Function describeStatement

Returns {
    directive: boolean;
    run: RunToken[];
    runEnds: (number | null)[];
    whole: boolean;
}

• directive: boolean

• run: RunToken[]

• runEnds: (number | null)[]

  For each run token, the offset in text just past the WORD it came from — where the arguments of a command whose verb is that token begin (#202c) — or null where that offset would not mean that.

  Per RUN TOKEN rather than one run-wide offset, because the run does not stop at the verb: runTokens consumes every leading bare word, so /interface print detail puts detail IN the run. Once the boundary rule has decided the verb, everything after it is by construction an argument — measured at 60% of the corpus's CRUD-verb commands, so ending the argument list at the run instead of at the verb would abstain on most of them.

  Null in two cases, both fail-closed:

  • the token is not the LAST part of its word (/ip/route/add/x splitting its verb mid-word): the following parts are neither path nor arguments of this reading;
  • the word walk stopped before reaching that token (the run and the words disagree), so no offset was established for it at all.

  A word END is the only offset that survives a \<newline> continuation — the joined bytes are interior to the word (see asciiWordSpans), which is why no per-token START is published.

• whole: boolean