Version of Lua module is not normalized about l3build

[Udi-Fogiel]

Sure, sorry for not doing already, I just got out of home and did not want to forget to report. I'll add an example when I'll return to my computer.

If you don't want to wait, you can look at reutenauer/polyglossia@ce56af7

[Udi-Fogiel]

OK, let's keep this ticket open until it is addressed somewhere else?

[josephwright]

Can you provide a short demo? This is not something I've seen in our .tlg files.

[Udi-Fogiel]

This is build.lua

bundle = ""
module = "foo"
checkengines = {"luatex"}

In the testfiles directory a test with

\input{regression-test.tex}
\START
\directlua{require'foo'}
\END

and in the testfiles/support there is a file foo.lua which contains

local module = {
    name          = "foo",
    version       = "this is the version",
    date          = "2024/05/09"
}

luatexbase.provides_module(module)

the .tlg is

This is a generated file for the l3build validation system.
Don't change this file in any respect.
Lua module: foo ....-..-.. this is the version

[josephwright]

We already normalise things that 'look like' version strings: this isn't dependent on being from a Lua module. The current patterns cover classical LaTeX-like version strings (vX.Ya) and semantic versioning like (vX.Y.Z). The issue with the one you've pointed to is there's no indicator it's a version string: no leading v :(

[Udi-Fogiel]

Is it more appropriate that I'll ask from the fontspec maintainer to use version tags that fall under this scheme?

[josephwright]

@Udi-Fogiel I wonder if this would be better addressed in ltluatex? It would be easy to check if the version string starts with v and print it if not, so that there is consistency in how they are logged.

[muzimuzhi]

In the LaTeX2e module ltluatex.dtx, currently the optional version field is loosely documented as a (lua) string and it happens both fontspec and polyglossia provide the version without leading v.

Maybe the luatexbase.provides_module() in ltluatex.dtx can add a leading v if version doesn't start with it.

Implementation of luatexbase.provides_module(), LaTeX2e 2023-11-01 Patch level 1

https://github.com/latex3/latex2e/blob/7cac63f1fa933a4bc643d934feca3407133ab698/base/ltluatex.dtx#L1011-L1026

local function provides_module(info)
  if not (info and info.name) then
    luatexbase_error("Missing module name for provides_module")
  end
  local function spaced(text)
    return text and (" " .. text) or ""
  end
  luatexbase_log(
    "Lua module: " .. info.name
      .. spaced(info.date)
      .. spaced(info.version)
      .. spaced(info.description)
  )
  modules[info.name] = info
end
luatexbase.provides_module = provides_module

[josephwright]

@muzimuzhi Yes, I was thinking about something as simple as

spaced(string.gsub(info.version,"^(%d)","v%1"))

[Udi-Fogiel]

currently the optional version field is loosely documented as a (lua) string and it happens both fontspec and polyglossia provide the version without leading v.

Yes, the version of polyglossia.lua was not updated since 2013/05/11, which was probably before this convention, and probably should be changed :)

I addressed this in latex3/latex2e#1364, so I'm closing here.
Thanks for the help!