Correct Pagination Ordering

Ensure algorithmic results (ordering, counts, and pagination) remain correct after filtering/slicing and that sorting/dedup logic is both efficient and deterministic.

Practical rules: 1) Order/recency claims must match the actual data returned

• If you only fetch a partial subset (e.g., MaxKeys=10), you cannot claim “newest-first” unless the subset is actually chosen by recency.
• Either change the fetch strategy or adjust the output wording to avoid misleading semantics.

2) Returned totals must reflect what the caller truly receives

• If you fetch extra results and then filter/slice client-side, recompute the effective total based on the post-processed result set (e.g., return len(docs) rather than a pre-filter server count).

3) Dedup must be hash-based and collision-safe

• Avoid result_copy not in results style membership checks (quadratic + projection collisions).
• Track seen identifiers in a set (prefer stable IDs like slug).

4) Sorting keys must define a total order

• Never return mixed-type sort keys (int vs str) from the same comparator.
• Use a tuple-based key with a type/group discriminator.

Example patterns:

Dedup with set:

limit = 10
seen_slugs: set[str] = set()
results = []

for m in all_matches:
    slug = m.get('slug', '')
    if slug in seen_slugs:
        continue
    seen_slugs.add(slug)
    results.append({k: v for k, v in m.items() if k != 'tags'})
    if len(results) >= limit:
        break

Total-order sort key (mixed numeric/non-numeric):

def line_key(k: str):
    return (0, int(k), '') if k.isdigit() else (1, 0, k)

sorted_line_numbers = sorted(line_numbers, key=line_key)