def _aggregateOutputs()

in libraries/python/aggregate_classification_results.py [0:0]

• 47 lines of code
• 9 McCabe index (conditional complexity)

    def _aggregateOutputs(self, outputs):
        results = {}
        for one_output in outputs:
            for key in one_output:
                value = one_output[key]
                if key not in results:
                    results[key] = value
                else:
                    results[key]["values"].extend(value["values"])
        pattern = re.compile(r"(\w+)_of_top(\d+)_corrects")
        # finally patch up the summary
        for res in results:
            one_result = results[res]
            one_result["type"] = one_result["type"]
            values = one_result["values"]
            match = pattern.match(one_result["metric"])
            if not match:
                continue
            if match.group(1) == "number":
                data = sum(values)
                one_result["summary"] = {
                    "num_runs": len(values),
                    "p0": data,
                    "p10": data,
                    "p50": data,
                    "p90": data,
                    "p100": data,
                    "mean": data,
                    "std": 0,
                    "MAD": 0,
                }
            elif match.group(1) == "percent":
                data = sum(values) * 100.0 / len(values)
                one_result["summary"] = {
                    "num_runs": len(values),
                    "p0": data,
                    "p10": data,
                    "p50": data,
                    "p90": data,
                    "p100": data,
                    "mean": data,
                    "std": 0,
                    "MAD": 0,
                }
            one_result["metric"] = "total_" + one_result["metric"]
            # there may be too many values, only keep the summary
            if len(values) > 200:
                del one_result["values"]
        return results