对文本文件中的每一行进行排序

Question

I have a text file which contains on each line some words, for example:

stackoverflow coding programming
tag question badges

I must sort each line and preserve the order of lines. For example, for the above example the output should be:

coding programming stackoverflow
badges question tag

My solution until now is to create a temp file, in which all the lines are sorted. The bash script looks like this:

FILE_TMP=$FILE".tmp" 
while read line
do
echo $line | xargs -n1 | sort | xargs >>$FILE_TMP
done < $FILE

mv $FILE_TMP $FILE

It works fine, but I'm not pleased that I must create a duplicate file, especially because the files are big.

So, my question is there any solution to sort in place each line of the file?

Thank you,

Answer 1

试试这个（如果文件没有空格分隔，你可能需要更改sed）：

cat datafile.dat | while read line; do echo $line | sed 's/ /\n/g' | sort | gawk '{line=line " " $0} END {print line}' ; done

Answer 2

If Python were an option, this would be quite easy using the in-place support from the fileinput module

>>> import os
>>> import fileinput
>>> for line in fileinput.input('file.txt', inplace=1):
...     line = line.rstrip(os.linesep)
...     print(' '.join(sorted(line.split())))
...

Answer 3

You could script a text editor (vim or emacs, for example) to do it "in place", but that wouldn't really help you avoid using a temp file since text editors will internally use temp files.

If your real problem is that it is slow to run, that is probably because it is spawning 3 different processes for each line in the source file. You could get around that by using a scripting language like perl that could go through the file sorting lines without spawning any additional processes. You'd still have an additional file for the output.

Answer 4

The accepted answer is somewhat slow. Try this:

awk ' {split( $0, a, " " ); asort( a ); for( i = 1; i <= length(a); i++ ) printf( "%s ", a[i] ); printf( "\n" ); }' input >output

Note: Your awk must be GNU, so as to have asort().

Answer 5

I think that the following awk goodness should do the job:

prompt$ cat foo.awk
{
    n = split($0, words)
    do {
        change_occured = 0
        for (idx = 1; idx <= n; ++idx) {
            if (words[idx] > words[idx + 1]) {
                t = words[idx]
                words[idx] = words[idx + 1]
                words[idx + 1] = t
                change_occured = 1
            }
        }
    } while (change_occured != 0)
    for (idx in words) {
        printf("%s ", words[idx])
    }
    split("", array)
    print ""
}
prompt$ awk -f foo.awk <<EOF
heredoc> stackoverflow coding programming
heredoc> tag question badges
heredoc> EOF
coding programming stackoverflow  
badges question tag

EDIT note that this is not an in place edit. It acts as a filter from stdin to stdout. You can use awk for this as well but reading and writing files there feels "clunky". If you really want to avoid the temporary file, use something like Perl.

Answer 6

Practically any "reasonable" solution for this problem will write the new contents to a new temporary file and then rename. Even things like perl "in place" processing ( perl -pi... ) or text editors actually do that. If you want to do it really in place , writing to the same physical disk position, it could be done (the new contents occupy exactly the same space as the old) but it's rather painful .

You can compile the code from this answer into a overwrite executable, and then run (WARNING: this is dangerous, backup your file first!)

while read line ; do echo $line | xargs -n1 | sort | xargs ; done < f | ./overwrite f

This is rather fragile, for example, you should be absolutely sure that the sorting that does the script does not mess with blank characters (what about DOS newlines? and consecutive blanks?), the script must spit the same amount (or less) of bytes per line as it eats.