Extracting a random pattern after matching a word in following lines - unix

Question

Extract household data corresponding to a keyword.
Z1/NEW "THE_PALM" 769 121003 1545
NEW HOUSE IN
SOMETHING SOMETHING
SN HOUSE CLASS
FIRST PSD93_PU 1579
CHAIRS
WOOD
SILVER SPOON
GREEN GARDEN
Z1/OLD "THE_ROSE" 786 121003 1343
NEW HOUSE OUT
SOMETHING NEW
SN HOUSE CLASS
FIRST_O PSD1000_ST 1432
CHAIRS
WOOD
GREEN GARDEN
BLACK PAINT
Z1/OLD "The_PURE" 126 121003 3097
NEW HOUSE IN
SOMETHING OLD
SN HOUSE CLASS
LAST_O JD4_GOLD 1076
CHAIRS
SILVER SPOON
I have a very large sized file. There is a list of items about the house at the end of every description. Corresponding to the houses containing SILVER SPOON, I want to extract the HOUSE ID as in data PSD93_PU and date 121003. I tried the following:
awk 'c-->0;$0~s{if(b)for(c=b+1;c>1;c--)print r[(NR-c+1)%b];print;c=a}b{r[NR%b]=$0}' b=7 a=0 s="SILVER" infile > outfile
But the problem is that the number of lines above the keyword SILVER are so random, that I can't figure out the solution.

Answer 1

assuming each new house starts with Z1
$ awk '$1 ~ /^Z1/ { date=$4; id=""; f=0; next; } \
$1 == "SN" { f=1; next; } \
f == 1 { id=$2; f=0; next; } \
$1" "$2 == "SILVER SPOON" { print id,date }' file
that, on a new house, reset all vars and get the date
if an SN is matched then the next line contains the id
get the id from the line
if "SILVER SPOON" is found print the id and date
if it is not found, a new house will be met and the vars are reset.
test with given data:
$ awk '$1 ~ /^Z1/ { date=$4; id=""; f=0; next; } $1 == "SN" { f=1; next; } f == 1 { id=$2; f=0; next; } $1 == "SILVER SPOON" && $2 == "SPOON" { print id,date }' file
PSD93_PU 121003
JD4_GOLD 121003
note :
if anybody knows how and if $1 == "SILVER" && $2 == "SPOON" can be merge together in one statement that'd be nice :) -- like: $1,$2 == "SILVER SPOON"
edit:
it can be done with $1" "$2 == "SILVER SPOON".
one could possibly omit the space and do $1$2 == "SILVERSPOON" but that would match even if $2 was empty and $1 contained the whole string, or $1 was SILVERSPO and $2 was ON. So the space in acts as a strict match.

Answer 2

Using sed:
sed -n -e 's/^Z1[^"]*"[^"]*"[ \t]*[0-9]*[ \t]*\([0-9]*\).*/\1/p'
-e '/^SN[ \t]*HOUSE/ { n; s/^[^ \t]*[ \t]*\([^ \t]*\).*/\1/p }'
Firstly, we call sed with the -n option in order to tell it to print only what we tell it to.
The first command will search for a particular pattern to extract the date. The pattern consists of:
^Z1: A line starting with the string "Z1".
[^"]*: zero or more characters that aren't double quotes
": double quote character
[^"]*: zero or more characters that aren't double quotes
[ \t]*: zero or more characters that are either tabs or spaces
[0-9]*: zero or more digits
[ \t]*: zero or more characters that are either tabs or spaces
\([0-9]*\): zero or more digits. The backslashed parenthesis are used in order to capture the match, ie. the match is stored into an auxiliary variable \1.
.*: zero or more characters, effectively skipping all characters until the end of the line.
This matched line is then replaced with \1, which holds our captured content: the date. The p after the command tells sed to print the result.
The second line contains two commands grouped together (inside braces) so that they are only executed on the "address" before the braces. The address is a pattern, so that it is executed on every line that matches the pattern. The pattern consists of a line that starts with "SN" followed by a sequence of spaces or tabs, followed by the string "HOUSE".
When the pattern matches, we first execute the n next command, which loads the next line from input. Then, we extract the ID from the new line, in a way analogous to extracting the date. The substitute pattern to match is:
^[^ \t]*: a string that starts with zero or more characters that aren't spaces or tabs (whitespace).
[ \t]*: then has a sequence of zero or more spaces and/or tabs.
\([^ \t]*\): a sequence of non whitespace characters is then captured
.*: the remaining characters are matched so that they are skipped.
The replacement becomes the captured ID, and again we tell sed to print it out.
This will print out a line containing the date, followed by a line containing the ID. If you want a line in the format ID date, you can pipe the output of sed into another sed instance, as follows:
sed -n -e [...] | sed -e 'h;n;G;s/\n/ /'
This sed instance performs the following operations:
Reads a line, and the h command tells it to store the line into the hold space (an auxiliary buffer).
Read the next line with the n command.
The G get command will append the contents of the hold space into the pattern space (the working buffer), so now we have the ID line followed by the date line.
Finally, we replace the new line character by a space, so the lines are joined into a single line.
Hope this helps =)

Answer 3

If your records are separated by two or three blank lines and the line spacing before the household items is consistent, you could use GNU awk like this:
awk -r 'BEGIN { RS="\n{3}\n*"; FS="\n" } /SILVER SPOON/ { split($1, one, OFS); split($6, two, OFS); print two[2], one[4] }' file.txt
Results:
PSD93_PU 121003
JD4_GOLD 121003