Discussion
Bonus thread for today...
Text is only native to programming to the extent that we tend to write our programs in plain text. Though early home computers had I/O which natively spoke ASCII, or something extremely close.
I've designed a hypothetical computer which natively processes text!
An entire string string library would be hardwired circuits, or close enough! Including ASCII character classification, Unicode would be implemented in software in part so it can be updated.
1/3?
For text input... Physical keyboards remain the simplest & most ergonomic solution!
Even if converting the pressed keys into ASCII then UTF-8 text requires a conversion pass or 2. The 2nd pass might need some additional UI to show what's happening.
If you're editing text, there's nuances around the arrow keys to get straight.
Though yes, we could support other text-entry mechanisms. For instance we could use machine learning to convert your verbal speech into text, might need an NPU.
2/4?
To output text verbally... My recollection is that we'd need one pass to evaluate the appropriate language's pronunciation rules, & another pass to finalize the emphasis rules. And from there these "phonemes" need to be converted into operations upon a simulated vocal tract.
Our string-centric CPU would then need to offload to a coprocessor for simulating that vocal tract (with a sine wave simulating breathing), which'll produce the actual audio.
3/4!
Outputting text visually is more complex! And graphics in generally would need coprocessors.
I'd use our string-centric CPU to perform some lexing & normalize text-direction. Then our "layout unit" GPU would need the ability to consult lookuptables to perform simplified text-classification. As it determines which "glyphs" it should use to represent the text & where (relatively) it should place them.
Certain languages like Arabic, Thai, or Korean needs special handling.
4/5!!
Once we know which glyphs to render where to depict the given text, we need to actually composite them there! For which I designed a 2ndary GPU.
And typically our fonts store glyphs in a compact & cleanly-resizable vector format, which needs to be converted rasters for output. So throw in a offscreen-rendering pass to produce a "glyph atlas"!
I'm synopsizing *heavily*! And I didn't mention line-wrapping.
P.S. I wound up with LibGettext being practically builtin to the CPU!
5/5 Fin!
Install bonfire.cafe
Get the full app experience
