Edit produces documents with no LF/CRLF after the last line

[DHowett]

I feel like ensuring the presence of a final EOL is something a text editor is supposed to do.

[lhecker]

Windows Terminal for instance has automatically generated project files that don't end in a newline. While it's fine if they're added, how would I edit those files if I don't want to introduce the extra newline this time?

VS Code solves this by making files.insertFinalNewline a setting and having a "Save without Formatting" action. Should we do the same?

[DHowett]

vim doesn't seem to care, and adds one anyway. I don't mind that behavior, I just end up not committing those when I git add -p. That's likely a workaround specific to my workflow though. :)

[Aerocatia]

FWIW, nano's behavior is to ensure newlines by default while also providing a launch argument to disable it (-L --nonewlines).

[Kissaki]

POSIX defines text files to include a final newline. Linux follows POSIX. But Windows and its applications do not, and have traditionally not cared about them.

Linux tools adding final newlines may not be the best comparison in this regard because of that.

Either way, preserving the existing state seems like the most reasonable default.

[rhubarb-geek-nz]

I feel like ensuring the presence of a final EOL is something a text editor is supposed to do.

If you type a newline at the end of the file it should be there, if you delete it should be gone. The editor should not be adding text you did not type.

[DHowett]

If the internal representation of the text within the editor cannot tell the difference between a newline at the end of the file or not then that needs fixing.

Nobody has said this. Please avoid conjecture.

[rhubarb-geek-nz]

Nobody has said this. Please avoid conjecture.

I have had a play and it looks as I would expect, if you type newline at the end it stays, if you remove it, it is gone. It is not compulsory to have a newline at the end of the file.

Often XML and JSON files are generated without a trailing linefeed, I would not expect an editor to add one that was not there, especially if they are stored in GIT and it then counts as a change when you exit.

[lhecker]

This may be something where Windows defaults to not adding them, while NIX does add them by default. Then, once we have settings (#22) you can change it to your liking.

Or, we make the behavior smart by default: If you type text on the last line of a file, a newline gets automatically appended. If you delete the newline explicitly, it won't be readded. (= It only gets added when writing, and only when writing on the last line.)

[Aerocatia]

Second option please. Having edit work one way, then then suddenly work differently because of running it through ssh on something that happens to be Linux does not sound ideal, and I want it to add the newlines even on Windows, because the files being edited are shared across systems.

[rhubarb-geek-nz]

then then suddenly work differently

I don't understand the issue at all. There is no "suddenly work differently", if you want a newline then type a newline. If you don't want it, then don't type a newline. Don't start adding characters that were never typed.

It should be possible on both Windows and POSIX to edit and maintain files both with and without a newline at the end by the simple expedient of typing what you want.

[DHowett]

Don't start adding characters that were never typed.

This is a simple document encoding question.

Just like on POSIX you don't type LF and on Windows you don't type CRLF, for a document to be a valid text document as specified by POSIX you don't type a newline at the end. It's just there.

You may disagree on whether the editor should add it, but it is a category error to say that the user should have to type it or not type it to make their document a valid document, when we can determine that they are editing a document.

[Aerocatia]

There is good reason for it too, the commands cat file1.txt file2.txt >bigfile.txt and echo Add a new line to my text file please >>somefile.txt are only going to consistently do what you would expect them to do if everything is terminated with a newline.

[rhubarb-geek-nz]

I suggest that on Windows it has the same behaviour as edit.com, notepad.exe and does not add what you did not type.

If the expectation on POSIX systems is that it behaves like nano and vim to provide least astonishment then you can do that.

So on Windows it works how Windows users expect, and on POSIX it works how POSIX users expect.

However there are use cases where that is not the desired behaviour, when dealing well-formed XML documents, or JSON files that do not contain redundant whitespace for example.

are only going to consistently do what you would expect them to do if everything is terminated with a newline.

It should be about mechanism, not policy, I can choose what characters are appropriate in my files and I can quite happily have one half of a sentence in one file and another half in another and use cat to join them for further processing.

[lhecker]

After thinking about it for a while, I'm likely going to go with the second option here: #41 (comment)
...if that approach is disliked by a significant amount of people, I think I'll consider different alternatives.

Reasons for this:

• A very convincing argument: Edit produces documents with no LF/CRLF after the last line #41 (comment)
• You can change and save your document, and the newline won't be added
• ...it'll only be added if you add (!) a character at the end of the document = it's rarely intrusive