UNIX for Dummies Questions & Answers

I want to delete all the blank lines from a file before a certain line number. e.g.

Input file (n: denotes line number)

Code:

1: a
2: 
3: b
4: c
5: 
6: d

I want to delete all blank lines before line number 3, such that my output is:

Code:

a
b
c

d

I see that

Code:

 sed '/^$/d' in_file

works fine but works on the whole file.

Code:

 sed '1,3 /^$/d' 1

is incorrect. How can I achieve this?

Code:

awk 'NR>3 || ! /^$/ {print}' in_file

---------- Post updated at 03:12 AM ---------- Previous update was at 03:00 AM ----------

You can also do it in sed with:

Code:

sed -n '1,3p;n
/^$/d
p' in_file

Try the sed version

Code:

sed '1,3 {/^$/d}' infile

You got it right - almost - but missed by a hairs width.

Any command (like "s", etc.) can be preceeded by a range clause. A range clause can be either a range of lines, determined by line numbers or a starting/ending regexp, or it can be a single line number or regexp. In the second case the command would either apply to the single line number or to all lines matching the regexp:

Code:

1,3 s/x/y/
/^aa/,/^bb/ s/x/y/
1 s/x/y/
/^abc/ s/x/y/

The first line would apply "s/x/y/" only to lines 1-3, the second line would apply it to all lines from a line starting with "aa" to a line starting with "bb".

The third line would apply the command only to line number 1 and the fourth to all lines starting with "abc".

This you did correctly, but range clauses cannot be nested, and

Code:

/^$/d

is also a range clause: one, that applies the command "d" only to lines of the form "/^$/", like in my fourth example. To correct your code you simply have to correctly nest the commands:

Code:

1,3 {
     /^$/ {
          d
     }
}

or, in short form:

Code:

1,3 {;/^$/d;}

I hope this helps.

bakunin

All of the sed documentation I've seen says that the "{" command format is:
The use you're showing here should be documented as:
[QUOTE][2addr] {command
command
. . .
}[/QOTE]
because "function" doesn't include an address. I wonder if this is something that all implementation of sed do (but never documented) {and should be included in a future version of the POSIX and UNIX standards}, or if it is only provided by some implementations of sed?

"function" is probably meant recursively. A function itself is a list of commands and may contain other functions as well. This is a principle in probably any programming language.

To be honest i haven't read the POSIX standard papers as intensively as you seem to have done, so i can only guess. What i have explained above is true for all the sed-variants i have ran across in my nearly 30 years of Unix, so even if my explanation is not completely POSIXly-correct it seems to cover the truth from a practical point of view.

Still, i would be interested if you could find out how it is really supposed to work as i would prefer to be sure instead of almost sure. It seems like you could shed some light on it.

I hope this helps.

bakunin

Check your local sed(1) man page... I bet it'll say something like:
And later give a list of sed functions and the number of addresses that need to be supplied with them to form a command.

I'll post a question to the POSIX mailing list to see if I can find an example where:
doesn't work.