[ChiragAgg5k]

QueryDepth and QueryComplexity re-walk a fragment's selection set at every spread. When nested fragments spread each other more than once (F0 { ...F1 ...F1 }, F1 { ...F2 ...F2 }, …), the work doubles per level, so a document of under a kilobyte can keep validation busy for seconds or longer.

Both rules now compute each fragment once per validated document and reuse the result at every spread:

• QueryComplexity caches each fragment's complexity by name. Variables, and therefore @include/@skip and field arguments, are fixed for a validation run, and field definitions are resolved per AST node, so a fragment scores the same wherever it is spread. The cache is reset in getVisitor().
• QueryDepth caches each fragment's depth relative to its spread point and adds the spread depth, which gives the same result as walking it in place. Cycle detection via calculatedFragments is kept; fragments found to be part of a cycle are remembered so repeated spreads into a cycle also stay linear.

Tests in QueryDepthTest and QueryComplexityTest cover a doubling-fragment document up to 50 levels (which would never finish before) with exact depth/complexity, a fragment spread at different depths, @include with repeated spreads across variable changes, and a cycle reached through repeated spreads.

Changelog

Fixed

• Avoid exponential validation time in QueryDepth and QueryComplexity for documents that spread nested fragments repeatedly

[ChiragAgg5k]

Closing; not pursuing this one for now. Thanks!