Generate new file with specific structure from text file

Question

On Linux I have the file orig-file.txt. This file include 4 fields now, but they could be less or more (this file is generated by other application).

What is the best option to translate the orig-file.txt to a file like output-file.txt file (it could be with a shell script or awk, etc.)

orig-file.txt

CREATE_TIMESTAMP              TELEPHONE_NUMBER             ID TYPE
-------------------           -------------------- ---------- -----------------
24-09-2009 16:17:45           33633333333                  20 other_mmm_phone
24-09-2009 17:45:07           33644444444                  20 other_mmm_phone
07-10-2009 10:45:49           12312312312                  20 legacyphone
07-10-2009 11:46:38           59320000043                  20 other_mmm_phone

output-file.txt

CREATE_TIMESTAMP -> 24-09-2009 16:17:45
TELEPHONE_NUMBER -> 33633333333
ID               -> 20
TYPE             -> other_mmm_phone



---



CREATE_TIMESTAMP -> 24-09-2009 16:17:45
TELEPHONE_NUMBER -> 33633333333
ID               -> 20
TYPE             -> other_mmm_phone

---

example from awk lang -( but its not work -:(

# awk 'NR>2 {
 > printf "\
 > %-16s -> %s\n\
 > %-16s -> %s\n\
 > %-16s -> %s\n\
 > %-16s -> %s\
 > \n\n\n---\n\n\n",\
 >         "CREATE_TIMESTAMP", $1" "$2,\
 >         "TELEPHONE_NUMBER", $3,\
 >         "ID", $4,\
 >         "TYPE", $5}\
 > '   orig-file.txt
awk: newline in string near line 2
awk: syntax error near line 3
awk: illegal statement near line 3
awk: newline in string near line 7

Answer 1

Here's some plain ksh:

{
  read t1 t2 t3 t4
  maxlen=$(printf "%s\n" ${#t1} ${#t2} ${#t3} ${#t4} | sort -n | tail -1)
  fmt=$(printf "%%-%ds -> %%s" $maxlen)
  read line
  while read date time tel id type; do
    printf "$fmt\n" $t1 "$date $time" $t2 $tel $t3 $id $t4 $type
    print "\n\n\n---\n\n"
  done
} < orig-file.txt

Update for flexible number of fields:

I substitute the space in the datetime field to make things easier to parse

sed '3,$s/ /@@/' orig-file.txt | 
{
    read line
    set -A headings $line
    max=0
    for head in "${headings[@]}"; do (( max < ${#head} )) && max=${#head}; done
    fmt=$(printf "%%-%ds -> %%s" $max)

    read line

    while read line; do
        set -A fields $line
        i=0
        while (( i < ${#headings[@]} )); do
            printf "$fmt\n" ${headings[$i]} ${fields[$i]} | sed 's/@@/ /'
            (( i=i+1 ))
        done
        print "\n\n\n---\n\n"
    done
}

Answer 2

This does the job in this case. Trivial modification needed if more fields are added.

awk 'NR>2{
    printf "\
%-16s -> %s\n\
%-16s -> %s\n\
%-16s -> %s\n\
%-16s -> %s\
\n\n\n---\n\n\n",\
        "CREATE_TIMESTAMP", $1" "$2,\
        "TELEPHONE_NUMBER", $3,\
        "ID", $4,\
        "TYPE", $5}\
' orig-file.txt > output-file.txt

"CREATE_TIMESTAMP" needs both $1 and $2 since the date itself is whitespace separated.

---

It can be modified to read the field names from the header, but there is the problem with the date being whitespace separated. If other fields also are allowed to contain whitespace, manual modifications will always be needed to compensate such as in the $1" "$2 case.