Open reynoldsnlp opened 3 years ago
I made a suggestion based on what I used in teahcing few years ago and the existing stuff for disambiguator.cg3
, I don't know if the other cg comes from shared so I didn't dare to touch yet: see here:
https://github.com/giellalt/template-lang-und/commit/778e2ac6acca3ee07d2075d2b8c8ddd06fd92b9f
I think lexcies and twolc are still quite simple they seem like what I made originally most of them.., is there anything specific to improve there though?
I agree with the original suggestion, i.e. paring down the generated LEXC files for stems and affixes to only ones that one might expect for a majority of languages (e.g. nouns, verbs, particles) and which exemplify some key LEXC design principles - to list the ones that come first to mind:
pos_affixes.lexc
and pos_stems.lexc
, ornon-standard-forms.lexc
In addition one could explicitly include the LEXC files that are part of shared set up (punctuation, symbols, proper names).
It is somewhat weird to have lots of LEXC files hanging around that one doesn't use and are perhaps excluded by being commented out in root.lexc
, e.g. for adjectives and so on.
When initializing new repositories, it is great to have example files to show the way, but some of the initialized files are surprisingly complex. This makes it so that a linguist starting on a new repository has to de-clutter before she can actually start building.
lexc
,twolc
, andcg3
files down to trivial toy examples, so that new projects only have to change/delete a few lines to get started with a new language. Anything that's important to include for explanatory purposes could be in comments, so it has no effect on how the project builds.