Feature #4615

Add IO#split and iterative form of String#split

Added by Joey Zhou over 4 years ago. Updated over 4 years ago.

Assignee:Yukihiro Matsumoto


file.each_line(sep=$/) {|line| ... } can be used to iterate on lines.

But the separator is just a string or nil, it cannot be a regexp.

Sometimes I may want to iterate on "sentences", which are strings separated by (simply say) punctuations ".;!?".

So if I can write it like this:

file.split(/[.;!?]/) {|sentence| ... }

I think it will be very convenient.

You may say I can write it like this:

file.gets(nil).split(/[.;!?]/).each {|sentence| ... }

But this code will: (1) slurp in the whole file; (2) create a temporary array. It the file is a big one, those 2 steps seem both expensive and unnessary.

So I suggest a flexible IO#split: (also available for File and ARGF)

io.split(pattern=$/) {|field|...} -> io # default pattern is $/, not $;
io.split(pattern=$/) -> enumerator # not array

(I think adding a new method is better, rather than modifying the IO#each_line, making it accept regexp as argument.)

Well, String#split has only one form:

str.split(pattern=$;, limit=0) -> array

Maybe add a iterative form, when with a block:

str.split(pattern=$;, limit=0) {|field| ... } -> str

Joey Zhou

Related issues

Related to Ruby trunk - Feature #4780: String#split with a block Assigned 05/25/2011


#1 Updated by Yui NARUSE over 4 years ago

  • Status changed from Open to Assigned
  • Assignee set to Yukihiro Matsumoto



#2 Updated by Yui NARUSE over 4 years ago

Just a thought,

String#split drops a separator.
In this use case, you want to drop the separator?

Anyway on 1.9.2, StringScanner#scan_until seems the one you want.

#3 Updated by Joey Zhou over 4 years ago

Yes, I've made a mistake. The split regexp should be /(?<=[.;!?])/ if I want to iterate on "sentences".

Well, the key points here are: (1)more flexible separator; (2)iterative idiom.

Ruby is just like Perl. $/ in Perl is just a string too. I saw in perldoc that "the value of $/ is a string, not a regex. awk has to be better for something." Maybe awk can set the record separator to a regexp?

So, if there's an IO#split or IO#each can take a regexp as separator, I think it's powerful.

IO and String classes have a few same methods: #each_line #each_char #each_byte, maybe a #split for IO is OK.

and I think (({str.split(pattern) {|filed| ...}})) is a pure Ruby idiom:)

Thank you for telling me that StringScanner has such a method. I'm not familiar with the standard libs. Thank you:)

#4 Updated by Nobuyoshi Nakada over 4 years ago

Use scanf.rb.

#5 Updated by Yukihiro Matsumoto over 4 years ago

  • Status changed from Assigned to Rejected

Use scanf, or read then split. Besides that File#split does not well describe the method's behavior (read then split). It makes me feel it splits the file contents into several files.



#6 Updated by Joey Zhou over 4 years ago

Well, how about (({string.split {|filed| ... }})) ?

#7 Updated by Nobuyoshi Nakada over 4 years ago

It should be another feature.

Also available in: Atom PDF