. The result is that the selection also includes the following = character. I don't see any way to select only foo[bar].

[hakonrk]

Just noticed that this could be a duplicate of #945.

[hakonrk]

Tried cherry-picking the four patches in martinetd's selection branch (#1007) onto the master branch. They had no effect on the issue I describe in this ticket.

[martinetd]

What's your select_by_word_characters ? By default it includes =, so the equal sign is just part of the word for kitty.

[hakonrk]

It makes no difference. I tried removing the = from select_by_word_characters and the same thing happens. There are many ways to reproduce this bug. The easiest is using spaces:

foo   bar

Double-click foo, then drag the selection towards `bar. Observe that, while moving over the spaces, the selection is always one character too wide when you hover the mouse over the spaces (or any other non-word character for that matter).

[trygveaa]

I'm also seeing this. I don't think it's a duplicate of #945 because it's not just that it selects the next word too early, but you can also get a selection which doesn't include whole words.

E.g. with the text some words, if you double-click on some and drag the mouse to the right, the selections will be some, then some w and at last some words. The selection some w should never have been active, it should go directly from some to some words.

[hakonrk]

Just wanted to clarify my answer to martinetd's question, quoted below:

What's your select_by_word_characters ? By default it includes =, so the equal sign is just part of the word for kitty.

When I said it did not matter what I put in select_by_word_characters, I was talking about how kitty behaved on the master branch. With the four patches on your selection branch cherry-picked to master, removing the = character from select_by_word_characters makes it possible to select foo[bar] without including the = character immediately following the closing bracket.

[hakonrk]

kovidgoyal, thanks for the patch. It changes the way selection works for this issue, but sadly it still doesn't make it possible to include an arbitrary number of non-word characters in the selection. Here's another example:

matrix[x][y]

Double-click on matrix, then try to grow the selection to matrix[x]. With patch 5f33d90, non-word characters never affect the selection, it does not grow until the mouse cursor reaches the next word character.

[kovidgoyal]

Well, yeah. Why would word selection allow you to select non-word
characters? If you want to select them, either use non-word selection
mode or add the characters you are interested in to the list of
characters consisting of word characters.

[hakonrk]

That's the way it works in xterm, gnome-terminal and konsole (maybe more, didn't bother to check others). It makes sense in my opinion, because when I need to copy a mix of word and non-word characters, I usually want whole words + one or more non-word characters following the word, as shown in my matrix example. Word selection doesn't mean the selection must start and end with word-characters. Also, why change the behavior compared to all(?) other terminal emulators?

[trygveaa]

Thanks for the patch. After thinking about it, I agree with @hakonrk though. It would be more practical to be able to control the inclusion of non-word characters, even though I said in my last comment that the selection should go directly from some to some words, having the selection some (with the space included) between would be nice.

[kovidgoyal]

OK, I dont see an issue with doing that.

[hakonrk]

I have played around with it a little bit, and here's a patch that I've applied on top of 5f33d90. It seems to make kitty do word selection the same as other terminal emulators, but it's barely been tested, and I don't claim to fully understand the code I'm modifying, so it might have some unintended side effects.

diff --git a/kitty/screen.c b/kitty/screen.c
index f5f68570..970c61f0 100644
--- a/kitty/screen.c
+++ b/kitty/screen.c
@@ -2013,7 +2013,10 @@ screen_selection_range_for_word(Screen *self, index_type x, index_type *y1, inde
     Line *line = visual_line_(self, *y1);
     *y2 = *y1;
 #define is_ok(x) (is_word_char((line->cpu_cells[x].ch)) || is_opt_word_char(line->cpu_cells[x].ch))
-    if (!is_ok(x)) return false;
+    if (!is_ok(x)) {
+        *e = x;
+        return true;
+    }
     start = x; end = x;
     while(true) {
         while(start > 0 && is_ok(start - 1)) start--;

[hakonrk]

Should probably also set the start value, *s = x, before returning.

[kovidgoyal]

Umm this is already implemented in master. 9849a69

[hakonrk]

Yes, I noticed that after I had posted it. :) Thanks again for fixing this! 👍