[kman42]

I'm new to regular expression thing and could use some help using sed. I'm trying to remove some numbers from a string of text (DNA sequence, actually).

I've tried this command:

% sed 's/[0-9]//g' file-name

but it didn't seem to have any effect. Here's my text sequence if that helps:

GCAAAACTCCAATGAAGTCAGAGCCCCCTACTCGCCGCCGCGGCCAGGCC CCCCAACATG61GACTGTCT^MCTGTATAGTGACAACCAAG$
AGTATGAGGAGATCACATTGGAAAGGGG^MTAACTCAGGTCTGGGCTTCA GCATCGCAGGT301GGCACCGACAACCCACACATCGGTGAC$
TCCATCGTTCGCCTCTACGTCATGCGCCGGAAACCCCCAGCTGAGAAGAT CATAGAG^MATC541AAGCTTATCAAAGGGCCTAAAGGACT$
GATGTCATGCATGAG^MGACGCCGTGGCAGCCCTGAAGAACACATATGAC GTTGTGTACCTA781AAGGTGGCCAAGCCCAGCAATG^MCC$
1CTAGGGGAGGAAGATATTCCCCGGGAACCAAGGCGGATCGTGA^MTCCA TCGGGGCTCCACC1021GGCCTGGGCTTCAACATTGTGGGC$
C1^M201GCGGGTCAGACGGTCACGATCATCGCTCAGTATAAACCAGAAG AGTATAGCCGATTCGAG1261GCC^MAAGATCCATGATCTT$
CTTCAT1441GTAATTGACGCCAGCGACGAA^MGAGTGGTGGCAAGCGCG GCGGGTCCACTCTGACAGTGAG1501ACCGATGACATTGGC$
TCATCCTTGGG1681CCTACCAAAGACCGTGCCAACGATGATCTTCTCTC CGAGTTCCCC^MGACAAGTTTGGATCC1741TGTGTCCCTC$
CGAGAGGTAGCAGAG192^M1CAGGGGAAGCACTGCATCCTTGATGTCTC AGCCAATGCCGTGCGGCGGCTGCAGGCGGCC1981CACCT^$
ACAGCTTTGAAGAGATCTAT2161CACAAAGTGAAACGTGTCATCGA^MA GACCTCTCAGGCCCCTACATCTGGGTCCCAGCCCGA2221G$


I'd also like to get rid of the $ and ^, but I figure that should be easy if I can get the numbers out.

thanks!
kman

[wataru]

I have no idea what that command you tried was, but couldn't you do what you want in TextEdit with search-and-replace?

[Millennium]

Originally posted by wataru:
I have no idea what that command you tried was, but couldn't you do what you want in TextEdit with search-and-replace?

Not really. Not nearly as easily, at any rate, giventhat TextEdit doesn't support regular expressions in its search and replace.

But there's another reason why he might want to use the Terminal: it's easier to integrate this into other applications than many search-and-replace GUI solutions. The idea is to use the sed language (which is very powerful for working with streams of bytes) in conjunction with other applications.

Unfortunately, I don't know sed. But I know this belongs in the Unix forum. You're more likely to find people who know sed there.

[jayg]

You might want to put a -e before the expression, e.g.:
sed -e 's/[0-9]//g' filename

I would think both ways would work though. They both do on Solaris. Are you using a file with Mac or Unix style eols? Maybe sed doesn't like Mac style.

[kman42]

Well, it does seem to work with either unix or mac linebreaks.

I wasn't outputting the result to a file so I wasn't getting the substitution to stick. Another unix lesson learned.

Thanks for the help.

kman