insights while writing a plugin

[samuelstroschein]

• a plugin doesn't import a bundle but messages
  • bundle alias is nowhere used
• optimizing a serialized format that can be hand edited is a lost cause
  • third party tooling (inlang ecosystem) will make hand editing redundant anyways
  • git merge conflicts will happen because everything is on one line but will become redundant with the lix host
  • the inlang message format is redundant after the lix host. the .inlang file will eventually become the exchange format (becuase it's a file :P )
• declaration = input | variable if true:
  • then variable reference is wrong. it should be declaration ref.
  • inputs can't be expressions because inputs can't go through a function
  • splitting declarations from inputs makes sense IMO
  • message format stopped matching on expressions. they match on variables now (but what are variables? declarations?) Match on variables instead of expressions unicode-org/message-format-wg#877

[samuelstroschein]

@martin.lysk1 @nils.jacobsen can we sync tomorrow? my brain melted away. either i am not understanding something, our data model doesn't make sense, or message format 2.0 is broken

https://www.loom.com/share/a4a7a1a386df405aa3c7f186635bf343

[martin-lysk]

Back from the rabbit hole.

Seems like we are not the only ones who are confunsed:

I really suggest looking into this discussion: unicode-org/message-format-wg#819

TL:DR;

• Values in ICU2 are immutable
• Functions do not mutate a value but store annotation to a referenced value
• The Function is (in some implementation) evaluated lazy - last annotation wins

For a deeper dive see loom:
https://www.loom.com/share/b9e105354fb74e4396f22b0d05b0dfba?sid=3dc609c9-525b-4fd1-b1eb-eaebe72b6cfb

[samuelstroschein]

I will do research on Monday with an end to end import -> crud -> export flow

[samuelstroschein]

succeded by MESDK-239