v4 CL-B1b — exact number of listed notes that could not be admitted.
v4 CL-B1b — vault-relative paths of skipped notes, each with the reason.
Capped at 20 entries so a vault full of malformed notes cannot inflate the
response; skipped_note_count stays exact and uncapped.
v3.6.2 HN-1 — count of ALL matching notes in the vault, NOT just the
returned slice. Pre-3.6.2 this was matches.length after the limit
cap, which underreported when more matches existed than limit.
Callers can now reliably tell when a result was truncated by
comparing total_matched > matches.length (or check truncated).
EXACT ONLY IF total_matched_exact is true. Notes the walk listed but
could not admit are skipped rather than rejecting the query, which makes
this a lower bound; an incomplete bounded walk still rejects outright,
because there the unseen remainder is unbounded and unnamed.
v4 CL-B1b — false iff at least one listed note was skipped, i.e.
total_matched is a LOWER BOUND rather than an exact count. Check this
before reporting total_matched as "the number of matching notes".
v3.6.2 HN-1 — true iff total_matched > matches.length (i.e. the
limit capped the response).
Predicates the parser couldn't evaluate (formula calls, linksTo, etc). v3.6.2 HN-2 — under strict mode (the new default) these now exclude the row instead of admitting it. Listed verbatim so callers can see what was REJECTED — empty array = all predicates fully evaluated.
Sub-set of matches (truncated to limit).