As soon as a new commit is made, a new commit ID means that the metric backend had no data for that particular version. Figure out how to incorporate old metrics data.
Idea: traverse back from current version ID until we find one which exists in the DB. Use this to fetch and display feedback.
This will work for unchanged files. Need to tell the user that this feedback might be stale.
For changed files: figure out how to map previously profiled lines to their new locations in the file editor. Idea: instead of line numbers, use a hash of the line's contents. This can be used a lookup table by the client to annotate lines. Problem: if a line looks the same, but has different performance, this won't work.
Another idea: figure out the difference between the branch and the last version that we have. Adjust line numbers according to the inserted lines.
As soon as a new commit is made, a new commit ID means that the metric backend had no data for that particular version. Figure out how to incorporate old metrics data.
Idea: traverse back from current version ID until we find one which exists in the DB. Use this to fetch and display feedback.
This will work for unchanged files. Need to tell the user that this feedback might be stale.
For changed files: figure out how to map previously profiled lines to their new locations in the file editor. Idea: instead of line numbers, use a hash of the line's contents. This can be used a lookup table by the client to annotate lines. Problem: if a line looks the same, but has different performance, this won't work.
Another idea: figure out the difference between the branch and the last version that we have. Adjust line numbers according to the inserted lines.