Splitting strings on spaces, unless inside quotes


#1

I want to split a string into words, but group quoted words together
such that…

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

So far I’m drawing a blank on the ‘Ruby way’ to do this and the only
solutions I can think of are turning out to be fairly ugly.

Any advice would be great. Thanks in advance.


#2

On 2006.01.07 09:08, Richard L. wrote:

solutions I can think of are turning out to be fairly ugly.

Any advice would be great. Thanks in advance.

Naively, you can try something like this:

s = ‘foo bar “baz quux” roo’
s.scan(/(?:"")|(?:"(.*[^\])")|(\w+)/).flatten.compact

Elaborate as necessary (add support for single quotes or something).

R.Livsey

E


#3

Richard L. removed_email_address@domain.invalid writes:

I want to split a string into words, but group quoted words together
such that…

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

How about the csv module? Despite the name, you don’t have to use
commas.

require ‘csv’
CSV::parse_line(‘some words “some quoted text” some more words’, ’ ')

I hope this helps,

Tim


#4

On Jan 6, 2006, at 6:08 PM, Richard L. wrote:

only solutions I can think of are turning out to be fairly ugly.

Any advice would be great. Thanks in advance.

I agree that CSV is the way to go, but here’s a direct attempt:

example = %Q{some words “some quoted text” some more words}
=> “some words “some quoted text” some more words”

example.scan(/\s+|\w+|"[^"]*"/).
?> reject { |token| token =~ /^\s+$/ }.
?> map { |token| token.sub(/^"/, “”).sub(/"$/, “”) }
=> [“some”, “words”, “some quoted text”, “some”, “more”, “words”]

Hope that gives you some fresh ideas.

James Edward G. II


#5

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

s = 'some words “some quoted text” some more words

sa = s.split(/"/).collect { |x| x.strip }
(0…sa.size).to_a.zip(sa).collect { |i,x| (i&1).zero? ? x.split : x
}.flatten


#6

(0…sa.size).to_a.zip(sa).collect { |i,x| (i&1).zero? ? x.split : x }.flatten

Just realized that Range responds to zip, so the to_a is unnecessary.

This looks slightly cleaner to me:

(1…sa.size).zip(sa).collect { |i,x| (i&1).zero? ? x : x.split }.flatten


#7

On Jan 7, 2006, at 1:08, Richard L. wrote:

I want to split a string into words, but group quoted words
together such that…

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

Curiously, someone asked exactly that on freenode#perl tonight.

If the input is that simple and is assumed to be well-formed this is
enough:

irb(main):005:0> %q{some words “some quoted text” some “” more
words}.scan(/"[^"]*"|\S+/)
=> [“some”, “words”, ““some quoted text””, “some”, “”"", “more”,
“words”]

Since nothing was said about this, it does not handle escaped quotes,
and it assumes quotes are always balanced, so a field cannot be %q
{"foo}, for example.

– fxn


#8
  • James Edward G. II (removed_email_address@domain.invalid) wrote:

example = %Q{some words “some quoted text” some more words}
=> “some words “some quoted text” some more words”

example.scan(/\s+|\w+|"[^"]*"/).
?> reject { |token| token =~ /^\s+$/ }.
?> map { |token| token.sub(/^"/, “”).sub(/"$/, “”) }
=> [“some”, “words”, “some quoted text”, “some”, “more”, “words”]

impressive

So long

Michael ‘entropie’ Trommer; http://ackro.org

ruby -e “0.upto((a=‘njduspAhnbjm/dpn’).size-1){|x| a[x]-=1}; p
‘mailto:’+a”


#9

Hi –

On Sat, 7 Jan 2006, James Edward G. II wrote:

So far I’m drawing a blank on the ‘Ruby way’ to do this and the only
solutions I can think of are turning out to be fairly ugly.

Any advice would be great. Thanks in advance.

I agree that CSV is the way to go, but here’s a direct attempt:

Me too (end of disclaimer :slight_smile:

example = %Q{some words “some quoted text” some more words}
=> “some words “some quoted text” some more words”

example.scan(/\s+|\w+|"[^"]*"/).
?> reject { |token| token =~ /^\s+$/ }.
?> map { |token| token.sub(/^"/, “”).sub(/"$/, “”) }
=> [“some”, “words”, “some quoted text”, “some”, “more”, “words”]

I think you could do less work:

example.scan(/"[^"]+"|\S+/).map { |word| word.delete(’"’) }

(Or am I overlooking some reason you’d want to capture sequences of
spaces?)

I changed the \w+ to \S+ (and moved it after the | to avoid having it
sponge up too much) in case the words included non-\w characters.

I guess with zero-width positive lookbehind/ahead one could do it
without the map operation.

David


David A. Black
removed_email_address@domain.invalid

“Ruby for Rails”, from Manning Publications, coming April 2006!


#10

Richard L. wrote:

I want to split a string into words, but group quoted words together
such that…

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

Try this:


#11

On Sat, 7 Jan 2006, Tim Heaney wrote:

How about the csv module? Despite the name, you don’t have to use
commas.

require ‘csv’
CSV::parse_line(‘some words “some quoted text” some more words’, ’ ')

I hope this helps,

briliant!

-a


#12

On Jan 6, 2006, at 8:33 PM, removed_email_address@domain.invalid wrote:

(Or am I overlooking some reason you’d want to capture sequences of
spaces?)

I changed the \w+ to \S+ (and moved it after the | to avoid having it
sponge up too much) in case the words included non-\w characters.

You’re right, that’s better all around.

I guess with zero-width positive lookbehind/ahead one could do it
without the map operation.

You can drop the map(), if you’re willing to replace it with two
other calls:

example = %Q{some words “some quoted text” some more words}
=> “some words “some quoted text” some more words”

example.scan(/"([^"]+)"|(\S+)/).flatten.compact
=> [“some”, “words”, “some quoted text”, “some”, “more”, “words”]

James Edward G. II


#13

On Mon, 2006-01-09 at 18:13 +0900, William J. wrote:

s = ‘some words “some quoted text” some more words’
p s.split( / "(.?)" *| / )

Which along with the CSV solution can’t handle complex cases:

s=‘one two" "‘with quotes’ "three "’

s.split( / "(.?)" *| / )
=> [“one”, “two”, " ", “'with”, “quotes’”, "three "]

require ‘csv’
CSV::parse_line(s)
=> []

but Shellwords can:

require ‘shellwords’
Shellwords.shellwords(s)
=> [“one”, “two with quotes”, "three "]


#14

Richard L. wrote:

I want to split a string into words, but group quoted words together
such that…

some words “some quoted text” some more words

would get split up into:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

s = ‘some words “some quoted text” some more words’
p s.split( / "(.?)" *| / )


#15

Geoff Jacobsen wrote:

require ‘csv’
CSV::parse_line(s)
=> []

but Shellwords can:

require ‘shellwords’
Shellwords.shellwords(s)
=> [“one”, “two with quotes”, "three "]

Another option is to use scan instead of split:

‘some words “some quoted text” some more words’.scan
%r{"(?:(?:[^"]|\.)*)"|\S+}
=> [“some”, “words”, ““some quoted text””, “some”, “more”, “words”]

With some additional effort even the quotes can be removed (using
grouping
for example).

r=[];‘some words “some quoted text” some more
words’.scan(%r{"((?:[^"]|\.)*)"|(\S+)}) {|m| r << m.detect {|x|x}};r
=> [“some”, “words”, “some quoted text”, “some”, “more”, “words”]

Kind regards

robert

#16

On Tue, 2006-01-10 at 04:23 +0900, William J. wrote:

[“some”, “words”, “some quoted text”, “some”, “more”, “words”]

s = ‘some words “some quoted text” some more words’
p s.split( / "(.?)" *| / )

Which along with the CSV solution can’t handle complex cases:

s=‘one two" "‘with quotes’ "three "’

s.split( / "(.?)" *| / )
=> [“one”, “two”, " ", “'with”, “quotes’”, "three "]

The shellwords “solution” is a solution to a different problem, not
to this one. It can’t even handle a simple case:

require ‘shellwords’
s = “why can’t you think?”
Shellwords.shellwords(s)

ArgumentError: Unmatched single quote: 't you think?

I agree my example doesn’t match the originators request but I think
there is enough ambiguity about the post to postulate that they may want
more real-world cases such as:

s=‘symbol “William said: “why can’t you think?”” 123 “foo”’
Shellwords.shellwords(s)

=> [“symbol”, “William said: “why can’t you think?””, “123”,
“foo”]

So Shellwords may indeed be a solution to this problem but the problem
is not stated precisely enough to know.


#17

Geoff Jacobsen wrote:

require ‘csv’
CSV::parse_line(s)
=> []

but Shellwords can:

require ‘shellwords’
Shellwords.shellwords(s)
=> [“one”, “two with quotes”, "three "]

This is not a “more complex case”; it is an invalid case.
The original poster simply wanted to avoid splitting on spaces
within double quotes, not within single quotes.

The shellwords “solution” is a solution to a different problem, not
to this one. It can’t even handle a simple case:

require ‘shellwords’
s = “why can’t you think?”
Shellwords.shellwords(s)

ArgumentError: Unmatched single quote: 't you think?