Must a gsub hash key be a string, not a regexp?

I want to do a sequence of gsubs against one string, so I utilized the fact that gsub can take a hash as the second argument. One thing I wanted to do with gsub is to convert a sequence of one or more space/tab into a single space, so I have something essentially as follows:

gsub(/[ \t]+/, {/[ \t]+/ => ' '})

In my actual code, the first argument is a union of the regexp I gave here, and the second argument includes more key-value pairs.

Now开发者_高级运维, when I apply this to a string, all of the space/tabs are deleted. I suppose this is because the match to the first argument is not regarded as matching to the key [ \t] in the second argument (hash). Does the match in the second argument hash only looks for exact string match, not regexp match? If so, is there any way to get around it?

This is a related question. If you need to use the hash because many things have to be substituted, this might work:

list = Hash.new{|h,k|if  /\s+/ =~ k then ' ' else k end}
list['foo'] = 'bar'
list['apple'] = 'banana'

p "appleabc\t  \tabc  apple foo".gsub(/\w+|\W+/,list)
#=> "appleabc abc banana bar"
p list
#=>{"foo"=>"bar", "apple"=>"banana"} no garbage

According to the docs, gsub with a hash as the second parameter only matches against literal strings:

'hello'.gsub(/[eo]/, 'e' => 3, 'o' => '*')    #=> "h3ll*"

If you want to supply multiple hashes you could work around it by creating a hash, where the key/value pairs are the search => replacement pairs, iterate over the hash, and pass those into the gsub. Because Ruby 1.9+ maintains the insertion order of the hash, you're guaranteed that the search will occur in the order you want.

search_hash = {
  '1' => 'one', 
  'too' => 'two', 
  /[\t ]+/ => ' '
}
str = "1, too,\t3 ,    four"
search_hash.each { |n,v| str.gsub!(n, v) }
str #=> "one, two, 3 , four"

If you just want the spaces/tabs replaced with one space, why not just specify that as the replacement, and omit the whole hash?

gsub(/[ \t]+/, ' ')

UPDATE: based on your comment, you can use the block syntax of gsub

gsub(/[ \t]+/) {|match| *do stuff here* }