Count sorted groups and flush the last run

Result = []
CurrentKey = empty; Count = 0
foreach X in RecordsSortedByKey:
    if CurrentKey is not empty AND key(X) != CurrentKey:
        Result = Result ++ [[CurrentKey,Count]]
        Count = 0
    CurrentKey = key(X)
    Count = Count + 1
if CurrentKey is not empty:
    Result = Result ++ [[CurrentKey,Count]]

Sorting gathers equal keys into runs. Keep the current key and a count. On a key change, save the previous group, reset the count and begin the new group. Save the final run after the loop because no following change will trigger it. An empty input produces [].

Actual city counts include Chennai 8, Bengaluru 5, Madurai 5, Trichy 3 and Vellore 3. Singleton cities include Ambur, Nagercoil, Salem and Theni. The procedure works because sorting grouped equal cities; it does not require a geographic ordering.

Change the unit before changing the key #

The source applies the same idea to item-level [Customer,Shop,Category] triples. Sorting those by customer counts item rows per customer; sorting by shop counts item rows per shop. Our complete 30-record bill-summary lab counts BILLS instead: Srivatsan 4 and Ahmed 4, for example. Bill counts, item-row counts and purchased quantities are different units. Full item counting requires original item rows, not category summaries.