OK, step by step! Try this one (replace "<t>" by literal tab characters):

Code:

sed '/Pt[ <t>]*\([0-9]\.[0-9]\{14\}[ <t>]*\)\{2\}[0-2]/ s/$/  bottom!/' /path/to/input > /path/to/output

This should mark the bottom lines with the word "bottom!". I find it easier to verify something i see therefore i would first modify the lines, verify that there are no false positives, false negatives, etc., and only then proceed. If everything is indeed correct replace the substitution with a simple delete-command:

Code:

sed '/Pt[ <t>]*\([0-9]\.[0-9]\{14\}[ <t>]*\)\{2\}[0-2]/d' /path/to/input > /path/to/output

You surely want to know how this works: we start with the innermost part:

Code:

[0-9]\.[0-9]\{14\}[ <t>]*

This matches a digit, followed by a ".", followed by another 14 digits, followed by any number of whitespace (i do not know if your numbers are separated by blanks or tabs). This matches one cooordinate.

Code:

Pt[ <t>]*\(<coordinate>\)\{2\}[0-2]

This matches a "Pt", followed by whitespace, followed by two such coordinates, followed by a digit between 0 and 2, hence a 0,1 or 2 in the first digit of the z-coordinate. This should identify the lines of the bottom layers.

We use this regexp in a "one-address-command" of sed. The command (in our case a substution or a deletion) is executed only in a line which fits onto the regexp:

Code:

/^a/ s/x/y/g

Change all "x" to "y", but only in lines which begin with "a".

If this does what you need we proceed to the next part of your problem.

I hope this helps.

bakunin