Use Semver in Dependency Constants

[uniboi]

If a mod for a dependency constant is loaded, this will create a sqvm constant that's a SQTable holding the dependency's major, minor and patch version as integer. All other subversions are omitted.
If a dependency is not loaded, the constant will be 0 instead.

This does not break compatability with old versions since a table is a truthy value like 1, as it is now.

This allows modders to compile code for ranges of mod versions.

For example:

#if HAS_UEBERBLICK
// This compiles only if the dependency is loaded. No matter the version of the dependency
#endif

#if HAS_UEBERBLICK && HAS_UEBERBLICK.major == 1
// This compiles ONLY if the dependency version is 1.x.x
#endif

#if HAS_UEBEBLICK && HAS_UEBERBLICK.major == 1 && HAS_UEBERBLICK.minor >= 2 && HAS_UEBERBLICK.minor <= 5
// This compiles only if the dependency's version is between 1.2.x - 1.5.x
#endif

Invalid version numbers are represented as -1.

• "1.2.test"
  major: 1
  minor: 2
  patch: -1

• "1."
  major: 1
  minor: -1
  patch: -1

• "heheheha"
  major: -1
  minor: -1
  patch: -1

etc.

[GeckoEidechse]

I assume v1.12.1-rc would then be:

major: 1
minor: 12
patch: 1

?

[uniboi]

v1.12.1-rc would be

mj: -1
mn: 12
pt: -1

since I did not bother writing an semver parser because
a) I have never seen a mod with a leading v in the version
b) I don't think anyone is playing on release candidates

[GeckoEidechse]

v1.12.1-rc would be

mj: -1
mn: 12
pt: -1

since I did not bother writing an semver parser because
a) I have never seen a mod with a leading v in the version

the leading v was a force of habit, forgot that it's not in the mods version

b) I don't think anyone is playing on release candidates

sadly cause we really need more playtesters x_x

[uniboi]

I could just terminate on a -, I guess. parsing rc subversion doesn't make sense imo

[GeckoEidechse]

I could just terminate on a -, I guess. parsing rc subversion doesn't make sense imo

Yeah that's fine. Was just wondering whether it would be a -1 or whether the -rcX is stripped and turned into just the 1 for patch ^^

[uniboi]

Ignores now anything after a -

[BobTheBob9]

genuinely surprised any kind of table stuff works in #ifs?

[uniboi]

Conditions for #if are normal squirrel expressions, they just need to be able to be evaluated at compile time (ie be constant).

Even stuff like this works #if { k = 1 }.k == int(CLIENT), even if unnecessary

[uniboi]

fixed again.
dep_test.zip
To test: Install both mods in the zip. In the mp lobby, the globals a = 1, b = 2 and c = 3 should be defined while d = 4 isn't.

[EladNLG]

Small request:

Make the version available as a string as well :)

[uniboi]

Make the version available as a string as well :)

you can now do checks like #if CONSTANT.version == "1.2.3" #elseif CONSTANT.version == "1.2.3-rc1" #endif

[ScureX]

very good i like so hot and sexy mmhmmm nice code yes

[catornot]

tested the pr; works well except it crashed once ( used the last ci build btw )

used a script command to test this : script printt(HAS_DEP_TEST_2.version, "major:",HAS_DEP_TEST_2.major,"minor:",HAS_DEP_TEST_2.major,"patch:",HAS_DEP_TEST_2.patch)

with dep_test_2

1.2.3 major: 1 minor: 1 patch: 3

without dep_test_2

[NORTHSTAR] [info] sq_compilebuffer returned SQRESULT_NULL
[NORTHSTAR] [info] sq_call returned SQRESULT_ERROR

with rc for patch

1.2.3-rc major: 1 minor: 1 patch: 3

another note it also works by just using #if DEP :)

nslog2023-03-20 19-28-52.txt

[uniboi]

It looks like the game crashes when accessing a dependency constant table and then recompiling scripts

[ASpoonPlaysGames]

@uniboi is this still going to be worked on? Or has the crashing discouraged you?

[uniboi]

I think it might just be a refcount issue, I haven't tested any further though.
Might try soon:tm:

[GeckoEidechse]

@uniboi heads-up that there are merge conflicts here ^^

[uniboi]

fixed in #685