How to parse multi line records (with awk?) - regex

Question

I'm trying to figure out how to extract particular fields from multi line records separated by \n\n.
In this instance, it happens to be output from apt-cache akin to DEBIAN control files. See output of apt-cache show "$package"
Package: caffeine
Priority: optional
Section: misc
Installed-Size: 641
Maintainer: Reuben Thomas <rrt#sc3d.org>
Architecture: all
Version: 2.8.3
Depends: python3:any (>= 3.3.2-2~), python3, gir1.2-gtk-3.0, gir1.2-appindicator3-0.1, python3-xlib, python3-pkg-resources, libnet-dbus-perl
Filename: pool/main/c/caffeine/caffeine_2.8.3_all.deb
Size: 58774
MD5sum: 4438db3f6d1cf43a4f4b49cc7f24cda0
SHA1: e748370ac5ccd7de6fc9466ce0451d2e90d179d4
SHA256: ae303b4e32949cc1e1af80df7217e3406291679e3f18fa8f78a5bbb97504c4f6
Description-en: Prevent the desktop becoming idle in full-screen mode
Caffeine stops the desktop becoming idle when an application
is running full-screen. A desktop indicator ‘caffeine-indicator’
supplies a manual toggle, and the command ‘caffeinate’ can be used
to prevent idleness for the duration of any command.
Description-md5: 7c14f8adc007b10f6ecafed36260bedb
Package: caffeine
Priority: optional
Section: misc
Installed-Size: 655
Maintainer: Reuben Thomas <rrt#sc3d.org>
Architecture: all
Version: 2.6+555~ubuntu14.04.1
Depends: python:any (<< 2.8), python:any (>= 2.7.5-5~), python, gir1.2-gtk-2.0, gir1.2-appindicator3-0.1, x11-utils, python-dbus
Filename: pool/main/c/caffeine/caffeine_2.6+555~ubuntu14.04.1_all.deb
Size: 58604
MD5sum: 1051c3f7d40d344f986bb632d7436849
SHA1: 5e5f622595e8cbba8fb7468b3cffe2914b0ba110
SHA256: 11c5bbf2d28dcda6a7b82872195f740f1f79521b60d3c9acea3037bf0ab3a60e
Description: Prevent the desktop becoming idle
Caffeine allows the user to prevent the desktop becoming idle,
either manually or when certain applications are run. This
prevents screen-blanking, locking, suspending, and so on.
Description-md5: 738866350e5086e77408d7a9c7ffa59b
Package: caffeine
Status: install ok installed
Priority: optional
Section: misc
Installed-Size: 794
Maintainer: Isaiah Heyer <freshapplepy#gmail.com>
Architecture: all
Version: 2.4.1+478~raring1
Depends: dconf-gsettings-backend | gsettings-backend, python (>= 2.6), python-central (>= 0.6.11), python-xlib, python-appindicator, python-xdg, python-notify, python-kaa-metadata
Description: Caffeine
A status bar application able to temporarily prevent the activation
of both the screensaver and the "sleep" powersaving mode.
Description-md5: 1c29acf1ab0f2e6636db29fbde1d14a3
Homepage: https://launchpad.net/caffeine
Python-Version: >= 2.6
My desired output is one line per record in the format apt-get download $pkg=$ver -a=$arch. Basically a list of the installation commands for available packages...
So far what I've got is apt-cache show "$package" | awk '/^Package: / { print $2 } /^Version: / { print $2 } /^Architecture: / { print $2 }' | xargs -n3 | awk '{printf "apt-get download %s=%s -a=%s\n", $1, $3, $2}'
This is the actual output:
apt-get download caffeine=2.8.3 -a=all
apt-get download caffeine=2.6+555~ubuntu14.04.1 -a=all
apt-get download caffeine=2.4.1+478~raring1 -a=all
The is as desired but it appears to be a fluke only because the order of the fields is consistent in this example. It would break if the order of fields was different.
I can do parsing like this using object orientation in Python but I'm having difficulty getting this done in one awk command. The only way I can see doing this correctly would be to split each record into individual tmp files (using split or something along those lines) and then parse each file individually (which is straightforward). Obviously I'd really like to avoid unnecessary I/O as this seems like something that awk is well equipped for. Any awk pro's know how to solve this? I'd even be open to a Perl one-liner or utilizing bash but I'm really interested in learning how to better leverage awk.

Answer 1

$ package=sed
$ apt-cache show "$package" | awk '/^Package: /{p=$2} /^Version: /{v=$2} /^Architecture: /{a=$2} /^$/{print "apt-get download "p"="v" -a="a}'
apt-get download sed=4.2.1-10 -a=amd64
How it works
/^Package: /{p=$2}
Save the package information in variable p.
/^Version: /{v=$2}
Save the version information in variable v.
/^Architecture: /{a=$2}
Save the architecture information in variable a.
/^$/{print "apt-get download "p"="v" -a="a}
When we reach a blank line, print out the information in the desired form.
My version of apt-cache always outputs a blank line after each package. Your sample output is missing the last blank line. If your apt-cache genuinely does not produce that last blank line, then we will need to add a little bit more code to compensate.
As a matter of style, some may prefer printf to print. In which case, replace the above with:
/^$/{printf "apt-get download %s=%s -a=%s\n",v,p,a}'

Answer 2

I find the best way to deal with data that contains name to value pairings is to create an array of those pairs and then just access the values by their names:
$ cat tst.awk
BEGIN { RS=""; FS="\n" }
{
delete n2v
for (i=1;i<=NF;i++) {
if ($i !~ /^ /) {
name = gensub(/:.*/,"","",$i)
value = gensub(/[^:]+:\s+/,"","",$i)
n2v[name] = value
}
}
printf "apt-get download %s=%s -a=%s\n",
n2v["Package"], n2v["Version"], n2v["Architecture"]
}
$ awk -f tst.awk file
apt-get download caffeine=2.8.3 -a=all
apt-get download caffeine=2.6+555~ubuntu14.04.1 -a=all
apt-get download caffeine=2.4.1+478~raring1 -a=all
The above uses a couple of gawk extensions but is easily adapted to any awk if necessary.