Why Is My Contenteditable Cursor Jumping to the End in Chrome?

问题

Scenario

I have a contenteditable <div> area, and within this area I may have some <span contenteditable="false"></span> containing some text. The idea is, these span elements will represent styled text that can not be edited, but may be deleted from the <div contenteditable="true"></div> area by pressing the backspace key.

Issue

The cursor placement is the big issue here. if you delete one of these <span> elements, the cursor jumps to the end of the <div>. More interesting, if you type some text while the cursor is "at the end," the text placement is just fine... Then, if you delete the newly typed text, the cursor jumps back!

I have prepared a fiddle which will demonstrate this. I need this to work only in Chrome, and other browsers are either of non-concern for now or have workarounds in place. (Also note the prepared Fiddle is crafted to demonstrate this in Chrome only).

Fiddle

Fiddle Instruction: Chrome Version 39.0.2171.95 m (64-bit) reproduced in 32-bit as well

Click into <div> area
Type "123"
Backspace "3" Backspace "2" Backspace "1"

Related Details

Researching this extensively, I have come across various SO question that are similar, but borrowing the associated solutions has not proved to be the silver bullet I am after. I have also found issues for the Chrome project which seem to target (perhaps not in the exact manner) the issue described above, and can be viewed below.

Issue 384357: Caret position inside contenteditable region with uneditable nodes
Issue 385003: Insert caret style is wrong when reaching the end of a line in a contenteditable element
Issue 71598: Caret in wrong position after non-editable element at the end of contentEditable

The closest SO solution I have found can be here. The idea in this solution is to place &zwnj; characters after the <span> elements, but if I want to now delete my <span>, I instead delete the &zwnj;... forcing my cursor to jump to the end, offering a weird UI experience by not deleting my <span> on my "initial delete key stroke."

Question

Has anyone experienced this issue and found a work around? I welcome any possible solution, even as JS hacky as they come. I've come to learn that leveraging contenteditable comes with a laundry list of struggle, but this seems to be the last remaining difficulty I currently have.

回答1:

I don't know why this happens, but I had the feeling it has something to do with the sizing of the <div>, so I tried playing with the display property. After setting it to inline-block and playing a little with the text I found that the issue is gone after I make some edits to it, specifically adding a new line.

I saw that, for some reason, the <br/> tag is kept in div.main after I delete my new line, but the appearance of the <div> and the way it responds to arrow keys is the same as if there is no new line in it.

So I restarted the fiddle with the CSS change and a <br/> tag in div.main and viola!

So to conclude:

Add display: inline-block to div.main
add a <br/> at the end of div.main

JSFiddle Link

回答2:

The problem is that your contenteditable element is a div, which by default is display: block. This is what causes your caret position problem. This can be fixed by making the outermost div non-editable and adding a new editable div that will be treated as inline-block.

The new HTML would have a new div just inside the outer one (and the corresponding closing tag at the end):

<div id="main" class="main"><div id="editable" contenteditable="true">...

And add this to your CSS:

div#editable {
    display: inline-block;
}

For the sake of seeing the caret better when it is between span elements, I've also added margin: 2px to the rule for div.main span in the CSS but this is not necessary to prevent the caret jumping issue reported in the question.

Here is a fiddle.

As you've started discovering, contenteditable is handled inconsistently across browsers. A few years back, I started working on a project (in-browser XML editor) where I thought contenteditable would make everything easier. Then as I developed the application, I soon found myself taking over the functions that I thought contenteditable would give me for free. Today, the only thing contenteditable give me is that it turns on keyboard events on elements I want to edit. Everything else, including caret movement and caret display, is managed by custom code.

回答3:

So, I've got a slightly cleaner implementation of your approach, which relies on the input event that is triggered when a contenteditable field is updated. This event is not supported by IE, but it looks like contenteditable is pretty wonky in IE anyways.

It basically injects elements around every element marked as contenteditable="false". It could probably be done on initial load rather than just on the input event, but I figured you might have ways of injecting more span's to the region, so input should account for this.

Limitations include some weird behavior surrounding the arrow keys, as you've seen yourself. Mostly what I've seen is that the cursor maintains its proper position when you move backward through the spans, but not when you move forward.

JSFiddle Link

Javascript

$('#main').on('input', function() {
    var first = true;
    $(this).contents().each(function(){
        if ($(this).is('[contenteditable=false]')) {
            $("<i class='wrapper'/>").insertAfter(this);
            if (first) {
                $("<i class='wrapper'/>").insertBefore(this);
                first = false   
            }
        }
    });
}).trigger('input');

CSS

div.main {
    width: 400px;
    height: 250px;
    border: solid 1px black;
}

div.main span {
    width:40px;
    background-color: red;
    border-radius: 5px;
    color: white;
    cursor: pointer;
}

i.wrapper {
    font-style: normal;
}

回答4:

I have found a hacky way to solve the initial problem of the cursor misbehaving at the end of the <div> with some JQuery. I am basically inserting an empty <i> for the cursor to "latch" on to at the end of the <div> contents. This tag works for me because there is no UI difference to the end user when backspacing these from normal text due to overriding font-style: normal (see fiddle), and I also needed something different than <span> to insert

I am not particularly thrilled with this workaround, especially choosing an <i> just because, but I have not come across a better alternative, and I still welcome better solutions- if any exist!. I plan to keep working at this to hopefully figure out the arrow keys as well, but luckily I am only bothered by this glitch in the Fiddle and not my code.

Fiddle (Chrome only)

<div id="main" contenteditable="true">
    <span contenteditable="false">item</span>
    <span contenteditable="false">item</span>
    <span contenteditable="false">item</span>
</div>

function hackCursor() {    
    return $('#main :last-child').get(0).tagName === 'SPAN' ? 
        $('#main span:last-child').after('<i style="font-style: normal"></i>') : 
        false;
}

$(function(){

    hackCursor();

    $('#main').bind({
        'keyup': function(e) {
            hackCursor();
        }
    })
});

回答5:

Just add a new line char (\n) before closing the contenteditable tag. For example in JavaScript: var html = '<div id="main" contenteditable="true">' + content + "\n" + '</div>'

来源：https://stackoverflow.com/questions/27786048/why-is-my-contenteditable-cursor-jumping-to-the-end-in-chrome

标签

javascript

html

google-chrome

contenteditable