The idea is to make everything pipeline based. If we're clever about nodes and datatypes passed around, it should suffice for most conlanging needs. The pipedream - pun intended - is to eventually integrate drawio into it and just draw pipelines. Here's a quick mockup of what I was thinking:
Though initially pipes will be text based. Rest of the wishlist is:
- Import and export to and from most common formats for storing lexicons, incl. spreadsheets and databases.
- Be as generic as possible: The program won't dictate structure of lexicon entries at all. So users will have to help themselves in this regard, incl. indicating which column contains the conlang lexical items.
- Providing a builtin generic SCA engine, but also allowing for external SCA plugins.
- Chaining together multiple pipes in complex logic, so that you can derive lexicons through multiple diachronic stages. You could also have dedicated pipes to derive e.g. pronouns or inflection.
- Eventually supporting some generic builtin vocab and inflection preview.
- Nodes would include: Data input, Data output, Filter, SCA, Inflection (essentially an SCA that applies a table of changes over every input) and Overwrite (for irregular developments, folk etymologies, etc.)
- Of course it'll have worker threads, with each thread executing the pipe or even a single node for a single lexical entry. We mean business. Also add some kind of real-time profiling (avg node execution times, progress bars, etc)
I don't know. Maybe I'll start doing the skeleton of it in the coming weeks. I've been also itching to learn a sexy new language, but that would pose additional difficulties, and using 3rd party libraries and projects may be an uphill battle for anything other than the few well-supported languages.
