Issue with parsing apk list --installed

[shomodj]

Describe the bug

When parsing installed alpine linux packages, the parser is skipping some packages due to the incomplete regex which results in pyinfra constantly trying to install a package that is already installed on the system

To Reproduce

Attached is the list of all the weird versions I quickly found on my system, note the openssh-server-9.3_p2-r0 x86_64 or syslinux-6.04_pre1-r13 x86_64 and less-633-r0

Expected behavior

Returned list of packages in the APK Fact class should return the same number of packages as apk list -I

Meta

Some weird examples

acct-6.6.4-r2 x86_64 {acct} (GPL-3.0-or-later) [installed]
alpine-keys-2.4-r1 x86_64 {alpine-keys} (MIT) [installed]
argon2-libs-20190702-r4 x86_64 {argon2} (Apache-2.0 OR CC0-1.0) [installed]
busybox-openrc-1.36.1-r5 x86_64 {busybox} (GPL-2.0-only) [installed]
ca-certificates-20230506-r0 x86_64 {ca-certificates} (MPL-2.0 AND MIT) [installed]
ca-certificates-bundle-20230506-r0 x86_64 {ca-certificates} (MPL-2.0 AND MIT) [installed]
chrony-4.3-r4 x86_64 {chrony} (GPL-2.0-or-later) [installed]
chrony-openrc-4.3-r4 x86_64 {chrony} (GPL-2.0-or-later) [installed]
coreutils-9.3-r1 x86_64 {coreutils} (GPL-3.0-or-later) [installed]
device-mapper-event-libs-2.03.21-r3 x86_64 {lvm2} (GPL-2.0-or-later AND LGPL-2.1-or-later AND BSD-2-Clause) [installed]
kmod-30-r3 x86_64 {kmod} (GPL-2.0-or-later) [installed]
kmod-libs-30-r3 x86_64 {kmod} (LGPL-2.1-or-later) [installed]
less-633-r0 x86_64 {less} (GPL-3.0-or-later OR BSD-2-Clause) [installed]
libedit-20221030.3.1-r1 x86_64 {libedit} (BSD-3-Clause) [installed]
libgcc-12.2.1_git20220924-r10 x86_64 {gcc} (GPL-2.0-or-later AND LGPL-2.1-or-later) [installed]
libncursesw-6.4_p20230506-r0 x86_64 {ncurses} (X11) [installed]
libpanelw-6.4_p20230506-r0 x86_64 {ncurses} (X11) [installed]
libstdc++-12.2.1_git20220924-r10 x86_64 {gcc} (GPL-2.0-or-later AND LGPL-2.1-or-later) [installed]
linux-firmware-none-20230515-r6 x86_64 {linux-firmware} (custom) [installed]
ncurses-terminfo-base-6.4_p20230506-r0 x86_64 {ncurses} (X11) [installed]
openssh-client-common-9.3_p2-r0 x86_64 {openssh} (BSD) [installed]
openssh-server-9.3_p2-r0 x86_64 {openssh} (BSD) [installed]
py3-certifi-2023.7.22-r0 x86_64 {py3-certifi} (MPL-2.0) [installed]
py3-certifi-pyc-2023.7.22-r0 x86_64 {py3-certifi} (MPL-2.0) [installed]
salt-3006.1-r0 x86_64 {salt} (Apache-2.0) [installed]
salt-minion-3006.1-r0 x86_64 {salt} (Apache-2.0) [installed]
syslinux-6.04_pre1-r13 x86_64 {syslinux} (GPL-2.0-or-later) [installed]
utmps-libs-0.1.2.1-r1 x86_64 {utmps} (ISC) [installed]

This is the best I could do, I don't think its even humanly possible to write a correct regex for this, that is why I split it into parts and searched for part which begins with a number in order to find where name and version were split

import re

with open("list.txt") as fh:
    lines = fh.readlines()

    packages = dict()

    for line in lines:
        line = line.strip()
        columns = line.split(" ")
        name_and_version = columns[0]
        parts = name_and_version.split("-")

        idx = 1
        while idx < len(parts):
            if re.match(r"^\d", parts[idx]):
                name = "-".join(parts[0:idx])
                version = "-".join(parts[idx:])
                packages[name] = version

            idx += 1

    print(packages)