Finding unique/obsolete binary Debian packages

Question

I have a script I use to prune packages that have gone stale from my custom Debian repository (i.e. they no longer exists on any of the official repositories):

import apt_pkg
import gzip
import subprocess
CUSTOM_REPO = ("/home/tshepang/.custom_repo/dists/tshepang/main/"
 "binary-amd64/Packages.gz")
TEMPLATE = ("/var/lib/apt/lists/http.debian.net_debian_dists_{}_{}_"
 "binary-amd64_Packages")
CODENAMES = 'jessie sid experimental'.split()
ARCHIVE_AREAS = "main contrib non-free".split()
def main():
 custom_repo = apt_pkg.TagFile(gzip.open(CUSTOM_REPO))
 wheezy_packages = list()
 for codename in CODENAMES:
 for archive_area in ARCHIVE_AREAS:
 repo = TEMPLATE.format(codename, archive_area)
 repo = apt_pkg.TagFile(gzip.open(repo, "rb"))
 wheezy_packages.extend([pkg["Package"] for pkg in repo])
 for package in custom_repo:
 package_name = package["Package"]
 if package_name not in wheezy_packages:
 cmd = "apt-cache policy " + package_name
 subprocess.call(cmd.split())
 choice = raw_input("remove from cache [Y/n]? ")
 if not choice or choice.lower().startswith("y"):
 cmd = ("reprepro -vv --basedir /home/tshepang/.custom_repo/ "
 "remove tshepang " + package_name)
 subprocess.call(cmd.split())
if __name__ == "__main__":
 main()

I strongly suspect the code can be better.

As a sidenote, here is sample UI:

apache2.2-common:
 Installed: (none)
 Candidate: 2.2.22-13
 Version table:
 2.2.22-13 0
 500 file:/home/tshepang/.custom_repo/ tshepang/main amd64 Packages
 remove from cache [Y/n]? 
 removing 'apache2.2-common' from 'tshepang|main|amd64'...
 Exporting indices...
 Deleting files no longer referenced...
 deleting and forgetting pool/main/a/apache2/apache2.2-common_2.2.22-13_amd64.deb

Answer 1

Your code looks very good, so I'll just propose some minor (and subjective) improvements:

• Imports: PEP8 has some recommendations about grouping imports that sound pretty reasonable.

• ("s1" "s2"): I don't know what the orthodoxy says, but I prefer an explicit + in between.

• list(): Why not []?

• wheezy_packages.extend([pkg["Package"] for pkg in repo]) -> wheezy_packages.extend(pkg["Package"] for pkg in repo).

• repo = a, then repo = f(repo). IMHO this is as common a pattern as is undesirable. Different values deserve different variable names (key point of functional programming).

• Function too long: I'd split the code in at least two auxiliary functions, one to get wheezy_packages and another to run the commands.

• tshepang: list() was inspired by stackoverflow.com/a/2745292
• tshepang: Is the explicit + in this case common practice?
• tshepang: Regarding the different values comment, see codereview.stackexchange.com/a/1719.
• tokland: 1) list() A.M. is certainly a reference to follow. 2) + I prefer to write it, but not really sure what's better/more idiomatic. 3) I strongly disagree with that answer. In that example I'd use message_pattern and pattern, as I said in the answer, different values, different names. Do you know some Ruby. Those are my ideas about FP (99% can be applied to Python): slideshare.net/tokland/functional-programming-with-ruby-9975242
• tshepang: Am not really familiar with Ruby.