I have a ~23000 line SQL dump containing several databases worth of data. I need to extract a certain section of this file (i.e. the data for a single database) and place it in a new file. I know both the start and end line numbers of the data that I want.
Does anyone know a Unix command (or series of commands) to extract all lines from a file between say line 16224 and 16482 and then redirect them into a new file?
sed -n '16224,16482p;16483q' filename > newfile
From the sed manual:
p - Print out the pattern space (to the standard output). This command is usually only used in conjunction with the -n command-line option.
n - If auto-print is not disabled, print the pattern space, then, regardless, replace the pattern space with the next line of input. If there is no more input then sed exits without processing any more commands.
q - Exit
sedwithout processing any more commands or input. Note that the current pattern space is printed if auto-print is not disabled with the -n option.
Addresses in a sed script can be in any of the following forms:
number Specifying a line number will match only that line in the input.
An address range can be specified by specifying two addresses separated by a comma (,). An address range matches lines starting from where the first address matches, and continues until the second address matches (inclusively).
sed -n '16224,16482 p' orig-data-file > new-file
Where 16224,16482 are the start line number and end line number, inclusive. This is 1-indexed. -n suppresses echoing the input as output, which you clearly don't want; the numbers indicate the range of lines to make the following command operate on; the command p prints out the relevant lines.
There is another approach with awk:
awk 'NR==16224, NR==16482' file
If the file is huge, it can be good to exit after reading the last desired line. This way, it won't read the following lines unnecessarily:
awk 'NR==16224, NR==16482-1; NR==16482 {print; exit}' file
awk 'NR==16224, NR==16482; NR==16482 {exit}' file
Standing on the shoulders of boxxar, I like this:
sed -n '<first line>,$p;<last line>q' input
e.g.
sed -n '16224,$p;16482q' input
The $ means "last line", so the first command makes sed print all lines starting with line 16224 and the second command makes sed quit after printing line 16428. (Adding 1 for the q-range in boxxar's solution does not seem to be necessary.)
I like this variant because I don't need to specify the ending line number twice. And I measured that using $ does not have detrimental effects on performance.
I wrote a Haskell program called splitter that does exactly this: have a read through my release blog post.
You can use the program as follows:
$ cat somefile | splitter 16224-16482
And that is all that there is to it. You will need Haskell to install it. Just:
$ cabal install splitter
And you are done. I hope that you find this program useful.
I wanted to do the same thing from a script using a variable and achieved it by putting quotes around the $variable to separate the variable name from the p:
sed -n "$first","$count"p imagelist.txt >"$imageblock"
I wanted to split a list into separate folders and found the initial question and answer a useful step. (split command not an option on the old os I have to port code to).
Just benchmarking 3 solutions given above, that works to me:
Credits on the 3 solutions goes to:
I'm using a huge file I find in my server:
# wc fo2debug.1.log
10421186 19448208 38795491134 fo2debug.1.log
38 Gb in 10.4 million lines.
And yes, I have a logrotate problem. : ))
Getting 256 lines from the beginning of the file.
# time sed -n '1001,1256p;1256q' fo2debug.1.log | wc -l
256
real 0m0,003s
user 0m0,000s
sys 0m0,004s
# time head -1256 fo2debug.1.log | tail -n +1001 | wc -l
256
real 0m0,003s
user 0m0,006s
sys 0m0,000s
# time awk 'NR==1001, NR==1256; NR==1256 {exit}' fo2debug.1.log | wc -l
256
real 0m0,002s
user 0m0,004s
sys 0m0,000s
Awk won. Technical tie in second place between sed and "head+tail".
Getting 256 lines at the end of the first third of the file.
# time sed -n '3473001,3473256p;3473256q' fo2debug.1.log | wc -l
256
real 0m0,265s
user 0m0,242s
sys 0m0,024s
# time head -3473256 fo2debug.1.log | tail -n +3473001 | wc -l
256
real 0m0,308s
user 0m0,313s
sys 0m0,145s
# time awk 'NR==3473001, NR==3473256; NR==3473256 {exit}' fo2debug.1.log | wc -l
256
real 0m0,393s
user 0m0,326s
sys 0m0,068s
Sed won. Followed by "head+tail" and, finally, awk.
Getting 256 lines at the end of the second third of the file.
# time sed -n '6947001,6947256p;6947256q' fo2debug.1.log | wc -l
A256
real 0m0,525s
user 0m0,462s
sys 0m0,064s
# time head -6947256 fo2debug.1.log | tail -n +6947001 | wc -l
256
real 0m0,615s
user 0m0,488s
sys 0m0,423s
# time awk 'NR==6947001, NR==6947256; NR==6947256 {exit}' fo2debug.1.log | wc -l
256
real 0m0,779s
user 0m0,650s
sys 0m0,130s
Same results.
Sed won. Followed by "head+tail" and, finally, awk.
Getting 256 lines near the end of the file.
# time sed -n '10420001,10420256p;10420256q' fo2debug.1.log | wc -l
256
real 1m50,017s
user 0m12,735s
sys 0m22,926s
# time head -10420256 fo2debug.1.log | tail -n +10420001 | wc -l
256
real 1m48,269s
user 0m42,404s
sys 0m51,015s
# time awk 'NR==10420001, NR==10420256; NR==10420256 {exit}' fo2debug.1.log | wc -l
256
real 1m49,106s
user 0m12,322s
sys 0m18,576s
And suddenly, a twist!
"Head+tail" won. Followed by awk and, finally, sed.
(some hours later...)
My analysis above ends up being an example of a basic flaw in doing an analysis.
The flaw is not knowing in depth the resources used for the analysis.
In this case, I used a log file to analyze the performance of a search for a certain number of lines within it.
Using 3 different techniques, searches were made at different points in the file, comparing the performance of the techniques at each point and checking whether the results varied depending on the point in the file where the search was made.
My mistake was to assume that there was a certain homogeneity of content in the log file.
The reality is that long lines appear more frequently at the end of the file.
Thus, the apparent conclusion that longer searches (closer to the end of the file) are better with a given technique, may be biased. In fact, this technique may be better when dealing with longer lines. What remains to be confirmed.
People trying to wrap their heads around computing an interval for the head | tail combo are overthinking it.
Here's how you get the "16224 -- 16482" range without computing anything:
cat file | head -n +16482 | tail -n +16224
Explanation:
The + instructs the head/tail command to "go up to / start from" (respectively) the specified line number as counted from the beginning of the file.
Similarly, a - instructs them to "go up to / start from" (respectively) the specified line number as counted from the end of the file
The solution shown above simply uses head first, to 'keep everything up to the top number', and then tail second, to 'keep everything from the bottom number upwards', thus defining our range of interest (with no need to compute an interval).
I wrote a small bash script that you can run from your command line, so long as you update your PATH to include its directory (or you can place it in a directory that is already contained in the PATH).
Usage: $ pinch filename start-line end-line
#!/bin/bash
# Display line number ranges of a file to the terminal.
# Usage: $ pinch filename start-line end-line
# By Evan J. Coon
FILENAME=$1
START=$2
END=$3
ERROR="[PINCH ERROR]"
# Check that the number of arguments is 3
if [ $# -lt 3 ]; then
echo "$ERROR Need three arguments: Filename Start-line End-line"
exit 1
fi
# Check that the file exists.
if [ ! -f "$FILENAME" ]; then
echo -e "$ERROR File does not exist. \n\t$FILENAME"
exit 1
fi
# Check that start-line is not greater than end-line
if [ "$START" -gt "$END" ]; then
echo -e "$ERROR Start line is greater than End line."
exit 1
fi
# Check that start-line is positive.
if [ "$START" -lt 0 ]; then
echo -e "$ERROR Start line is less than 0."
exit 1
fi
# Check that end-line is positive.
if [ "$END" -lt 0 ]; then
echo -e "$ERROR End line is less than 0."
exit 1
fi
NUMOFLINES=$(wc -l < "$FILENAME")
# Check that end-line is not greater than the number of lines in the file.
if [ "$END" -gt "$NUMOFLINES" ]; then
echo -e "$ERROR End line is greater than number of lines in file."
exit 1
fi
# The distance from the end of the file to end-line
ENDDIFF=$(( NUMOFLINES - END ))
# For larger files, this will run more quickly. If the distance from the
# end of the file to the end-line is less than the distance from the
# start of the file to the start-line, then start pinching from the
# bottom as opposed to the top.
if [ "$START" -lt "$ENDDIFF" ]; then
< "$FILENAME" head -n $END | tail -n +$START
else
< "$FILENAME" tail -n +$START | head -n $(( END-START+1 ))
fi
# Success
exit 0
Since we are talking about extracting lines of text from a text file, I will give an special case where you want to extract all lines that match a certain pattern.
myfile content:
=====================
line1 not needed
line2 also discarded
[Data]
first data line
second data line
=====================
sed -n '/Data/,$p' myfile
Will print the [Data] line and the remaining. If you want the text from line1 to the pattern, you type: sed -n '1,/Data/p' myfile. Furthermore, if you know two pattern (better be unique in your text), both the beginning and end line of the range can be specified with matches.
sed -n '/BEGIN_MARK/,/END_MARK/p' myfile
The -n in the accept answers work. Here's another way in case you're inclined.
cat $filename | sed "${linenum}p;d";
This does the following:
I was looking for an answer to this but I had to end up writing my own code which worked. None of the answers above were satisfactory. Consider you have very large file and have certain line numbers that you want to print out but the numbers are not in order. You can do the following:
My relatively large file
for letter in {a..k} ; do echo $letter; done | cat -n > myfile.txt
1 a
2 b
3 c
4 d
5 e
6 f
7 g
8 h
9 i
10 j
11 k
Specific line numbers I want:
shuf -i 1-11 -n 4 > line_numbers_I_want.txt
10
11
4
9
To print these line numbers, do the following.
awk '{system("head myfile.txt -n " $0 " | tail -n 1")}' line_numbers_I_want.txt
What the above does is to head the n line then take the last line using tail
If you want your line numbers in order, sort ( is -n numeric sort) first then get the lines.
cat line_numbers_I_want.txt | sort -n | awk '{system("head myfile.txt -n " $0 " | tail -n 1")}'
4 d
9 i
10 j
11 k
