Terminal in Jupyter notebook break line for unicode incorrectly

[yuyichao]

Steps to reproduce

1. make a file txt with content

   啊啊啊啊啊啊啊
   

2. cat this file in the jupyter notebook terminal (availabe at url /terminal/*)

The terminal seems to break the line before the last non-ascii character.

It also seems to be confused about the line number since when I cat this file a couple times, the height of the terminal window is larger than the browser window and the scroll bar appears.

[takluyver]

Again, I think this is an issue with term.js, though hopefully an easier one to fix than the other one you reported. I have checked what gets sent over the websocket, and that is correct '啊啊啊啊啊啊\r\n'.

I'll try to dig into that today. I feel a bit bad just pointing all these issues to term.js without doing anything to help improve it.

[takluyver]

OK, I've worked out more or less what's going on here.

First, even with monospace fonts, 啊 is wider than standard Latin characters, at least on my computer. You can see here if this is the same for you:

abcdef
啊啊啊啊啊啊

Since term.js uses the width of a standard character to calculate the number of columns in the terminal, this means that the row ends up wider than it should be, and it gets wrapped to a second line. Because the rest of the line is non-breaking spaces, and possibly because it's a particle rather than a word character, the browser breaks the line before the last 啊, so it appears to go on the start of the next line.

I can fix this by adding the character to the defined wide characters in the isWide() function. But I assume there are other nearby characters that should be treated similarly, and I don't know where the range starts and ends. Anyone who does know, feel free to make a PR adding that to term.js.

Closing here since I've identified it as a term.js issue.

[takluyver]

Reported as chjj/term.js#66