Weighting static analysis alerts

A system comprising:

• one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
• receiving a request to compute weights for a plurality of alerts generated by a static analysis system, wherein each alert represents a coding defect occurring in one or more software projects analyzed by the static analysis system, and wherein each alert has one or more respective attribute values for one or more respective alert attributes;
• generating a respective sample of alerts for each feature of a plurality of features, each sample comprising alerts occurring in one or more projects and having a particular attribute value for an alert attribute corresponding to the feature;
• computing, for each sample of alerts corresponding to each attribute value and for each feature of a plurality of features, one or more feature values for alerts having the attribute value for the feature;
• obtaining, for each feature, an importance distribution that maps each feature value of one or more feature values for the feature to a respective measure of importance for an alert having the feature value;
• computing, for each alert of the plurality of alerts and for each feature, a respective feature score for the feature using the importance distribution for the feature and the one or more feature values computed for alerts having an attribute value matching an attribute value of the alert;
• and
• computing, for each alert, a respective weight for the alert by combining the plurality of feature scores computed for the alert.

A computer program product, encoded on one or more non-transitory computer storage media, comprising instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:

• receiving a request to compute weights for a plurality of alerts generated by a static analysis system, wherein each alert represents a coding defect occurring in one or more software projects analyzed by the static analysis system, and wherein each alert has one or more respective attribute values for one or more respective alert attributes;
• generating a respective sample of alerts for each feature of a plurality of features, each sample comprising alerts occurring in one or more projects and having a particular attribute value for an alert attribute corresponding to the feature;
• computing, for each sample of alerts corresponding to each attribute value and for each feature of a plurality of features, one or more feature values for alerts having the attribute value for the feature;
• obtaining, for each feature, an importance distribution that maps each feature value of one or more feature values for the feature to a respective measure of importance for an alert having the feature value;
• computing, for each alert of the plurality of alerts and for each feature, a respective feature score for the feature using the importance distribution for the feature and the one or more feature values computed for alerts having an attribute value matching an attribute value of the alert;
• and
• computing, for each alert, a respective weight for the alert by combining the plurality of feature scores computed for the alert.