Problem

A data team at Acme Analytics receives raw event records as Python dictionaries. Implement a function that cleans these records and returns a per-user summary suitable for analysis.

Given a list of event dictionaries, process each record as follows:

1. Ignore records missing user_id, event, or value.
2. Normalize event by trimming whitespace and converting to lowercase.
3. Keep only records where normalized event is one of "click", "view", or "purchase".
4. Convert value to an integer if possible; otherwise ignore the record.
5. Aggregate valid records by user_id.

For each user, return:

• user_id
• total_value: sum of valid values
• event_counts: counts for each valid event type that appeared

Return the final result as a list of dictionaries sorted by:

1. total_value descending
2. user_id ascending

Input/Output

• Input: records, a list of dictionaries
• Output: a list of summary dictionaries

Examples

Example 1 Input: records = [{"user_id":"u1","event":" Click ","value":"3"},{"user_id":"u1","event":"view","value":2},{"user_id":"u2","event":"purchase","value":"5"}] Output: [{"user_id":"u1","total_value":5,"event_counts":{"click":1,"view":1}},{"user_id":"u2","total_value":5,"event_counts":{"purchase":1}}] Explanation: Both users total 5, so u1 comes first because of ascending user_id.

Example 2 Input: records = [{"user_id":"u1","event":"signup","value":4},{"user_id":"u2","event":"view","value":"bad"}] Output: [] Explanation: Both records are invalid after filtering.

Constraints

• 1 <= len(records) <= 10^4
• Each record is a Python dictionary with string keys
• value may be an integer or a numeric string
• user_id is compared lexicographically