Wrong version reported for Ruby

[voxik]

This is mostly clone of https://pagure.io/mdapi/issue/115

I was complaining, that the-new-hotness does not report new versions available, but the analysis discovered, that the mdapi is to blame, because it reports wrong version. This is the latest Ruby build:

https://koji.fedoraproject.org/koji/buildinfo?buildID=2093827

Apparently, it is ruby-3.1.3-172.fc38, but the mdapi reports version 6.4.0. There is actually rubygem-rdoc-6.4.0-172.fc38.noarch.rpm subpackage of Ruby package, but I have no idea why is this version represented as the Ruby version.

[voxik]

Analyzing the issue, the problem seems to be here:

mdapi/mdapi/services/__init__.py

Lines 77 to 85 in e9d4439

	async with dtbsobjc.execute(GET_PACKAGE_BY_SRC, ("%s-%%" % srcn,)) as dbcursor:
	pkgc = await dbcursor.fetchall()
	if pkgc:
	srcn = re.escape(srcn)
	ptrn = re.compile("%s-[0-9]" % srcn)
	for pkgx in pkgc:
	if ptrn.match(pkgx[3]):
	pckg = Packages(*pkgx)
	break

The GET_PACKAGE_BY_SRC returns somethig like this:

sqlite> SELECT pkgKey,pkgId,name,rpm_sourcerpm,epoch,version,release,arch,summary FROM packages WHERE rpm_sourcerpm LIKE 'ruby-%' ORDER BY epoch DESC, version DESC, release DESC;
pkgKey|pkgId|name|rpm_sourcerpm|epoch|version|release|arch|summary
38788|50f352b42e9c829c7ba7a8401fae72cdbd7ef2c46fe4518b59973c473d13f0ab|rubygem-rdoc|ruby-3.1.3-172.fc38.src.rpm|0|6.4.0|172.fc38|noarch|A tool to generate HTML and command-line documentation for Ruby projects
38565|f32ba51918ff49a02b2a121f699deb77fa7aa7bddae210d2c13c96d5af21dc7b|rubygem-minitest|ruby-3.1.3-172.fc38.src.rpm|0|5.15.0|172.fc38|noarch|Minitest provides a complete suite of testing facilities
38726|de9d4c528526558afb30053f712bd476d6703819e296eca15fb141db3b881c8c|rubygem-psych|ruby-3.1.3-172.fc38.src.rpm|0|4.0.4|172.fc38|x86_64|A libyaml wrapper for Ruby
38970|2974c671192386b3d0704862a58f88d3de645227a1dff1790e465419121228f8|rubygem-test-unit|ruby-3.1.3-172.fc38.src.rpm|0|3.5.3|172.fc38|noarch|An xUnit family unit testing framework for Ruby
39074|82386a71740e58baba01bf949c760146d8e74c420ac5322fda605f7552605be9|rubygems|ruby-3.1.3-172.fc38.src.rpm|0|3.3.26|172.fc38|noarch|The Ruby standard for packaging ruby libraries
39075|20119b07d66b912f68a191846258f1bb5fcac06286d610589e07dd49c60cf78e|rubygems-devel|ruby-3.1.3-172.fc38.src.rpm|0|3.3.26|172.fc38|noarch|Macros and development tools for packaging RubyGems
38811|132c918ff67d62c21ae2a88fdf43c55b419541391a8bd2dee242c043e3e79add|rubygem-rexml|ruby-3.1.3-172.fc38.src.rpm|0|3.2.5|172.fc38|noarch|An XML toolkit for Ruby
38023|7c82d3aecfc90fede9c52139886b653185bef8509707d1fe3db02142204b63dc|ruby|ruby-3.1.3-172.fc38.src.rpm|0|3.1.3|172.fc38|x86_64|An interpreter of object-oriented scripting language

... snip ...

38067|b578216f79903bce787b6d609a5180f989de80baf8999faf4ca48d4b95d097e3|ruby-romkan|ruby-romkan-0.4-27.fc37.src.rpm|0|0.4|27.fc37|noarch|Romaji <-> Kana conversion library for Ruby
39013|f083a8acc3dfbbff9232e8c0302146a3eae775205c7ecdb7793fe342cb09f850|rubygem-typeprof|ruby-3.1.3-172.fc38.src.rpm|0|0.21.3|172.fc38|noarch|TypeProf is a type analysis tool for Ruby code based on abstract interpretation
38842|3c08549df9601010b3204cd2e832d5f138f7638e2375ed02543c35bde01a34fa|rubygem-rss|ruby-3.1.3-172.fc38.src.rpm|0|0.2.9|172.fc38|noarch|Family of libraries that support various formats of XML "feeds"
38051|8a6464d83afd226369531f2a3c359a48a9231ca2994a37ad69143997ec89a721|ruby-icon-artist|ruby-icon-artist-0.1.92-24.fc37.src.rpm|0|0.1.92|24.fc37|noarch|Supporting libraries for icon artists

This is later filtered by:

mdapi/mdapi/services/__init__.py

Line 81 in e9d4439

ptrn = re.compile("%s-[0-9]" % srcn)

I.e. the rpm_sourcerpm has to match srcn-[0-9], which still matches the first item and that is the rubygem-rdoc. While this is certainly better then prior 1d7d14d, it is still not enough. I thing that if there is row, where name == srcn, that line should be preferred. Of course there might not be such line, if I am not mistaken, this is common for python packages. So there still should be some fallback.