push.cx
+6
-13
README.md
+6
-13
README.md
···
13
13
It works and has reasonable tests, including passing the TypeID suite.
14
14
It's probably not idiomatic Lua.
15
15
16
+
I've written up some notes on the implementation and alternatives in the [announcement blog post](https://push.cx/typeid-in-lua).
17
+
16
18
17
19
# Use
18
20
···
62
64
47 successes / 0 failures / 0 errors / 0 pending : 0.019008 seconds
63
65
```
64
66
65
-
66
-
# Ideas
67
+
# Likely Issues
67
68
68
-
* I experimented with style.
69
-
`TypeID` is more OO style, `Base32` and `UUID7` work on primitives.
70
-
After implementing, I guess users would probably prefer getting a primitive back, despite the lack of types.
71
-
* The nested conversions in the metatable `uuid` function suggests the internal representation of `suffix` is wrong,
72
-
but the syntactic distinction between `.field` and `:method()` means duplicating the data into two fields, exposing the internal representation by it being a field and the other a method, or getting away from what seems like common struct-y style and making both methods.
73
-
I don't really know what level to aim at between "nice consistent high-level interface" and "damn everything, all primitives and seams showing for perf".
74
69
* The rockspec probably does not correctly depend on `sys/time.h`.
75
-
* I ported Base32 from the official TypeID Golang implementation then wrote UUID7 in bytes to match it,
76
-
but all that intermediate bit twiddling could be simplified by generating a UUID7 directly into the Base32 encoding.
77
-
* Maybe I'm looking under the wrong name, but it seems odd there isn't a bitfield type I could use, given Lua's popularity in games.
78
-
Some searching turned up [a library](https://github.com/JohnHind/Lua_Bitfield) but the absence of multi-bit operations seems inconvenient.
79
-
* Spec says "Implementations SHOULD allow encoding/decoding of other UUID variants" but does not mandate validating them, so I don't.
70
+
* I don't understand why `luarocks` generated the version `0.1-1`, what is the `-1` for?
71
+
Is my version `0.1` or `0.1-1`?
72
+
* The TypeID spec says "Implementations SHOULD allow encoding/decoding of other UUID variants" but does not mandate validating them, so I don't.
80
73
Perhaps the spec should say implementations should not permit any old 128 bits.