From: Robert Klemme <shortcutter@...>
Date: 2010-10-28T19:36:35+09:00
Subject: Re: ctype functionality without a gem?

On Thu, Oct 28, 2010 at 12:02 PM, Anton Bangratz <tony@twincode.net> wrote:
> On 10/28/2010 11:16 AM, Ammar Ali wrote:
>> On Thu, Oct 28, 2010 at 11:47 AM, Robert Klemme
>> <shortcutter@googlemail.com> wrote:
>>> On Thu, Oct 28, 2010 at 9:50 AM, Ammar Ali <ammarabuali@gmail.com> wrote:
>>>> Is there a built-in method for identifying character types a la
>>>> ctype() in C? I would like to avoid requiring a gem dependency.
>>
>>> What do you need that for?  Why not directly use a regexp to match a
>>> string?  Often you can use capturing groups in a single regexp, e.g.
>>>
>>> irb(main):009:0> %w{foo bar 123}.each do |s|
>>> irb(main):010:1* if /(\d+)|(\w+)/ =~ s
>>> irb(main):011:2> puts "number" if $1
>>> irb(main):012:2> puts "chars" if $2
>>> irb(main):013:2> end
>>> irb(main):014:1> end
>>> chars
>>> chars
>>> number
>>> => ["foo", "bar", "123"]
>>
>> That's very cool. I'll have to remember that one.
>
> It is a bit brittle, though. Watch:
>
>  %w{foo bar 123 345bar}.each do |s|
>   if /(\d+)|(\w+)/ =~ s
>     puts "number" if $1
>     puts "chars" if $2
>   end
>  end
> #=>
> chars
> chars
> number
> number

Yes, of course.  The regexp was a quick hack only to demonstrate the
mechanism.  You can easily fix that by proper anchoring the regexp.

>> I do actually need to test if a character is of a certain type, all by
>> itself, as part of a parser I'm working on. I came up with this quick
>> solution for now, http://gist.github.com/650968
>
> This is similarily fragile. If you are only testing for one character,
> it seems good, if you are going to test for strings, it will not be
> enough.
>
> Also, the whole solution is not necessarily fast - I do love regular
> expressions, but I would search on for other methods to meet your
> requirements. It really depends on what your goal is. If it is about
> bit streams, you might want to look into bit-struct or similar
> approaches that use Array#pack and String#unpack.

If I have to implement a parser manually I would typically use regexp
for scanning.  You can even make this fairly readable by using /x.

input.scan %r{
  # white space
  (\s+)
  # integer
  ([-+]?\d+)
  # keyword: if
  (if)
  # etc
}x do |m|
  ...
end

Of course, if the number of tokens is large this will become very
awkward.  Better use a proper parser generator then.

Cheers

robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/