Canonicalize and Group Stack Traces

Problem

You’re on-call for a fintech payments platform processing millions of card authorizations per day. A React checkout page occasionally crashes, and your error pipeline receives huge volumes of stack traces from different CDNs and builds (cache-busting query strings). To triage quickly, you need to deduplicate traces that are “the same” after normalization.

Each stack trace is a list of frame strings in this format:

<function>@<url>:<line>:<column>

Example frame: render@https://cdn.app.com/static/js/main.9f3a.js?build=123:120:9

Canonicalization rules

For each frame, convert it to:

<function>@<file_name>:<line>

Where:

1. Ignore the host and path; keep only the last path segment (file_name).
2. Remove any query string (anything after ?) before extracting the file name.
3. Keep the line number.
4. Ignore the column number.
5. Within a single trace, collapse consecutive duplicate canonical frames (e.g., A, A, A, B → A, B).

Task

Implement:

def group_stack_traces(traces: list[list[str]]) -> dict[str, int]:

Return a dictionary mapping a stringified tuple of canonical frames to the number of traces that normalize to that exact canonical sequence.

Note: The key must be the Python str() of the tuple (e.g., "('a@app.js:1', 'b@app.js:2')") to match the evaluation harness.

Constraints

• 1 <= len(traces) <= 2 * 10^4
• 1 <= len(trace) <= 200
• sum(len(trace) for trace in traces) <= 2 * 10^6
• Each frame contains exactly one '@' and at least two ':' after the URL