skip() and limit()

`skip()` and `limit()`

For any query q, q.skip(...) and q.limit(...) build queries whose output is like q's output, but cropped at the beginning (skip) or at the end (limit). Together they are wrappers for SQL's LIMIT and OFFSET.

For any query q, and any value n of type uint32_t, q.skip(n) and q.limit(n) return querys with the following characteristics.

Their value types are the same as q's
Their value mappers are identical to q's.
Their output consists of a contiguous subsequence of q's output. More specifically:
- q.skip(n) produces all the records that q would produce after the first n.
- q.limit(n) produces all the records that q would produce, up to a maximum of n.

It would be a mistake to read skip as a direct translation of SQL's OFFSET. If you add the clause LIMIT 10 OFFSET 8 to some SQL query then, as long as the original query would have produced at least 18 rows, the modified query will produce ten rows, viz. the 9th thru 18th. By contrast, the quince expression q.limit(10).skip(8) produces at most two items ^[11] .

If you want the behaviour of LIMIT 10 OFFSET 8, then put the skip first: q.skip(8).limit(10).

^[11] The motive for this policy is to uphold compositionality. In q.limit(10).skip(8), you can replace q.limit(10) by any other query that produces the same ten items, and overall output will not be affected. But imagine if skip() had the semantics of OFFSET: the output of q.limit(10).skip(8) would depend on records that even q.limit(10) itself does not produce.