LF at the End of File

[yandex]

LF at the End of File

[dmikis]

First of all, a whole bunch of unix utils rely upon the fact that every line in a text file ends with newline. Second, some editors inserts last newline automatically, so you may accidentally become author of the last line.
This rule has the very same meaning as not BOM rule, UTF8 as an encoding and prohibition of trailing whitespaces: keep project files in a uniform and most portable way.

[twirl]

What are those UNIX tools?
If we say we need portability, then why LF not CR LF?

[dmikis]

What are those UNIX tools?

Mainly, editors: they just add last newline on saving a file (and you'll become the author of the last line). According to POSIX in Unix-like systems text file is the one with newlines at the end of every line, including last line. Otherwise a program has every right to treat a file as a binary one.

If we say we need portability, then why LF not CR LF?

CR LF is Win-only, whereas LF is for everything else we actually use. Majority wins.

[twirl]

What's the problem with being an author of empty line? And why state this in a codestyle?

[dmikis]

Not empty but the last one, which can be meaningful.

[twirl]

Meaningful to people? How?

[dmikis]

You know you can write code and the last line of a file? :)

[twirl]

I don't understand.

[dmikis]

I don't want to argue about this anymore. There is an implication from POSIX, which may or may not be strictly necessary, and since we mostly use POSIX-compliant tools I suggest this rule continues to live. There is no any reason whatsoever to change this, only nasty little repercussions.

[twirl]

We live without this rule, and everything is fine. Problem is greatly exaggerated, no contemporary tool gives any damn to absence of blank line.

[dmikis]

Again, what about ones adding the last newline?

[twirl]

See no problem with them (== with being an author of blank line).

[dmikis]

It's not just that: you may accidentally commit files you haven't actually changed.

[twirl]

Does this impose any problem?

[dmikis]

I believe it does. SCMs are not only for ability to revert some changes, they're also able to provide sometimes crucial info about a piece of code. Thus checking in unrelated crap is not exactly desired.

[twirl]

git blame and svn blame provide information regarding every single line, so you wouldn't have problem with unrelated crap.

[dmikis]

You do understand that would be informational noise which would not ease an investigation.

[twirl]

Requirement to add LFs to all files makes much more noise.

[dmikis]

No, it's not. One in a lifetime setting to your editor.

[twirl]

Nope. As you mentioned, there are tons of different tools, some of them add LF, and some do not.

[dmikis]

Actually, if you don't use windows, only tools we're using that don't add the last LF by default are WebStorm and SublimeText (these are known to me).

[tarmolov]

I think LF is more convenient because:

1. The majority of our developers use OSX or *nix. Not Windows.
2. Our production code is deployed to servers with unix like systems.

Using LF is a good rule in my opinion.

[twirl]

Please, explain the logic to me then:

1. Codestyle requires to have blank line at the end of the file because some (sic!) tools may require it, and you want to have as much portability as possible.
2. Codestyle requires to use LFs instead of CRLFs, though some tools obviously won't work with LFs. There you somehow don't need portability.
   So, do you need portability or not?

[dmikis]

Yeah, "portability" might be a little stretch. What I meant is portability across POSIX-compatible systems we use.

Never the less, I think I've exhausted my arguments here. If the majority of concerned people is pro removing the rule, let's just do that.

[twirl]

Let me note that other teams could have just opposite situation, i.e. much more windows than osx/linux systems.

I vote for remove this rule from codestyle.

[tarmolov]

my vote: keep the rule in the codestyle

[vtambourine]

Keep the rule.

[dodev]

Windows IDEs and editors can be configured to use POSIX EOL characters and automatically place new line chars at the end of files in a few clicks. @twirl why is this an issue?

P.S. Even git will fire warnings if the file doesn't end with a new line.

[twirl]

Not every Windows tool could be configured to use LFs.

[dodev]

Notepad can't for sure.
What is your suggestion?

[twirl]

My suggestion is: do not state rules which are defined by developer tools stack since different teams use different stacks.

[dodev]

Yep, you're right, but I think EOL in the end of file is a good rule to follow.
We should decide if we should restrict the EOL character to POSIX, make POSIX EOL a recommendation or not state it at all.

[twirl]

We use POSIX systems and we haven't been facing any problems for at least last 3 years.

[dodev]

That doesn't guarantee that you won't in the future.
I think we should change the rule to It is recommended to end files with a line break character. Not specifying whether it should LFCR, CR or LF.

PS we should really discuss what goes in the codestyle - only how to write JS code, source file structure, Web API recommendation etc. I got really confused after all the issues you created.

[tarmolov]

For me personally, it's a good practice to standardize an end of line:

• Easier to work with multiline strings (you have only one char marks an end of line 😄 ).
• Decrease possible commits diffs (developers work with different an end of line preferences can commit unnecessary changes).

I suggest to keep this rule for a while. It's not a big deal to support it.
if other teams are nervous about this rule too, we'll go back to this discussion.

What do you think?

[twirl]

I think the issue is about trailing blank line, not LF vs CRLF (that's just side story).

[dmikis]

My final suggestion is to keep this rule if we OK with unix and cli specific aspect in the codestyle. If we do not want to state this, it's may be worth to remove the rule. Also it applies to no-BOM and LF-only rules.

[tarmolov]

I agree with @dmikis. Keep it.

[twirl]

Okay. So, who makes final decision?

[alt-j]

I think @ikokostya must, because he is primary author of this codestyle.

[twirl]

Hmm. Does @ikokostya act on behalf of all Yandex JavaScript developers?

[vtambourine]

@dmikis, @tarmolov and me voted for keeping the rule. @dodev wasn't so explicit, but he said "I think EOL in the end of file is a good rule to follow", so I will count him on that side :)
@twirl and @zloylos voted for removing the rule.
@alt-j suspends.

So, majority of participants of that discussion supports that rule. This issue can be closed.

[tarmolov]

Don't hurry, please ;)

[qfox]

@twirl I'm not sure why you against this rule.

Pros:

• POSIX tools that requires EOL at EOF (e.g. git, svn) wouldn't warn you and throw an error;
• Explicit (require last EOL) is better than implicit (allow both variants);
• Diffs would contain less noise;
• EOL at EOF also guarantees that every line in file will be ended with EOL that simplifies parsing a little;
• Your variants here.

Cons:

• Need to configure .editorconfig file and install plugin for your IDE (you working in notepad, right?)?

upd I'm working on Windows via putty (nano/mcedit) and in GUI (Sublime, earlier Intype/Idea/Netbeans/FAR) and OS X (Sublime/TextMate/terminal editors). And you really confusing me when saying "Not every Windows tool could be configured to use LFs" — can you tell me which one can't be configured?

[twirl]

POSIX tools that requires EOL at EOF (e.g. git) wouldn't warn you and throw an error;

I've never seen such a warning from git or other tools I use. I should mention that we develop 1M+ lines of code project and never have any problem with EOLs since we stopped appending one file to another with cat and started using specialized building tools four years ago.

Cons:
Need to configure .editorconfig file and install plugin for your IDE (you working in notepad, right?)?

What about "Plurality is not to be posited without necessity" rule?
This stupid rule comes from ancient days. Nobody really tested its necessity.
All the discussion lively reminds me famous experiment on monkeys — where finally nobody knows why really it's prohibited to do something, but still fiercely insists on keeping the rule.

[qfox]

I vote for remove this rule from codestyle.

@zloylos Removing is even worse than an opposite. Here we should decide to use EOL at EOF or don't. Explicitly.

@tarmolov I know that Yandex codestyle is more common than BEM community, but canonical version has this rule and new one is based on Yandex jscs preset that also has this rule. It means that some other guys will vote for this rule (I don't know if it weights anything for ya), just fyi.

[qfox]

I've never seen such a warning from git or other tools I use.

It's not a problem ;-) Now you see:

and never have any problem with EOLs since we stopped appending one file to another with cat and started using specialized building tools four years ago.

I will agree with you when git will stop noticing me about this. I know just 99% of tools can handle this. But it's free to add this line automatically, so no reason to make git sad.

What about "Plurality is not to be posited without necessity" rule?
This stupid rule comes from ancient days. Nobody really tested its necessity.
All the discussion lively reminds me famous experiment on monkeys — where finally nobody knows why really it's prohibited to do something, but still fiercely insists on keeping the rule.

Yep. And nope. See above. Risk of fail vs free adding EOL at EOF by editor — your choice is here.

Personally I really hate this message in git, and moreover I hate diffs that can't be resolved automatically at the EOF without EOL at EOF. That's my practical experience why we must use this rule. And it's in git. I don't even want to say about svn's conflict resolver...

[ikokostya]

This rule helps us keep our sources in same style and removes redundant git diffs. Of course you need setup your editor, but only once.