TermBuf::WidthProbe::Result
Everything the probe learned.
Constructors
Instance methods
Whether the terminal answered at all. When it did not, policy is whatever was passed in.
Samples the resulting policy still measures differently from the terminal. Terminals reach conclusions this design has no flag for, and naming what is left over beats modelling it wrong.
Keystrokes that arrived while the probe was running.
Whether the terminal counts a cluster's columns by adding its code
points up, or nil when it did not answer the sample that tells the
two apart.
The samples are already the right question. One of them has a per-code-point count that no policy can reach — four faces joined by zero width joiners own eleven that way, against two collapsed and eight laid out — so its answer names the terminal's bookkeeping outright. Measuring beats matching on the terminal's name, which can neither know about a version that fixed it nor about a terminal nobody has tried.