Time-Based Key-Value Map With Floor and Ceiling Queries

Process operations on a time-based key-value map. SET stores or replaces a value for a key at a timestamp and contributes null to the result. GET contributes the value at the greatest stored timestamp less than or equal to the query timestamp. CEIL contributes the value at the smallest stored timestamp greater than or equal to the query timestamp. A missing answer contributes the empty string.
SET timestamps may arrive out of order. When the same key and timestamp is set again, the latest value replaces the earlier one. Process operations in array order and return one result per operation.

Function
runVersionedMap(operations: String[], keys: String[], values: String[], timestamps: int[]) → String[]

Examples
Example 1
operations = ["SET","SET","GET","CEIL"]
keys = ["a","a","a","a"]
values = ["late","early","",""]
timestamps = [10,2,7,7]
return = ["null","null","early","late"]
The out-of-order writes are sorted by timestamp for both floor and ceiling lookup.
Example 2
operations = ["SET","SET","GET","CEIL","SET","GET"]
keys = ["x","x","x","x","x","x"]
values = ["old","new","","","replacement",""]
timestamps = [5,9,4,6,5,5]
return = ["null","null","","new","null","replacement"]
Missing floor lookup is empty, ceiling finds timestamp 9, and a repeated timestamp is overwritten.

Constraints
1 <= operations.length <= 100000
All four input arrays have equal length.
Each operation is SET, GET, or CEIL.
1 <= timestamps[i] <= 10^9
Keys and SET values are nonempty strings of length at most 100.

Map each key to a sorted structure of timestamps and values. GET is a floor lookup, CEIL is a ceiling lookup, and both are binary search on that sorted structure. The catch is that SET arrives out of order and can overwrite an existing timestamp. A plain array with append breaks, because you'd need to insert in the middle, which is O(n) per insert. In Java or C++, use a TreeMap or std::map per key with floorKey and ceilingKey or lower_bound. In Python, no built-in sorted map exists, so use bisect with insort and a parallel dict for values. Check for an existing timestamp first, then overwrite without inserting a duplicate. Return the string null for SET and the empty string for misses. Pitfall: mixing those two up. If the structure feels heavy and you freeze live, StealthCoder is the hedge that gets you a working version.

If this hits your live OA and you blank, StealthCoder solves it in seconds, invisible to the proctor.