why do we skip printing all multi index results?
Is there a test for multi value?

If my understanding of the existing C code is correct, it does currently avoid to emit the same doc id multiple time.
It goes:

QueryIterator *iter = NewInvIndIterator_NumericQuery(range->entries, sctx, &fieldCtx, NULL, NULL, range->minVal, range->maxVal);
ReplyIteratorResultsIDs(iter, sctx->redisCtx);

which in turn invokes:

InitInvIndIterator(&numIt->base, INV_IDX_NUMERIC_ITERATOR, idx, NewNumericResult(), fieldCtx, sctx, &decoderCtx, NumericCheckAbort);

which in turn uses this helper to decide if it should skip multi-values:

// Returns true if the iterator should skip multi-values from the same document
static inline bool ShouldSkipMulti(const InvIndIterator *it) {
  return IndexReader_HasMulti(it->reader); // The index holds multi-values (if not, no need to check)
}

which basically always evaluates to "skip multivalues in read, just don't bother checking if there are no multivalues".

I guess the reasoning goes as follows: we are only returning the doc ids, without their values, so what's the point of printing the same doc id N times?
I'm happy to change it, but I assume this will be a user-visible change if the debug commands are exposed?

Changed it to include all ids.

consider moving to a general inverted index debug file. it essentially implements
debug_commands.c

Converted this into a method on the Summary itself.

This function as well can be move to inverted index debug file

That was my first instinct too, but this is actually specific to a numeric index since it makes two specific assumptions:

• All entries are numeric
• And the index is tracking the number of entries explicitly (i.e. it uses our EntriesTrackingIndex wrapper around the "bare" index)

It is not used for anything else, and in fact, its C counterpart gets deleted in #8306.

so lets rename it

Changed to reply_numeric_index_debug 👍🏻

Multi-value doc IDs returned repeatedly

Medium Severity

debug_dump_index emits IDs via range.reader() in reply_range_entries, which returns every numeric entry. Unlike the C DUMP_NUMIDX path, this bypasses iterator logic that skips duplicate IDs for multi-value docs. Multi-valued documents are therefore repeated, producing a different command output shape and inconsistent counts against numDocs.

so lets rename it