mjhjmtu

I have a netlist file. I has splited lines. The split sign is '+' at the beginning of one line.

MP1 Y A VDD VDD pch_25 W=1u L=550n M=1 
MP0 Y B 
+VDD VDD pch_25 W=1u L=550n M=1 
MN1 v0 A VSS VSS nch_25_dnw W=500.0n L=550n M=1 
MN0 Y B v0 VSS nch_25_dnw W=500.0n L=550n M=1 

I'd like to replace the beginning M of each line with X if that line contain pch_25. I have the following command:

sed -e :a -e '$!N;s/n+/\/;ta' -e 'P;D' file | sed '/pch_25/s/^M/X/' | sed -e 's/\/n+/g' 

How to replace this command with a single sed without pipeline?

Obviously the format allows multiple line breaks, so the N;P;D pattern doesn't fit here really good: If you collect more lines, you need to print and remove all but one, but P;D doesn't do that, that's why you do the replacing stuff.

Here is the first way I can think of:

sed -e :a -e '$be' -e 'N;/n[^+]/!ba' -e :e -e 'h;s/n[^+].*//;/pch_25/s/^M/X/;p;g;/n[^+]/!d;s/.*n//;ba' file

We loop to collect lines until we hit the last line or a line not starting with +. Then save the pattern in the hold space and remove the last line from the space, if it's not starting with +. Now all lines that form one record are together in the pattern space; we can do the desired replacement and print it.

Now the g gets back the pattern before removing the beginning of the next record. If it doesn't contain a newline followed by a non-+, it's been the last record and we can delete it and exit. Otherwise remove everything but the last line and go back to mark :a.

(Note that you can't simply use s/.*n([^+])/1/;ta, because the t could be triggered by an earlier s replacement.)

Update: Alternative dirty trick to use N;P;D anyhow:

sed -e '/^end$/d;/^+/P;D;;:a' -e '$x;s/.*/end/;H;x;;$!N;/n[^+]/!ba' -e '/pch_25.*n/s/^M/X/;P;D' file