I have a large file, about 10GB. I have a vector of line numbers which I would like to use to split the file. Ideally I would like to accomplish this using command-line utilities. As a regex:
File:
1 2 3
4 5 6
7 8 9
10 11 12
13 14 15
16 17 18
Vector of line numbers:
2 5
Desired output:
File 1:
1 2 3
File 2:
4 5 6
7 8 9
10 11 12
File 3:
13 14 15
16 17 18
Using awk:
$ awk -v v="2 5" ' # space-separated vector if indexes
BEGIN {
n=split(v,t) # reshape vector to a hash
for(i=1;i<=n;i++)
a[t[i]]
i=1 # filename index
}
{
if(NR in a) { # file record counter in the vector
close("file" i) # close previous file
i++ # increase filename index
}
print > ("file" i) # output to file
}' file
Sample output:
$ cat file2
4 5 6
7 8 9
10 11 12
Very slightly different from James's and kvantour's solutions: passing the vector to awk as a "file"
vec="2 5"
awk '
NR == FNR {nr[$1]; next}
FNR == 1 {filenum = 1; f = FILENAME "." filenum}
FNR in nr {
close(f)
f = FILENAME "." ++filenum
}
{print > f}
' <(printf "%s\n" $vec) file
$ ls -l file file.*
-rw-r--r-- 1 glenn glenn 48 Jul 17 10:02 file
-rw-r--r-- 1 glenn glenn 7 Jul 17 10:09 file.1
-rw-r--r-- 1 glenn glenn 23 Jul 17 10:09 file.2
-rw-r--r-- 1 glenn glenn 18 Jul 17 10:09 file.3
This might work for you:
csplit -z file 2 5
or if you want regexp:
csplit -z file /2/ /5/
With the default values, the output files will be named xxnn where nn starts at 00 and is incremented by 1.
N.B. The -z option prevents empty elided files.
Here is a little awk that does the trick for you:
awk -v v="2 5" 'BEGIN{v=" 1 "v" "}
index(v," "FNR" ") { close(f); f=FILENAME "." (++i) }
{ print > f }' file
This will create files of the form: file.1, file.2, file.3, ...
Ok, I've gone totally mental this morning, and I came up with a Sed program (with functions, loops, and all) to generate a Sed script to make what you want.
Usage:
put the script in a file (e.g. make.sed) and chmod +x it;
then use it as the script for this Sed command sed "$(./make.sed <<< '1 4')" inputfile¹
Note that ./make.sed <<< '1 4' generates the following sed script:
1,1{w file.1
be};1,4{w file.2
be};1,${w file.3
be};:e
¹ Unfortunately I misread the question, so my script works taking the line number of the last line of each block that you want to write to file, so your 2 5 has to be changed to 1 4 to be fed to my script.
#!/usr/bin/env -S sed -Ef
###########################################################
# Main
# make a template sed script, in which we only have to increase
# the number of each numbered output file, each of which is marked
# with a trailing \x0
b makeSkeletonAndMarkNumbers
:skeletonMade
# try putting a stencil on the rightmost digit of the first marked number on
# the line and loop, otherwise exit
b stencilLeastDigitOfNextMarkedNumber
:didStencilLeastDigitOfNextMarkedNumber?
t nextNumberStenciled
b exit
# continue processing next number by adding 1
:nextNumberStenciled
b numberAdd1
:numberAdded1
# try putting a stencil on the rightmost digit of the next marked number on
# the line and loop, otherwise we're done with the first marked number, we can
# clean its marker, and we can loop
b stencilNextNumber
:didStencilNextNumber?
t nextNumberStenciled
b removeStencilAndFirstMarker
:removeStencilAndFirstMarkerDone
b stencilLeastDigitOfNextMarkedNumber
###########################################################
# puts a \n on each side of the first digit marked on the right by \x0
:stencilLeastDigitOfNextMarkedNumber
tr
:r
s/([0-9])\x0;/\n\1\n\x0;/1
b didStencilLeastDigitOfNextMarkedNumber?
###########################################################
# makes desired sed script skeleton from space-separated numbers
:makeSkeletonAndMarkNumbers
s/$/ $/
s/([1-9]+|\$) +?/1,\1{w file.0\x0;be};/g
s/$/:e/
b skeletonMade
###########################################################
# moves the stencil to the next number followed by \x0
:stencilNextNumber
trr
:rr
s/\n(.)\n([^\x0]*\x0[^\x0]+)([0-9])\x0/\1\2\n\3\n\x0/
b didStencilNextNumber?
###########################################################
# +1 with carry to last digit on the line enclosed in between two \n characters
:numberAdd1
#i\
#\nprima della somma:
#l
:digitPlus1
h
s/.*\n([0-9])\n.*/\1/
y/0123456789/1234567890/
G
s/(.)\n(.*)\n.\n/\2\n\1\n/
trrr
:rrr
/[0-9]\n0\n/s/(.)\n0\n/\n\1\n0/
t digitPlus1
# the following line can be problematic for lines starting with number
/[^0-9]\n0\n/s/(.)\n0\n/\n\1\n10/
b numberAdded1
###########################################################
# remove stencil and first marker on line
:removeStencilAndFirstMarker
s/\n(.)\n/\1/
s/\x0//
b removeStencilAndFirstMarkerDone
###########################################################
:exit
# a bit of post processing the `w` command has to be followed
# by the filename, then by a newline, so we change the appropriate `;`s to `\n`.
s/(\{[^;]+);/\1\n/g
I have several file names, for ease I've put them in a file as follows:
01.action1.txt
04action2.txt
12.action6.txt
2.action3.txt
020.action9.txt
10action4.txt
15action7.txt
021action10.txt
11.action5.txt
18.action8.txt
As you can see the formats aren't consistent what I'm trying to do is extract the first numbers from these file names 1,4,12,2,20 etc
I have the following regex
(\.)?action\d{1,}.txt
Which is successfully matching .action[number].txt but I need to also match the leading 0 and apply it to my substitute with blank in sed so i'm only left with the leading numbers. I'm having trouble matching the leading 0 and applying the whole thing to sed.
Thanks
With GNU sed:
sed -r 's/0*([0-9]*).*/\1/' file
Output:
1
4
12
2
20
10
15
21
11
18
See: The Stack Overflow Regular Expressions FAQ
I don't know if the below awk is helpful but it works as well:
awk '{print $1 + 0}' file
1
4
12
2
20
10
15
21
11
18
I am trying to split a file with data like
2 0.2345
58 0.3608
59 0.3504
60 0.4175
65 0.3995
66 0.3972
67 0.4411
411 0.3455
2 1.3867
3 1.4532
4 1.2925
5 1.2473
6 1.2605
7 1.2463
8 1.1667
9 1.1312
10 1.1502
11 1.1190
12 1.0346
13 1.0291
409 0.8025
410 0.8695
411 0.9154
For this kind of data, I am trying to split this into two files:
File 1 : 2 -411 (first Column match)
File 2 : 2-411 (second occurrence in the first column)
For this, I wrote these two one liners:
awk '1;/411/{exit}' $1 > File1_$1 ;
awk '/411/,0' $1 | awk '{if (NR!=1) {print}}' > File2_$1
The problem is that if there is a match of "411" (as in "67 0.4411") on the second column, my script prematurely cuts from that line.
I am unable to make the match on the first column only as occurrence of 411 on the second column can be number of times and not of interest.
Any help would be greatly appreciated.
an idea could be to use this command combination
awk '{ if ($1 >= 2 && $1 <= 411) print $0 }{if ($1=="411") exit}' input > f1
then
grep -v -f f1 input > f2
if your input file is more bigger you should repeat step2.
I don't know nothing about Bash, but for regex i think you should indicate that the line begins with 411 like that \b411.
Why isn't my regex working? It just returns back the original file. My file looks like this (for a few hundred lines):
1 Germany 1765 0 Equal
2 Argentina 1631 0 Equal
3 Colombia 1488 1 Up
4 Netherlands 1456 -1 Down
5 Belgium 1444 0 Equal
6 Brazil 1291 1 Up
7 Uruguay 1243 -1 Down
8 Spain 1228 -1 Down
9 France 1202 1 Up
...
192 US Virgin Islands 28 -1 Down
And I want this:
Germany,1
Argentina,2
Colombia,3
...
US Virgin Islands,192
This is the regex I tried:
sed 's/\([0-9]*\)\t\([a-zA-Z]*\)/\2,\1/g' <fifa.csv >fifa.csv
But it just returns the original file.
EDIT:
Now I tried
sed 's/\([0-9]*\)\t\([a-zA-Z]*\)/\2,\1/g' <fifa.csv >fifa.csv
and got
,1 Germany,,1765Equal,0,
,2 Argentina,,1631Equal,0,
,3 Colombia,,1488Up,1,
,4 Netherlands,,1456-Down,1,
,5 Belgium,,1444Equal,0,
You could try the below sed command if the fields are tab-separated.
sed 's/^\([0-9]\+\)\t\([^\t]*\).*/\2,\1/' file
Add the inline-edit option -i to save the changes made.
sed -i 's/^\([0-9]\+\)\t\([^\t]*\).*/\2,\1/' file
^ means start of the line anchor. + would repeat the previous character one or more times. Basic sed uses BRE so you need to escape the + to do the functionality of repeating the previous character one or more times. [^\t]* matches any character but not of \t tab character zero or more times.
The following is what you are looking for. The -i option specifies that files are to be edited in-place.
sed -i 's/^\([0-9]\+\)\t\([^\t]*\).*/\2,\1/' fifa.csv
awk '{print( $2 "," $1)}' YourFile
not a sed but easier to manage
I have a whitespace delimited file with a variable number of entries on each line. I want to replace the first two whitespaces with commas to create a comma delimited file with three columns.
Here's my input:
a b 1 2 3 3 2 1
c d 44 55 66 2355
line http://google.com 100 200 300
ef jh 77 88 99
z y 2 3 33
And here's my desired output:
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line,http://google.com,100 200 300
ef,jh,77 88 99
z,y,2 3 33
I'm trying to use perl regular expressions in a sed command but I can't quite get it to work. First I try capturing a word, followed by a space, then another word, but that only works for lines 1, 2, and 5:
$ cat test | sed -r 's/(\w)\s+(\w)\s+/\1,\2,/'
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line http://google.com 100 200 300
ef jh 77 88 99
z,y,2 3 33
I also try capturing whitespace, a word, and then more whitespace, but that gives me the same result:
$ cat test | sed -r 's/\s+(\w)\s+/,\1,/'
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line http://google.com 100 200 300
ef jh 77 88 99
z,y,2 3 33
I also try doing this with the .? wildcard, but that does something funny to line 4.
$ cat test | sed -r 's/\s+(.?)\s+/,\1,/'
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line http://google.com 100 200 300
ef jh,,77 88 99
z,y,2 3 33
Any help is much appreciated!
How about this:
sed -e 's/\s\+/,/' | sed -e 's/\s\+/,/'
It's probably possible with a single sed command, but this is sure an easy way :)
My output:
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line,http://google.com,100 200 300
ef,jh,77 88 99
z,y,2 3 33
Try this:
sed -r 's/\s+(\S+)\s+/,\1,/'
Just replaced \w (one "word" char) with \S+ (one or more non-space chars) in one of your attempts.
You can provide multiple commands to a single instance of sed by just providing multiple -e arguments.
To do the first two, just use:
sed -e 's/\s\+/,/' -e 's/\s\+/,/'
This basically runs both commands on the line in sequence, the first doing the first block of whitespace, the second doing the next.
The following transcript shows this in action:
pax$ echo 'a b 1 2 3 3 2 1
c d 44 55 66 2355
line http://google.com 100 200 300
ef jh 77 88 99
z y 2 3 33
' | sed -e 's/\s\+/,/' -e 's/\s\+/,/'
a,b,1 2 3 3 2 1
c,d,44 55 66 2355
line,http://google.com,100 200 300
ef,jh,77 88 99
z,y,2 3 33
Sed s/// supports a way to say which occurrence of a pattern to replace: just add the n to the end of the command to replace only the nth occurrence. So, to replace the first and second occurrences of whitespace, just use it this way:
$ sed 's/ */,/1;s/ */,/2' input
a,b ,1 2 3 3 2 1
c,d ,44 55 66 2355
line,http://google.com 100,200 300
ef,jh ,77 88 99
z,y 2,3 33
EDIT: reading another proposed solutions, I noted that the 1 and 2 after s/ */,/ is not only unnecessary but plainly wrong. By default, s/// just replaces the first occurrence of the pattern. So, if we have two identical s/// in sequence, they will replace the first and the second occurrence. What you need is just
$ sed 's/ */,/;s/ */,/' input
(Note that you can put two sed commands in one expression if you separate them by a semicolon. Some sed implementations do not accept the semicolon after the s/// command; use a newline to separate the commands, in this case.)
A Perl solution is:
perl -pe '$_=join ",", split /\s+/, $_, 3' some.file
Not sure about sed/perl, but here's an (ugly) awk solution. It just prints fields 1-2, separated by commas, then the remaining fields separated by space:
awk '{
printf("%s,", $1)
printf("%s,", $2)
for (i=3; i<=NF; i++)
printf("%s ", $i)
printf("\n")
}' myfile.txt