CODEWITHSUNDEEP
csvawkfixed-width
user3204008
user3204008

Reputation: 98

Trouble converting a fixed-width file into a csv

sorry if this is a newbie question, but I didn't find the answer to this particular question on stackoverflow. I have a (very large) fixed-width data file that looks like this: simplefile.txt

ratno      fdate ratname                        typecode country        
12346 31/12/2010 HARTZ                              4    UNITED STATES
12444 31/12/2010 CHRISTIE                           5    UNITED STATES
12527 31/12/2010 HILL AIR                           4    UNITED STATES
15000 31/12/2010 TOKUGAVA  INC.                     5    JAPAN
37700 31/12/2010 HARTLAND                           1    UNITED KINGDOM
37700 31/12/2010 WILDER                             1    UNITED STATES  
18935 31/12/2010 FLOWERS FINAL SERVICES INC         5    UNITED STATES
37700 31/12/2010 MAPLE CORPORATION                  1    CANADA
48614 31/12/2010 SERIAL MGMT  L.P.                  5    UNITED STATES
 1373 31/12/2010 AMORE MGMT GROUP N A               1    UNITED STATES

I am trying to convert it into a csv file using the terminal (the file is too big for Excel) that would look like this:

ratno,fdate,ratname,typecode,country        
12346,31/12/2010,HARTZ,4,UNITED STATES
12444,31/12/2010,CHRISTIE,5,UNITED STATES
12527,31/12/2010,HILL AIR,4,UNITED STATES
15000,31/12/2010,TOKUGAVA  INC.,5,JAPAN
37700,31/12/2010,HARTLAND,1,UNITED KINGDOM
37700,31/12/2010,WILDER,1,UNITED STATES 
18935,31/12/2010,FLOWERS FINAL SERVICES INC,5,UNITED STATES
37700,31/12/2010,MAPLE CORPORATION,1,CANADA
48614,31/12/2010,SERIAL MGMT  L.P.,5,UNITED STATES
 1373,31/12/2010,AMORE MGMT GROUP N A,1,UNITED STATES

I dug a bit around on this site and found a possible solution that relies on the awk shell command: 

awk -v FIELDWIDTHS="5 11 31 9 16" -v OFS=',' '{$1=$1;print}'   "simpletestfile.txt"

However, when I execute the above command in the terminal, it inadvertently also inserts commas in all white spaces, inside the separate words of what is supposed to remain a single field. The result of the above execution is as follows:

ratno,fdate,ratname,typecode,country
12346,31/12/2010,HARTZ,4,UNITED,STATES
12444,31/12/2010,CHRISTIE,5,UNITED,STATES
12527,31/12/2010,HILL,AIR,4,UNITED,STATES
15000,31/12/2010,TOKUGAVA,INC.,5,JAPAN
37700,31/12/2010,HARTLAND,1,UNITED,KINGDOM
37700,31/12/2010,WILDER,1,UNITED,STATES
18935,31/12/2010,FLOWERS,FINAL,SERVICES,INC,5,UNITED,STATES
37700,31/12/2010,MAPLE,CORPORATION,1,CANADA
48614,31/12/2010,SERIAL,MGMT,L.P.,5,UNITED,STATES
1373,31/12/2010,AMORE,MGMT,GROUP,N,A,1,UNITED,STATES

How can I avoid inserting commas in white spaces outside of delineated fieldwidths? Thank you!

Upvotes: 0

Views: 160

Answers (2)

jas
jas

Reputation: 10865

Your attempt was good, but requires gawk (gnu awk) for the FIELDWIDTHS built-in variable. With gawk:

$ gawk -v FIELDWIDTHS="5 11 31 9 16" -v OFS=',' '{$1=$1;print}' file

ratno,      fdate, ratname                       , typecode, country
12346, 31/12/2010, HARTZ                         ,     4   , UNITED STATES
12444, 31/12/2010, CHRISTIE                      ,     5   , UNITED STATES
12527, 31/12/2010, HILL AIR                      ,     4   , UNITED STATES

Assuming you don't want the extra spaces, you can do instead:

$ gawk -v FIELDWIDTHS="5 11 31 9 16" -v OFS=',' '{for (i=1; i<=NF; ++i) gsub(/^ *| *$/, "", $i)}1' file
ratno,fdate,ratname,typecode,country
12346,31/12/2010,HARTZ,4,UNITED STATES
12444,31/12/2010,CHRISTIE,5,UNITED STATES
12527,31/12/2010,HILL AIR,4,UNITED STATES

If you don't have gnu awk, you can achieve the same results with:

$ awk -v fieldwidths="5 11 31 9 16" '
BEGIN { OFS=","; split(fieldwidths, widths) }
{
    rec = $0
    $0 = ""
    start = 1;
    for (i=1; i<=length(widths); ++i) {
        $i = substr(rec, start, widths[i])
        gsub(/^ *| *$/, "", $i)
        start += widths[i]
    }
}1' file

ratno,fdate,ratname,typecode,country
12346,31/12/2010,HARTZ,4,UNITED STATES
12444,31/12/2010,CHRISTIE,5,UNITED STATES
12527,31/12/2010,HILL AIR,4,UNITED STATES

Upvotes: 1

glenn jackman
glenn jackman

Reputation: 247012

perl is handy here:

perl -nE '                                     # read this bottom to top
    say join ",", 
        map {s/^\s+|\s+$//g; $_}               # trim leading/trailing whitespace
        /^(.{5}) (.{10}) (.{30}) (.{8}) (.*)/  # extract the fields
' simplefile.txt 

ratno,fdate,ratname,typecode,country
12346,31/12/2010,HARTZ,4,UNITED STATES
12444,31/12/2010,CHRISTIE,5,UNITED STATES
12527,31/12/2010,HILL AIR,4,UNITED STATES
15000,31/12/2010,TOKUGAVA  INC.,5,JAPAN
37700,31/12/2010,HARTLAND,1,UNITED KINGDOM
37700,31/12/2010,WILDER,1,UNITED STATES
18935,31/12/2010,FLOWERS FINAL SERVICES INC,5,UNITED STATES
37700,31/12/2010,MAPLE CORPORATION,1,CANADA
48614,31/12/2010,SERIAL MGMT  L.P.,5,UNITED STATES
1373,31/12/2010,AMORE MGMT GROUP N A,1,UNITED STATES

Although, for proper CSV, we need to be a bit cautious about fields containing commas or quotes. If I was feeling less secure about the contents of the file, I'd use this map block:

map {s/^\s+|\s+$//g; s/"/""/g; qq("$_")}

which outputs

"ratno","fdate","ratname","typecode","country"
"12346","31/12/2010","HARTZ","4","UNITED STATES"
"12444","31/12/2010","CHRISTIE","5","UNITED STATES"
"12527","31/12/2010","HILL AIR","4","UNITED STATES"
"15000","31/12/2010","TOKUGAVA  INC.","5","JAPAN"
"37700","31/12/2010","HARTLAND","1","UNITED KINGDOM"
"37700","31/12/2010","WILDER","1","UNITED STATES"
"18935","31/12/2010","FLOWERS FINAL SERVICES INC","5","UNITED STATES"
"37700","31/12/2010","MAPLE CORPORATION","1","CANADA"
"48614","31/12/2010","SERIAL MGMT  L.P.","5","UNITED STATES"
"1373","31/12/2010","AMORE MGMT GROUP N A","1","UNITED STATES"

Upvotes: 0

Related Questions