Help in finding a regex pattern

[Former Member]

Dear all REGEX Gurus.

We have a comma separated string like this:

132, 143, "222, 144, abc", 227, 888, "222#55#ab"

As you might have guessed, this comes from excel when we save as CSV file, and when one of the columns in excel has the value 222,144,abc. So excel itself puts " at beginning and end.

Now we want to split it based on ',' like this:

132

143

222,144,abc

227

888

222#55#ab

So what we thought is to use a REGEX to find a pattern that has anything beginning with ", ending with ", and has a comma (,) in between. We'd replace that comma with a special character.

System seems to perform greedy search if i search like this ",".

I tried this:

"([^"]+)" -> this works that it gives me all values within quotes. But I want only those which have a comma in them.

So in above, i do not want 222#55#ab to come.

Tried various combinations, but not able to get it work.

Can someone advise please how to achieve it?

Thanks in adv.

[former_member226519]

my first idea:

SPLIT line AT ',' INTO TABLE xyz

LOOP AT XYZ

and when you find a "' at first position -> concatenate with the following lines until you find a " at the end.

ENDLOOP

[Former Member]

Yes, we already tried that.

It has a problem that if there is an opening ", but no closing ", then our search becomes useless.

Moreover it slows down performance as our data can have many rows.

Hence I thought REGEX would be a faster and simpler way.

[Former Member]

Hi,

If " " will give atleast one , then give regex with + .

Well i didn tried it. Just give a try.

Please check the use of +(page 4) in [this link|http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/03a52be5-0901-0010-9da4-e9d5f8c5ce1c?QuickLink=index&overridelayout=true].

[kesavadas_thekkillath]

Hi,

I am in 4.6 version and dont have the facility to regex it for you.

Just wrote a sample code, if useful then use it.

DATA:lv_string(100) TYPE c.
DATA:lv_len TYPE i.
TYPES:BEGIN OF ty,
      field(100) TYPE c,
      END OF ty.
DATA:it TYPE TABLE OF ty.
FIELD-SYMBOLS:<fs> TYPE ty.
lv_string = '132, 143, "222, 144, abc", 227, 888, "222#55#ab"'.
CONDENSE lv_string NO-GAPS.
SPLIT lv_string AT '"' INTO TABLE it.
CLEAR lv_string.
LOOP AT it ASSIGNING <fs>.
  IF <fs> IS INITIAL.
    delete it index sy-tabix.
    CONTINUE.
  ENDIF.
  lv_len = strlen( <fs> ) - 1.
  IF ( <fs>+lv_len(1) CA '"' ) OR ( <fs>+lv_len(1) CA ',' ).
    <fs>+lv_len(1) = ' '.
  ENDIF.
  IF ( <fs>+0(1) CA '"' ) OR ( <fs>+0(1) CA ',' ).
    <fs>+0(1) = ' '.
  ENDIF.
  CONDENSE <fs>.
  WRITE:/ <fs>.
ENDLOOP.

[Former Member]

Thanks Keshav and everyone else.

But I was really looking for a REGEX to achieve the same task as I feel a good REGEX will be much faster than all this logic.

I have currently also written a logic to overcome this problem, but earlier I have used REGEX to find file names and extensions and I feel it is very powerful and wonderful.

Any advices on what regex can make it work for me here? Thanks again in adv