alternative for "sed -i" (and without temp file)

[w5000]

hello,
case: sed on a system doesn't support -i

I need to remove leading hashe(s) from matched patern lines in a file.
I can sed + temp file but how could I skip temp file and do change it directly in text file?

$ echo "        ############# 456\n\t ## 12345 \t\n     2345     # ## ####12345#\t   aa\n%### \t12345 bb\n11 22 33 44 55"
        ############# 456
         ## 12345
     2345     # ## ####12345#      aa
%###    12345 bb
11 22 33 44 55
$ D=12345
$ echo "        ############# 456\n\t ## 12345 \t\n     2345     # ## ####12345#\t   aa\n%### \t12345 bb\n11 22 33 44 55"|sed "/${D}/ s/^[[:space:]]*//;/${D}/ s/[[:space:]]*$//;/${D}/ s/[[:space:]][[:space:]]*/\ /g;/${D}/ s/^#[# ]*//"
        ############# 456
12345
2345 # ## ####12345# aa
%### 12345 bb
11 22 33 44 55
$

 

[feherke]

Hi

The usual answer is, use Perl ( or Ruby ) :

perl -ipe 's/^#+//' /input/file

[gray]# or[/gray]

ruby -ipe 'sub /^#+/,""' /input/file

Or if the file is small, use a variable instead of temporary file :

content="$( sed 's/^#[# ]*//' /input/file )"
echo "$content" > /input/file

how could I skip temp file and do change it directly in text file?

That is really difficult. Not even [tt]sed -i[/tt] does such thing. ( It writes the result to a new file and renames it after. )

( Note that I not reproduced the entire Sed code to to keep the examples simple and readable. )

Feherke.
feherke.github.io