where is grep

where-is-grep.md

One of Larry's purported factors in the design of Camelia (and by extension, Perl 6) was to appeal to 7 year old girls (and by extension, people new to programming in general).

In addition, another design principle of Perl6 was to make it "less Unix centric". Whereas Perl had s/// and tr///, Perl 6 retains these but adds subst and trans, which is friendlier to new programmers, and self-documenting to some degree.

It seems odd, then, that we still have the grep operator named so. Even in Perl, grep - as in globally search a regular expression and print - bears little resemblance to what it actually does. It's more of a filter, filtering elements out that do not evaluate True to a given expression.

What then would be it's alternative friendly name? Taking into account these design principles...

• Context Sensitive: A term or symbol might do different things in different context
• Maximum reuse: Syntax and concepts are reused wherever possible to minimize the stuff you have to remember

I propose where.

This make a lot of sense when you consider how similar a grep expression is to the context that where is currently used (in defining a subset).

Here's a demonstration to highlight the similarities

> (^10).grep(* %% 2) 
(0 2 4 6 8)
> subset Even of Int where * %% 2;
> (^10).grep(Even)
(0 2 4 6 8)
> my Int $even where * %% 2
> (^10).grep($even)
(0 2 4 6 8)
> 

Here it becomes very clear. The * %% 2 expression appears directly after both grep and where. Contextually they are similar. We can also give a subset to grep and it does as expected. Conceptually they are almost identical.

Here's some hypothetical code that uses a newly named where method that aliases grep.

my @even = @numbers.where(* %% 2)
my @prod = @servers.where(*.status eq 'Production')
my @matches = @data.where(/pattern/)

Please note, I'm not advocating for the removal of grep at all... I'm proposing an additional nickname for it. One more relevant, logical, and friendly to new (all?) programmers, including 7 year old girls.

Update - 2016-03-31 : Module!

This is now a (kind of) fully fledged module in the ecosystem.

https://github.com/0racle/p6-wig

Update - 2016-04-04 : Prior Art

Some prior art on the use of the word where to filter a list

Powershell

$evens = 0..10 | Where { $_ % 2 -eq 0 }

C#

int[] evens = array.Where(i => (i % 2) == 0).ToArray();

SQL

SELECT v INTO #evens FROM #nos WHERE v % 2 = 0

Haskell

import Data.Array
ary = listArray (1,10) [1..10]
evens = listArray (1,n) l where
  n = length l
  l = [x | x <- elems ary, even x]

Dart

numbers.where((i) => i.isEven);

Interesting to note, when looking at Filter - Rosetta Code, the only examples that have a grep operator are UNIX shell, Perl, and Sidef ( a language implemented in perl5 ).

Oh, there's a gist!

I'll quote myself:

<AlexDaniel> I'd love to see some alternative to “grep”. Just so that I don't have to explain what grep is to non-unixers

Meaning that “.where” or even “.filter” is much better for teaching purposes. That being said, I don't like that the word “where” will have two different meanings (and I don't really buy the idea that these are somewhat similar). In perl, different things are written differently, so I'd be happy to see it this way. But “filter” is not any better at all.

Any other alternatives? What about “grab”?

“keep” was also suggested: http://irclog.perlgeek.de/p6dev/2016-04-03#i_12279019

This idea of this in core is over for now. TimToady ruled against it...

TimToady | I really don't think we're changing anything here; 'where' in particular has
         | no current auto-iterative meaning, it's only used where something external is
         | looking for a boolean smartmatch, so that would be confusing to overload

I still like my suggestion, and at the least, it started a discussion which showed there is a desire for an alternate name. However it seems to have caused more fallout than I was prepared for.

I just wanted to put down a few remaining thoughts...

I updated the top post with some prior art, just to show that the use of where in this context is not unheard of.

To those who also liked my suggestion, I say thanks for making me feel a little less weird. Specifically, I want to thank lizmat for her support and kind words.

Thanks also to Larry for Perl. I was a late comer to perl5 (only a couple years ago) and I love it dearly.

Without a love of perl5 I might have been one of those people who dismissed Perl 6 off-hand, and subsequently would have missed out on what an utterly enjoyable language it is to use. Thanks to all those working on Perl 6 and it's ecosystem. I am exited for the future of Perl 6.

I hope the augment/precomp issues can be ironed out soon so that ideas like this can live better lives in the module ecosystem.

In the meantime, I still have a version in rakudobrew with this this idea built into core that I can fire up occasionally and reminisce about that time I almost changed Perl 6 fundamentally {[: )

filter has become a standard of sorts, but it is a misnomer because one does not know if the boolean should apply to what you're keeping or what you're filtering out.

my @nums = (^10).filter(* %% 2)

Am I keeping even numbers or rejecting them? This is the natural English thought, to me at least.

my @nums = (^10).grep(* %% 2)

grep has the same issue as filter for me. Compounded by the fact that it isn't even a recognizable English word.

---

For this reason, I like keep and reject.

my @nums = (^10).keep(* %% 2)

There is no doubt that @nums contain even numbers.

my @nums = (^10).reject(* %% 2)

Again, there is no doubt that @nums contain odd numbers.

---

Having said that, Perl6's gather/take functionality is fairly similar and quite clear. It's just a very new system to a lot of people.

my @nums = gather for ^10 { take if $_ %% 2 }