Hackety Hackers Before A Live Studio Audience

Great video! One great thing about it is the guy sounds like Tom Green, which keeps me listening :)

I knew Eric Mill before he was internet-famous. He seems grander now, like a platform diver gracefully hanging in the air above a pool.

Holy Firefox toolbars Batman!

Chris you are too kind. Especially since I am just so danged disorganized in this presentation! I did a better job at RailsConf, I promise.

The ^$().*? noise is madness! How shall the regular expression monster be slain by HH?

Offtopic, would you be a dear and fix the LINK tag (currently href=“/”) at the top of your RSS feed, _why?

I don’t think regexes can or should be slain—but perhaps we can find a better syntax. Brevity in regular expressions is overrated.

Someone want to try writing a library to make this work?

Regexp.build do
  anchor_start
  one_or_more { exactly "foo" }
  exactly "bar"
  capture { zero_or_more { any_char_from "a".."f" } }
end # => /^(?:foo)+bar([a-f]*)/

I think a more readable syntax is what yerejm meant, Mental, not removing them altogether. I really like the block approach. I wonder if you could do simple stuff in a line though, with a special string?

string.match(“starts with ‘gordon’, has any letters or numbers, has one ‘@’, has many letters or numbers, has one ‘.’, ends with many lowercase letters”)
=> string.match(/^gordon[a-zA-Z0-9]*@[a-zA-Z0-9]+\.[a-z]+$/)

This would match emails starting with “gordon”. I think this grammar is actually parsable.

It’s parseable, but I’ve learned to strongly dislike attempts at “natural language” programming because they create false expectations. People get repeatedly drawn to try things which are entirely natural in English but fall outside of the small subset the synthetic language corresponds to, so they’re effectively punished for experimenting.

[In terms of specific experiences, I would have to say the least unsuccessful example of a “natural language” approach is Inform 7, but even there I find it becomes very frustrating to learn until you finally internalize the limits of its pidgin.]

Contrast that with Ruby, where generalizing from examples is likely to lead to pleasant surprises, rather than a smack on the hand and violated expectations. That’s one of the things that makes Ruby so fun in the first place, isn’t it?

HH has a few methods like String#starts? and String#ends? and String#remove that I think help the situation a bit. And there’s String#count, too.

aanand: Is fixed. Thankyou for being so aanandamant.

I’m always amazed that ruby doesn’t have String#starts? and String#ends? methods. I think mental actually has a reasonable idea. It could probably even be a lot more succinct. Maybe I’ll play with something tonight :)

I like MenTaLguY’s example.

But perhaps the exactly operator is redundant?

And could we assign captured stuff to variables directly? Using magic $n variables is succinct but a bit implicit, and especially confusing when there are lots of nested captures to index.

> r = Regexp.build do
>   anchor_start
>   one_or_more "foo"
>   "bar"
>   capture(:named_thing) { zero_or_more { any_char_from "a".."f" } }
> end
=> [...]
>
> named_thing
NameError: undefined local variable or method `named_thing' [...]
>
> "foobarbaz" =~ r
=> 1
>
> named_thing == $1
=> true

Anthony: maybe it could return a subclass of MatchData which would define accessors for the match names.

Something like: capture(:named_thing){…} or capture_many(:other_things){…}.

Which could then be used:
result = built_expression.match(“Some text”)
result.named_thing => some match
result.other_things => array of stuff

It’s a little more explicit, and somewhat less magic.

If you’re trying to assign to a variable, why not use assignment?

named_thing = capture {zero_or_more {any_char_from “a”.."f"}}


And named_thing would be a scalar or an array, depending on whether 1 or more objects were found.

I like netghost’s idea.

By way of contrast, using assignment with the call to capture directly won’t work, since Regexp.build is not evaluating the regexp, but constructing it, which may not even happen in the binding where you need to do the match (in fact, if you’re using this more expensive way to create a regexp, you almost certainly don’t want it to be!). There are also problems if you need to nest the capture.

_why, sorry to bother you on a blog post, but HH forums and other bits and bobs aren’t working for me and I wondered what’s wrong? There isn’t a hint on the other parts of the HH site…

This sadly means some of the later HH leasons just fall down and die, is there something wrong at your end or mine?

Cheers
fd

Got the same problem…Hope you guys can fix it soon.