Delete 2-line paragraphs, keep 3-line paragraphs, maintain newline between all

Question

On a Linux machine with GNU tools I have an ascii file that consists of only 2-line and 3-line paragraphs, each paragraph separated by a newline. All 2-line paragraphs are to be deleted so that all 3-line paragraphs remain separated by a newline. I've been experimenting unsuccessfully with awk paragraphs. Original content:

aaa aaaaaa
bb bbb 
cc cccc ccccc cccc

ddd d
e eeeeee ee

fff ff
ggg gggg gggg
hhh hhh hh hhhh

ii
jjjj jjjj jjjjjj j jjj

The above should look like this:

aaa aaaaaa
bb bbb 
cc cccc ccccc cccc

fff ff
ggg gggg gggg
hhh hhh hh hhhh

I would prefer sed, awk, or perl for this.

Answer 1

With GNU awk:

awk 'BEGIN{FS=RS; RS=""} NF==3{print; print ""}' file

Output:

aaa aaaaaa
bb bbb 
cc cccc ccccc cccc

fff ff
ggg gggg gggg
hhh hhh hh hhhh

Output contains one trailing empty line.

See: 8 Powerful Awk Built-in Variables – FS, OFS, RS, ORS, NR, NF, FILENAME, FNR

Answer 2

Using any awk with a blank line at the end of the output:

$ awk -v RS= -F'\n' -v ORS='\n\n' 'NF != 2' file
aaa aaaaaa
bb bbb
cc cccc ccccc cccc

fff ff
ggg gggg gggg
hhh hhh hh hhhh

Using any awk without a blank line at the end of the output:

$ awk -v RS= -F'\n' 'NF != 2{print sep $0; sep=ORS}' file
aaa aaaaaa
bb bbb
cc cccc ccccc cccc

fff ff
ggg gggg gggg
hhh hhh hh hhhh

Answer 3

You can use perl for this (along with several other possibilities)

perl -n -00 -e 'chomp; print "$_\n\n" unless y/\n// == 1' datafile

Or expanded out somewhat,

perl -n -00 -e '
    # -n -00 = while { } loop reading a paragraph at a time
    # -e = execute the following code in the loop
    chomp;                                # Discard trailing newline (if any)
    print "$_\n\n" unless y/\n// == 1;    # Output paragraphs unless they have just 1 newline
' datafile

Result from your example data:

aaa aaaaaa
bb bbb
cc cccc ccccc cccc

fff ff
ggg gggg gggg
hhh hhh hh hhhh