Saving .csv into internal table - using dataset (',' comes between data)

Question

Hi experts,
I need to save .csv from application server to internal table.
i am using the below code.
gt_raw and gwa_raw are dxrawdata format.
OPEN DATASET gv_pfile FOR INPUT IN TEXT MODE ENCODING DEFAULT.
*--- Display error messages if any.
  IF sy-subrc NE 0.
    WRITE:/ 'FILE UPLOAD FAILED - ERROR NO. : ', sy-subrc.
    EXIT.
  ELSE.
    DO.
      READ DATASET gv_pfile INTO gwa_raw.
      IF sy-subrc NE 0.
        EXIT.
      ELSE.
        APPEND gwa_raw TO gt_raw.
        CLEAR gwa_raw.
      ENDIF.
    ENDDO.
*--Close the Application server file (Mandatory).
    CLOSE DATASET gv_pfile.
  ENDIF.
  DELETE DATASET gv_pfile.
  LOOP AT gt_raw into gwa_raw.
    IF SY-TABIX > 1.
      SPLIT gwa_raw at ',' into gwa_cust-cust_code
                                gwa_cust-cust_name
                                gwa_cust-grp_name
      APPEND gwa_cust TO gt_cust.
      CLEAR: gwa_cust, gwa_raw.
    ENDIF.
  ENDLOOP.
My program works fine.
But when the gwa_cust-grp_name contains the value for eg. -> panasonic co., ltd.
it takes till panasonic co., only
and leaves ltd. as i am using SPLIT command.
is there any other way to do this.
plz help me to solve this issue.
thanks.

Answer 1

Hi,
I notice you have marked the message as answered, but I just wanted to let you know there is a solution. The trick is to parse into an internal table and then to find and reassemble fields that were split because they contgain a comma. The ABAP program below is a commented example.
Rgds,
Mark
REPORT  zcsv_parse.
DATA:
  tokens       TYPE i.
TYPES: BEGIN OF ty_result,
  company      TYPE char20,
  compnr       TYPE i,
  city         TYPE char30,
  country      TYPE char30,
END OF ty_result.
DATA:
  gt_rawtab    TYPE TABLE OF string,
  gw_rawtab    LIKE LINE OF gt_rawtab,
  gt_result    TYPE TABLE OF ty_result,
  gw_result    LIKE LINE OF gt_result,
  gt_parse     TYPE TABLE OF string,
  gw_parse     LIKE LINE OF gt_parse.
DEFINE %csvline.
  gw_rawtab = &1.
  append gw_rawtab to gt_rawtab.
END-OF-DEFINITION.
START-OF-SELECTION.
* Create CSV lines, some with a comma inside a token
  %csvline '"CompanyOne NV",500,"Antwerp","Belgium"'.
  %csvline '"CompanyTwo,Inc",600,"New York,NY","USA"'.
  %csvline '"CompanyThree,Ltd",700,"Sydney,NSW","Australia"'.
* Parse the raw CSV
  LOOP AT gt_rawtab INTO gw_rawtab.
    REFRESH gt_parse.
    SPLIT gw_rawtab AT ',' INTO TABLE gt_parse.
    DESCRIBE TABLE gt_parse LINES tokens.
*   If extra commas: token count higher than field count
    IF tokens > 4.
      PERFORM reassemble.
    ENDIF.
*   At this point each entry in GT_PARSE contains exactly
*   one result field => build the result table
    LOOP AT gt_parse INTO gw_parse.
*     Strip quotes from text fields
      REPLACE ALL OCCURRENCES OF '"' IN gw_parse WITH ''.
      CASE sy-tabix.
        WHEN 1. gw_result-company = gw_parse.
        WHEN 2. gw_result-compnr = gw_parse.
        WHEN 3. gw_result-city = gw_parse.
        WHEN 4. gw_result-country = gw_parse.
      ENDCASE.
    ENDLOOP.
    APPEND gw_result TO gt_result.
  ENDLOOP.
* Show the formatted result
  LOOP AT gt_result INTO gw_result.
    WRITE: / gw_result-company, gw_result-compnr,
             gw_result-city, gw_result-country.
  ENDLOOP.
*&      Form  reassemble
*       Merges tokens that were split because they contain a comma
FORM reassemble.
  DATA: lastpos    TYPE i,
        lastchar   TYPE c,
        currtoken  LIKE sy-tabix,
        nexttoken  LIKE sy-tabix,
        gw_next    LIKE gw_parse.
  LOOP AT gt_parse INTO gw_parse.
    lastpos  = STRLEN( gw_parse ) - 1.
    lastchar = gw_parse+lastpos(1).
*   Token starts with quote but does not end with one =>
*   must merge with the next token
    IF gw_parse+0(1) = '"' AND lastchar <> '"'.
      currtoken = sy-tabix.
      nexttoken = sy-tabix + 1.
      READ TABLE gt_parse INTO gw_next INDEX nexttoken.
      CONCATENATE gw_parse gw_next INTO gw_parse SEPARATED BY ','.
      MODIFY gt_parse FROM gw_parse INDEX currtoken.
      DELETE gt_parse INDEX nexttoken.
    ENDIF.
  ENDLOOP.
ENDFORM.                    "reassemble