TREX ABAP Client: How to index file content?

Question

Hello Colleagues,
We have installed TREX search engine and writing own solution by using ABAP Client. Everything is ok except files content like XLS, DOC and so on. When we try to post binary content to the FM TREX_EXT_INDEX it is not processed by TREX and only attributes search is available. I think something wrong with data types.
Test example:
  data lt_data type table of char100.
  data l_size type i.
  data lt_doc type trext_index_docs.
  data ls_doc type trexs_index_doc.
  call function 'GUI_UPLOAD'
    exporting
      filename   = 'D:test.xls'
      filetype   = 'BIN'
    importing
      filelength = l_size
    tables
      data_tab   = lt_data.
  ls_doc-doc_key = '00001'.
  ls_doc-doc_langu = 'EN'.
  ls_doc-doc_type = 'B'.
  ls_doc-mime_type = 'application/excel'.
  call function 'SCMS_BINARY_TO_STRING'
    exporting
      input_length = l_size
    importing
      text_buffer  = ls_doc-content
    tables
      binary_tab   = lt_data.
  append ls_doc to lt_doc.
  call function 'TREX_EXT_INDEX'
    exporting
      i_index_id            = me->index_id
      i_rfc_destination     = me->rfc_dest
      i_index_document_list = lt_doc.
Document is indexed without content. Why?

Answer 1

Hi Evgeni,
I realise this is a little late, but just in case you are still interested - or anyone else out there is:
Basically somewhere internally the FM 'TREX_EXT_INDEX' does the following with your ls_doc-content:
l_xstring = p_content_in. "(p_content_in == ls_doc-content)
If you look at the conversion rules in ABAP the xstring then expects the string to contain only the characters '0-9A-F'. Which your string does not have after calling the FM SCMS_BINARY_TO_STRING.
Thus you have to format your ls_doc-content differently. Basically you need to move the hex characters in your xstring into 'real characters'. Which will expand the string massively.
We solved the problem using the following Form:
form conv_content using value(raw_content) type string
               changing content            type string.
  data: lv_char   type c,
        lv_string type string.
  field-symbols:  type x.
  clear content. " init output.
  while raw_content is not initial.
    lv_char = raw_content(1).
    assign lv_char to  casting.
    lv_string = .
    concatenate content lv_string into content in character mode respecting blanks.
    shift raw_content left.
  endwhile.
endform.
Not exactly efficient, but you can call it just as you would call the FM SCMS_BINARY_TO_STRING (except you don't need the filesize). Then the TREX will index your MS-Office documents as long as they are not of type Office 2007 or newer. In that case there is another bug, where the mime_type you pass in the interface is only 50 Chars long - which is too short to fit the full docx mime type for example.
Regards,
Robin