codeguru_profiler_agent/profiler.py (4 lines): - line 29: # TODO: Review the memory limit - line 36: # TODO FIXME Consider making this work for Windows. - line 170: # TODO delay metadata lookup until we need it - line 198: TODO Currently we can start multiple instances of Profiler which we want to avoid. Implement a singleton codeguru_profiler_agent/aws_lambda/lambda_handler.py (2 lines): - line 37: # TODO FIXME Review if the support for python 3.6 bootstrap can be improved. - line 43: invokeid='unknown_id') # FIXME invokeid is used for error handling, need to see if we can get it codeguru_profiler_agent/model/call_graph_node.py (2 lines): - line 74: # FIXME: Review use of tuple vs list vs dictionary, and linear search vs binary search - line 83: FIXME: We still need to review the memory vs cpu tradeoffs of using a tuple vs an list vs a dictionary here, codeguru_profiler_agent/profiler_builder.py (1 line): - line 12: # TODO: Should we move these logic to Profiler instead? codeguru_profiler_agent/sdk_reporter/sdk_reporter.py (1 line): - line 34: # TODO decide if we need to compress with gzip or not.