I suggest building your strings using Transform() to make sure all fields are first converted to text. The following worked for me:

lcString1 = 'count(pws_id + transform(' + field(3) + ') ;
          + transform(' + field(4) + ') + transform(' ;
          + field(6) + '))'
lcString2 = 'pws_id, ' + field(3) + ', ' + field(4) + ', ' + field(6)
select pws_id, &lcString1 kount ;
   from r6mile ;
   group by &lcString2 ;
   having kount > 1

For the test purposes, I used the Field() function to return the name of the field in the currently selected workarea so I would not have to manually type in the names of the fields.

>Hi everyone,
>
>I have an application called De-Duper, which identifies duplicates in a table based on one field or combination of few fields. Once duplicates are identified, user can review them, combine, delete, edit, etc.
>
>My colleague, who was testing this application, made a suggestion: this application should identify complete duplicates (e.g. almost all fields should be identical), then delete all duplicate records except one.
>
>I have one solution in mind, but not sure, it's the best. Besides, it would not be Rushmore optimizable, so may take a long time.
>
>Ok, here is what I have in mind:
>
>Table has a RecNum Identifier field.
>
>1) Create a list of all fields, which should be checked (comma-delimited) in a string. This part is already done.
>
>2)

lcStr='select *, count(*) as cnt_dups from BldMstr group by '+m.lcFieldList+ ;
>    ' having cnt_dups>1 into cursor curCompleteDups'
>    strtofile(m.lcStr, m.lcTempPrg)
>    compile (m.lcTempPrg)
>    do (m.lcTempPrg)
>    delete (m.lcTempPrg)
>
>3)  create similar string
>    m.lcDelStr='field1=curCompleteDups.field1 and field2=curCompleteDups.field2"...
>    lcStr='delete for '+m.lcDelStr+' and RecNum<>curCompleteDups.RecNum'
>    create a temp program and execute it.

>
> Do you think, it's feassible? Or can you suggest a better approach?
>
> Thanks in advance.