Re: Match against multiple patterns problem

Charles Hixson wrote in [#1106215]:
> The case expression might be a bit cleaner, but that would appear to
> coerce using the $~ pattern match value, which I don't like at all

$~ is the same object as your 'm', so you can do your `m[0]' thing with
it, or call `m.to_s' on it, or whatever you like to explicitly coerce
it.  I'll put some examples below.

> The current version is:
>  if  not s.is_a?(String) or s.length < 1 then return  ["1",""]
>  if  (m = /^[^\W_]+/.match(s) )  then  return  ["a", m[0]]
>  if  (m = /^\s+/.match(s) )      then  return  ["s", m[0]]
>  if  (m = /^[[:cntrl:]]+/)       then  return  ["c", m[0]]
>  if  (m = /^[[:punct:]_]+/)      then  return  ["p", m[0]]
>  if  (m = /^[^\w\s[:cntrl:][:punct:]]+/) != nil
>                                  then  return  ["r", m[0]]
> which looks a lot better.

Sure, except that it's syntactically invalid (missing all the 'end's)
and the final 3 if-statements assign the regexp (not the match data) to
m.  Hopefully this is just a copy-paste error; otherwise, decent unit
tests should catch it.

> Also, there would need to be a default case to handle uncovered
> options (which just prints an error message and returns ["2", ""].

Your original (and final) code missed that part.

If you want to not assign m in the condition, you could use $~.  Here's
a straight copy of your final code (with some fixes):

  if  !s.is_a?(String) or s.length < 1 then return ["1",""]; end
  if  /^[^\W_]+/.match(s)        then  return  ["a", $~[0]]; end
  if  /^\s+/.match(s)            then  return  ["s", $~[0]]; end
  if  /^[[:cntrl:]]+/.match(s)   then  return  ["c", $~[0]]; end
  if  /^[[:punct:]_]+/.match(s)  then  return  ["p", $~[0]]; end
  if  /^[^\w\s[:cntrl:][:punct:]]+/.match(s)
                                 then  return  ["r", $~[0]]; end
  puts "error message"
  ["2",""]

Or with a case statement:

  return ['1',''] unless s.is_a?(String) and !s.empty?
  case s
  when /^[^\W_]+/       ; ['a', $~[0]]
  when /^\s+/           ; ['s', $~[0]]
  when /^[[:cntrl:]]+/  ; ['c', $~[0]]
  when /^[[:punct:]_]+/ ; ['p', $~[0]]
  when /^[^\w\s[:cntrl:][:punct:]]+/
                          ['r', $~[0]]
  else  puts "error message"
                          ['2', '']
  end

Obviously you can arrange it however you like, I've just attempted to
match your layout style based on a guess.

Charles Hixson wrote in [#1106217]:
> Given that $1 means what I think it means from context, that would work,
> but I dislike the global variables with pattern matching.

The global assignment happens whether or not you like/use it.
Assertively not using the feature isn't a very strong way to express
your displeasure with it, and it won't just shrivel up and drop off lack
of use.  ;)

See also: https://bugs.ruby-lang.org/issues/8110

> But really my problem was being mislead by error messages about what the
> problem actually was.

That is the main thing.  As a tip, in future you should include the
error message in your post; possibly someone with more experience could 
help you find the root cause more quickly that way.

-- 
Posted via http://www.ruby-forum.com/.