Replace Values and Return Sums

You are given an integer array entries and a 2D integer array transactions. Each transaction is a pair [oldValue, newValue].
For each transaction, replace every occurrence of oldValue in entries with newValue. After applying the transaction, record the sum of all values in entries.
Return an array containing the recorded sums after each transaction.
The sums may exceed the range of a 32-bit integer.

Function
replaceValuesAndReturnSums(entries: int[], transactions: int[][]) → long[]

Examples
Example 1
entries = [1, 2, 1, 3]
transactions = [[1, 4], [2, 1], [4, 2]]
return = [13, 12, 8]
After replacing 1 with 4, the array is [4,2,4,3] with sum 13. After replacing 2 with 1, the sum is 12. After replacing 4 with 2, the sum is 8.
Example 2
entries = [5, 5]
transactions = [[1, 2], [5, 1]]
return = [10, 2]
The first transaction has no matching value, so the sum stays 10. The second transaction changes both values to 1, so the sum becomes 2.

Constraints
transactions[i].length == 2
The returned sums may require 64-bit integers.

Don't touch the array. Build a hash map of value to count, and compute the initial sum as a 64-bit number. For each transaction [old, new], look up count = map[old]. If it's zero or missing, or old equals new, record the current sum and move on. Otherwise add count * (new - old) to the sum, then add count to map[new] and delete map[old]. Each transaction is O(1), so the total is O(n + m). The classic pitfalls are overflow and the old == new case. Do the multiplication in long, not int. If old equals new and you delete after adding, you wipe out the count and corrupt every later answer. Example 2 shows the no-match case: the sum stays 10. If you freeze on the merge step in the live OA, StealthCoder is the safety net that can hand you the clean version.

If you see this problem in your OA tomorrow, the play is to recognize the pattern in 30 seconds. StealthCoder buys you that recognition.