Reading map: correct the measured no-op rate for markQuoted

0.9% (2 of 230), not the ~7% estimated before the endpoint was restricted to
indexed books — that filter, added for the byline fix, removed nearly every
orphan as a side effect.
This commit is contained in:
wes
2026-08-14 08:47:14 -04:00
parent bd6910a5dd
commit 3039ff45ed
+5 -4
View File
@@ -408,10 +408,11 @@ export function build(container, data) {
* Ring the node a quoted passage came from. `textHash` is the `text_hash` that
* api.c0smere.net/highlights/random returns beside the quote.
*
* Returns false when that passage has no node — about 1 highlight in 15, because its
* source book has never been indexed into library-rag and so was never placed. That is a
* normal outcome, not an error: the caller does nothing and the map simply shows no ring.
* The set shrinks as those sources get indexed.
* Returns false when that passage has no node. Measured at 2 of the 230 quotes the header
* can draw (0.9%) — the endpoint already restricts itself to books library-rag has
* indexed, so the only survivors are passages the export could not confirm inside the
* chunk they claim. A normal outcome, not an error: the caller does nothing and no ring
* is drawn.
*/
markQuoted(textHash) {
const node = nodeOfHash.get(textHash);