Re: [bug-gawk] funcsub func needed

[Kjetil Flovild-Midtlie]

Benchmarked on a 4core 1yr old i5 laptop, ca 2300 different regex rules

w gensub (pluss a 2nd pass awk to adjust grp-values ) == 5 min
w awk funcsub (1 pass) == 50 minutes !!!

PROBLEM:
funcsub version is VERY easy to understand and modify
for "dummy" regex personnel but not usable since it could take forever 
to find a bug in output

gensub+2pass is complex and need a hardcore "awky" to mod the 2pass awk 
and keep placeholder symbols in-sync and in correct position (see below 
@%@ and @=@)


# 
--------------------------------------------------------------------------------------------------------------
# sample awk file, 1pass [auto-generated from another awk file + tsv 
datafile]
#
/^data/ { $0 = 
gensub(/\<a\s([[:digit:]])\s([[:alpha:]])/,"/address@hidden@address@hidden@\\2","g") }
/^data/ { $0 = 
gensub(/\<b\s([[:digit:]])\s([[:alpha:]])/,"/address@hidden@address@hidden@\\2","g") }
/^data/ { $0 = 
gensub(/\<c\s([[:digit:]])\s([[:alpha:]])/,"/address@hidden@address@hidden@\\2","g") }
/^data/ { $0 = 
gensub(/\<d\s([[:digit:]])\s([[:alpha:]])/,"/address@hidden@address@hidden@\\2","g") }
/^data/ { $0 = 
gensub(/\<e\s([[:digit:]])\s([[:alpha:]])/,"/address@hidden@address@hidden@\\2","g") }

# ..(2295 more) !!!

{ print $0 }

# end 1pass file 
------------------------------------------------------------------------------------------

# sample awk file 2pass (typically small) 
---------------------------------------------------------
{
    gensub( /@address@hidden/,"/Alp","g")
    gensub( /@address@hidden/,"/Bet","g")
    gensub( /@address@hidden/,"/0","g")
    gensub( /@address@hidden/,"/1","g")

    gensub( /@address@hidden/,"X","g")
    gensub( /@address@hidden/,"IV","g")
    gensub( /@address@hidden/,"I","g")

    # etc
   # often more advanced IF-ELSE is needed !!
   #
    print $0
}
# end 2pass file 
--------------------------------------------------------------------------------------------

awk -f 1pass.awk input.txt | awk -f 2pass.awk > output.txt