sed question? 1

[cpjust]

I don't understand this. I have a text file with this in it:

    1234

(that's 4 spaces, not a tab)

If I run this command to remove the spaces, it works:

sed -i -e "s/^[ ]*//" out.txt

but if I run this command it doesn't work:

sed -i -e "s/^[ ]+//" out.txt

+ means 1 or more of the previous characters, and there are definitely 1 or more spaces, so why doesn't the 2nd command work?

 

[AdaHacker]

You're correct, but in this case sed thinks that you're using the + as a character to match, rather than to denote one or more occurrences. Try escaping that + and it should work.

sed -i -e "s/^[ ][COLOR=red]\[/color]+//" out.txt

 

[cpjust]

What the hell?? That worked.
Shouldn't that tell it to look for a space followed by a '+' if you escape it?
Why didn't I have to escape the * then?

 

[AdaHacker]

Good question. I actually never bothered to look that up until just now.

It appears the reason is that, by default, sed uses Basic Regular Expressions (BRE), which do not include +, ?, and a few others. Those are only in Extended Regular Expressions (ERE). Thus sed treats them as normal characters. It appears that escaping an ERE metacharacter will cause sed to treat give them special treatment.

Alternatively, you can just tell sed to use ERE by passing it the -r option. That way you don't even have to worry about it.

 

[cpjust]

Cool thanks. I thought it had to be something weird like that.