Gnu AWK split by pattern and insert columns , but get more delimiter

Question

I have used same awk file to process two different string. So wired.

cat test.awk
BEGIN{FPAT="([^,])*|(\"[^\"]+\")";OFS=","}{$4="TDP,-1,-1,0,0"OFS$4;print $0}

1. echo "a,b,b,b,b,b,b,b,b,b,b,b,\"a,b\"" | gawk -f test.awk

   a,b,b,TDP,-1,-1,0,0,b,,b,b,b,b,b,b,b,b,"a,b"

2. echo "a,b,,\"a,b\""|gawk -f test.awk

   a,b,,TDP,-1,-1,0,0,"a,b"

Actually,for the first one, there is ",," in the result. But I am expecting the second result in the first one.

Answer 1

[Not really an answer - but too big for a comment]

I think the behavior you're seeing is related to the first atom of your FPAT - which potentially has a zero-length match i.e. ([^,])* matches zero or more non-comma characters - but exactly how it is related eludes me at this point. Consider for example (GNU Awk 4.0.1):

$ echo "a,b,c,d,e,f,g,h,i,j,k,l,\"m,n\"" | 
  gawk '
    BEGIN{FPAT="([^,])*|(\"[^\"]+\")";OFS=","}
    {print $0; $4=$4; print $0; print NF}
  '      a,b,c,d,e,f,g,h,i,j,k,l,"m,n"
a,b,c,d,,e,f,g,h,i,j,k,l,"m,n"
14

whereas if we access the value of NF before the re-assignement

$ echo "a,b,c,d,e,f,g,h,i,j,k,l,\"m,n\"" | 
  gawk '
    BEGIN{FPAT="([^,])*|(\"[^\"]+\")";OFS=","}
    {print $0; print NF; $4=$4; print $0; print NF}
  '
a,b,c,d,e,f,g,h,i,j,k,l,"m,n"
13
a,b,c,d,e,f,g,h,i,j,k,l,"m,n"
13

Regardless, the behavior seems to be unambiguous if you change FPAT to only match non-empty sequences:

$ echo "a,b,c,d,e,f,g,h,i,j,k,l,\"m,n\"" | 
  gawk '
    BEGIN{FPAT="([^,]+)|(\"[^\"]+\")";OFS=","}
    {$4="TDP,-1,-1,0,0" OFS $4; print $0}
  '
a,b,c,TDP,-1,-1,0,0,d,e,f,g,h,i,j,k,l,"m,n"