Tips for negation on grep, first, etc.

[colomon]

One of the common use cases of grep is to filter against a negative test -- the classic -v option for command line grep. Negation of tests usually results in more complicated tests, and the negation approaches are different for different sorts of tests. A quick guide to handling this, either in the grep documentation or with a pointer from the grep documentation, would be very handy.

[AlexDaniel]

Great.

That said, it seems that people don't necessarily realize why this construct works, which is actually why we have this ticket in the first place. I think we can provide much more value by explaining why it works. See jnthn's explanation, which may be a bit too short for docs but it is a good starting point anyway.

Another interesting point is that we don't really explain how to do not-grep. For example, .grep({!Int}) is not going to work. Therefore I propose to recommend this idiom also:

say <a b c d e f>.grep(none Int)

[colomon]

Yes, while coke++'s documentation change was exactly what I needed this morning, I think Alex is right on both fronts -- a deeper explanation would be nice and a broader explanation would be nice.

[AlexDaniel]

OK, here's another very important point: https://irclog.perlgeek.de/perl6/2017-07-27#i_14927558

TL;DR none junction is 8 times faster than using a block.

[AlexDaniel]

I've submitted this ticket: https://rt.perl.org/Ticket/Display.html?id=131805

Turns out that the difference in 8 times was due to a regression. Also, this is only relevant when you're grepping lists of Ints (and trying to match them with a regex), which is not very common anyway.

none junction is still ≈1.45 times faster if we add ».Str, but this difference is getting close to being irrelevant.

Soooo… I guess none is still worth a mention, even though performance benefits of using it currently no longer seem to be that huge.