Fix Issue 10662 - byLine!(Char, immutable Char) won't compile

[ntrel]

http://d.puremagic.com/issues/show_bug.cgi?id=10662

This wouldn't compile because readln(buf) now requires a mutable buf.
Also make ByLine.line, ByLine.first_call private.

[ntrel]

I've just realized this will allocate a new string on each call to front.

[monarchdodra]

But if you don't allocate a new string, that basically means you are going to clobber the returned string...

[ntrel]

Yes, that's why I did it, but I think repeated calls to front without calling popFront shouldn't keep making copies. Will fix.

[ntrel]

repeated calls to front without calling popFront shouldn't keep making copies

Now fixed.

[monarchdodra]

Now fixed

Except now, you'll trigger a dupe even for normal ByLine!char. Also, you might as well just call .dup followed by a cast here. Who knows what to! is doing (or not...) behind the scenes?

[ntrel]

No. From http://dlang.org/phobos/std_conv.html#.to :

Converting a value to its own type (useful mostly for generic code) simply returns its argument.

@monarchdodra :

you might as well just call .dup followed by a cast here

I would need a static if. I think to is better.

[monarchdodra]

Hum.... That might warrant an extra comment (IMO). But OK, works for me.

[andralex]

Yah, this is an intended use case for to - just use it and it'll do the right thing.

[monarchdodra]

I'm not sure I agree with this "fix" (I explained why in your bug report). Having the behavior depend on the qualification of the input is weird for me: I think it would warrant a newly named function. If I understand you right, you are trying to do:

char => reusable buffer: OK
immutable(char) => duped before returning buffer: I guess... why not?
const(char) => ???

I'm leaning in favor leaving ByLine "as is" (we can add some constraints to check for mutable types if necessary), and instead introduce a byDupLine function. This function would dup, regardless of type (be it char or immutable char). This would mean having a "standard and not surprising behavior". I also think having a byLine that doesn't clobber your data, be it mutable or not, to be something very good to have (this has been requested for repeatedly).

[ntrel]

Changed to use Unqual, updated unittest, squashed commits.

As mentioned on the bug tracker, I'll probably start a discussion about improving byLine or alternatives to byLine in a few days, but I think this fix is still sensible.

[ntrel]

I'll probably start a discussion about improving byLine or alternatives

See http://forum.dlang.org/post/ksj7b6$86b$1@digitalmars.com

[ntrel]

Closed as monarchdodra and jmdavis are opposed, plus readText.splitter is probably more efficient than this.