Remove new line character by checking the expression, using sed

Question

Have to write a script which updates the file in this way.

raw file:

<?blah blah blah?>
<pen>
<?pineapple?>
<apple>
<pen>

Final file:

<?blah blah blah?><pen>
<?pineapple?><apple><pen>

Where ever in the file if the new line charter is not followed by

<?

We have to remove the newline in order to append it at the end of previous line.

Also it will be really helpful if you explain how your sed works.

Answer 1

Perl solution:

perl -pe 'chomp; substr $_, 0, 0, "\n" if $. > 1 && /^<\?/'

• -p reads the input line by line, printing each line after changes
• chomp removes the final newline
• substr with 4 arguments modifies the input string, here it prepends newline if it's not the first line ($. is the input line number) and the line starts with <?.

Answer 2

Sed solution:

sed ':a;N;$!ba;s/\n\(<[^?]\)/\1/g' file > newfile

The basic idea is to replace every

\n followed by < not followed by ?

with what you matched except the \n.

Answer 3

When you are happy with a solution that puts every <? at the start of a line, you can combine tr with sed.

tr -d '\n' < inputfile| sed 's/<?/\n&/g;$s/$/\n/'

Explanation:
I use tr ... < inputfile and not cat inputfile | tr ... avoiding an additional catcall.
The sed command has 2 parts.
In s/<?/\n&/g it will insert a newline and with & it will insert the matched string (in this case always <?, so it will only save one character).
With $s/$/\n/ a newline is appended at the end of the last line.

EDIT: When you only want newlines before <? when you had them already, you can use awk:

awk '$1 ~ /^<\?/ {print} {printf("%s",$0)} END {print}'

Explanation:
Consider the newline as the start of the line, not the end. Then your question transposes into "write a newline when the line starts with <?. You must escape the ? and use ^ for the start of the line.

awk '$1 ~ /^<\?/ {print}'

Next print the line you read without a newline character. And you want a newline at the end.