Remove lines between patterns nearest another pattern - awk

Question

I'm trying to remove certain objects from a PDF file. The objects all look like this:
40 0 obj
<<
/PieceInfo
/Subtype /Form
/Resources
<<
/Font
<<
/Fm1 35 0 R
>>
>>
/Type /XObject
/BBox [0 -22.5 131.05 0]
/Length 601
/Matrix [1 0 0 1 0 0]
>>
stream
. . .
A bunch of compressed gibberish here
. . .
endstream
endobj
What I found to work without breaking the PDF document is deleting the stuff between obj, stream, and endstream.
Is there a way in sed or awk to look for lines containing /Form, and then deleting everything between the nearest obj above and stream below, and that stream and the endstream below it, so that the final result looks like this:
40 0 obj
stream
endstream
endobj

Answer 1

Given:
$ echo "$pdf"
40 0 obj
<<
/PieceInfo
/Subtype /Form
/Resources
<<
/Font
<<
/Fm1 35 0 R
>>
>>
/Type /XObject
/BBox [0 -22.5 131.05 0]
/Length 601
/Matrix [1 0 0 1 0 0]
>>
stream
. . .
A bunch of compressed gibberish here
. . .
endstream
endobj
You can use perl:
$ echo "$pdf" | perl -0777 -lne 'print "$1$2$3\n" if /(^.*(?<=\bobj)\s*\R)[\s\S]*?\/Form[\s\S]*?^(stream\s*^)[\s\S]*?^(endstream\s+endobj)/m'
40 0 obj
stream
endstream
endobj
Demo and explanation of regex

Answer 2

perl -0777 -pe 's/(?<=obj)[\s\S]+?\/Form[\s\S]+?\n(?=endstream)/\nstream\n/g' pdf
There are a LOT of ways this regex can backfire (The key issue is "obj" or "endstream" appearing midstream or those fields or "/Form" being missing). You'll need a full script for something production quality, in which case you would definitely need to "show your work" to get help. Also you may need to remove or change the \n before (?=endstream) for an actual PDF. I'm not familiar with the line end characters it uses.
The jist, as a glob, would be that it looks for obj*/Form*endstream, and then clobbers everything not in a lookaround (?[etc]), and it manually readds the stream line.

Answer 3

awk may also do the job,
awk '/[^end]obj/||/[end]*stream/{print;if(d==1){s=""}d=1;next}{s=s $0}END{print s}' pdf
Brief explanation,
/[^end]obj/||/[end]*stream/: locate the string 'obj', 'stream', and 'endstream'
If above string is existed in the line, print it and enable the flag d
If the d is already enabled, clear the buffer str
Print the str in the end

Answer 4

This might work for you (GNU sed):
sed -r '/\<obj\>/{n;:a;/\<endobj\>/!{N;ba};s/.*\<(stream)\>.*\<(endobj)\>/\1\n\2/}' file
Gather up the lines between obj and endobj and remove the parts either side of stream.

Answer 5

$ cat tst.awk
$NF == "endobj" {
print (obj ~ "/Form" ? "stream" ORS "endstream" : obj)
obj = ""
inObj = 0
}
inObj { obj = (obj == "" ? "" : obj ORS) $0 }
!inObj { print }
$NF == "obj" { inObj = 1 }
$ awk -f tst.awk file
40 0 obj
stream
endstream
endobj