Change Details

See PHI1104. The older "differential.querydiffs" method includes the entire raw diff text for all the diffs associated with a revision in its response, but we: only care about the most recent diff; and don't care about the text at all. For reasonable large changes with several updates, this can be significantly slow. We can get this same information more efficiently from the modern "differential.diff.search", since D19386 (April 2018). The only trick is that we need a "revisionPHID", which we don't have on hand. For now, just fetch the revision PHID. In the future, we can likely make adjustments so that we have the revision PHID already by the time we get here. This may slow down the normal case very slightly (since we now do two calls instead of one), but it speeds up the bad cases dramatically.

See PHI1104. The older "differential.querydiffs" method includes the entire raw diff text for all the diffs associated with a revision in its response, but we: only care about the most recent diff; and don't care about the text at all. For reasonably large changes with several updates, this can be significantly slow. We can get this same information more efficiently from the modern "differential.diff.search", since D19386 (April 2018). The only trick is that we need a "revisionPHID", which we don't have on hand. For now, just fetch the revision PHID. In the future, we can likely make adjustments so that we have the revision PHID already by the time we get here. This may slow down the normal case very slightly (since we now do two calls instead of one), but it speeds up the bad cases dramatically.

See PHI1104. The older "differential.querydiffs" method includes the entire raw diff text for all the diffs associated with a revision in its response, but we: only care about the most recent diff; and don't care about the text at all. For reasonabley large changes with several updates, this can be significantly slow. We can get this same information more efficiently from the modern "differential.diff.search", since D19386 (April 2018). The only trick is that we need a "revisionPHID", which we don't have on hand. For now, just fetch the revision PHID. In the future, we can likely make adjustments so that we have the revision PHID already by the time we get here. This may slow down the normal case very slightly (since we now do two calls instead of one), but it speeds up the bad cases dramatically.